SOLAR RESOURCE AND PV POTENTIAL OF MALAWI 24 MONTH SITE RESOURCE REPORT December 2018 This report was prepared by Solargis, under contract to the World Bank. Solar Power Resource Mapping: Malawi [Project ID: P151289]. This activity is funded and supported by the Energy Sector Management Assistance Program (ESMAP), a multi-donor trust fund administered by the World Bank, under a global initiative on Renewable Energy Resource Mapping. Further details on the initiative can be obtained from the ESMAP website. The content of this document is the sole responsibility of the consultant authors. Any improved or validated solar resource data will be incorporated into the Global Solar Atlas. Copyright © 2018 THE WORLD BANK Washington DC 20433 Telephone: +1-202-473-1000 Internet: www.worldbank.org The World Bank does not guarantee the accuracy of the data included in this work and accept no responsibility for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because the World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for non-commercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: +1-202-522-2625; e-mail: pubrights@worldbank.org. Furthermore, the ESMAP Program Manager would appreciate receiving a copy of the publication that uses this publication for its source sent in care of the address above, or to esmap@worldbank.org. All images remain the sole property of their source and may not be used for any purpose without written permission from the source. Attribution Please cite the work as follows: World Bank. 2018. Solar resource and PV potential of Malawi: 24 Month Site Resource Report. Washington, DC: World Bank. Annual Solar Resource Report for solar meteorological stations after completion of 24 months of measurements Republic of Malawi Reference No. 141-07/2018 Date: 3 December 2018 Customer Consultant World Bank Solargis s.r.o. Energy Sector Management Assistance Program Contact: Mr. Marcel Suri Contact: Mr. Dhruva Sahai Mytna 48, 811 07 Bratislava, Slovakia 1818 H St NW, Washington DC, 20433, USA Phone +421 2 4319 1708 Phone: +1-202-473-3159 E-mail: marcel.suri@solargis.com E-mail: mailto: dsahai@worldbank.org http://solargis.com http://www.esmap.org/RE_Mapping Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 TABLE OF CONTENTS Table of contents ............................................................................................................................................. 4 Acronyms ........................................................................................................................................................ 6 Glossary .......................................................................................................................................................... 7 1 Introduction............................................................................................................................................. 9 1.1 Background................................................................................................................................................ 9 1.2 Delivered data sets .................................................................................................................................... 9 1.3 Information included in this report ......................................................................................................... 10 1.4 Benefits of the measurement campaign ............................................................................................... 10 2 Position of solar meteorological sites.................................................................................................... 11 3 Ground measurements .......................................................................................................................... 13 3.1 Instruments and measured parameters ................................................................................................ 13 3.2 Station operation and calibration of instruments ................................................................................. 14 3.3 Quality control of measured solar resource data.................................................................................. 16 3.3.1 Chileka ....................................................................................................................................... 16 3.3.2 Kasungu ..................................................................................................................................... 19 3.3.3 Mzuzu ........................................................................................................................................ 22 3.4 Recommendations on the operation and maintenance of the sites .................................................... 25 3.5 Instruments re-calibration at the end of measurement campaign ...................................................... 26 4 Solar resource model data ..................................................................................................................... 28 4.1 Solar model .............................................................................................................................................. 28 4.2 Site adaptation of the solar model: Method .......................................................................................... 29 4.3 Results of the model adaptation at three sites ..................................................................................... 31 5 Meteorological model data .................................................................................................................... 37 5.1 Meteorological model ............................................................................................................................. 37 5.2 Validation of meteorological data .......................................................................................................... 37 5.2.1 Air temperature at 2 meters ..................................................................................................... 38 5.2.2 Relative humidity ....................................................................................................................... 40 5.2.3 Wind speed and wind direction at 10 meters .......................................................................... 42 5.3 Uncertainty of meteorological model data ............................................................................................ 44 6 Solar resource: uncertainty of long-term estimates ............................................................................... 45 6.1 Uncertainty of solar resource yearly estimate....................................................................................... 45 6.2 Uncertainty due to interannual variability of solar radiation................................................................. 46 6.3 Combined uncertainty ............................................................................................................................. 47 7 Time series and Typical Meteorological Year data ................................................................................. 50 7.1 Delivered data sets .................................................................................................................................. 50 7.2 TMY method ............................................................................................................................................ 51 7.3 Results ..................................................................................................................................................... 51 8 Conclusions .......................................................................................................................................... 56 Annex 1: Site related data statistics ............................................................................................................... 57 Yearly summaries of solar and meteorological model outputs ....................................................................... 57 Monthly summaries of solar and meteorological model outputs .................................................................... 59 © 2018 Solargis page 4 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Frequency of occurrence of GHI and DNI daily model values for a period 1994 to 2017 .............................. 61 Frequency of occurrence of GHI and DNI 15-minute model values for a period 1994 to 2017 ..................... 64 Frequency of occurrence of measured and model GHI and DNI during two years of measurements .......... 67 Frequency of occurrence of GHI and DNI ramps .............................................................................................. 69 List of figures ................................................................................................................................................ 72 List of tables .................................................................................................................................................. 74 References .................................................................................................................................................... 76 Support information ....................................................................................................................................... 78 © 2018 Solargis page 5 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 ACRONYMS AP Atmospheric Pressure CFSR Climate Forecast System Reanalysis. The meteorological model operated by the US service NOAA (National Oceanic and Atmospheric Administration) CFS v2 Climate Forecast System Version 2 CFSv2 model is the operational extension of the CFSR (NOAA, NCEP) DIF Diffuse Horizontal Irradiation, if integrated solar energy is assumed. Diffuse Horizontal Irradiance, if solar power values are discussed DNI Direct Normal Irradiation, if integrated solar energy is assumed. Direct Normal Irradiance, if solar power values are discussed. GFS Global Forecast System. The meteorological model operated by the US service NOAA (National Oceanic and Atmospheric Administration) GHI Global Horizontal Irradiation, if integrated solar energy is assumed. Global Horizontal Irradiance, if solar power values are discussed. MACC Monitoring Atmospheric Composition and Climate – meteorological model operated by the European service ECMWF (European Centre for Medium-Range Weather Forecasts) Meteosat MFG Meteosat satellite operated by EUMETSAT organization. MSG: Meteosat Second and MSG Generation; MFG: Meteosat First Generation PWAT Precipitable water (water vapour) QC Quality control RH Relative Humidity at 2 metres TEMP Air Temperature at 2 metres WD Wind Direction at 10 metres WS Wind Speed at 10 metres © 2018 Solargis page 6 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 GLOSSARY Aerosols Small solid or liquid particles suspended in air, for example desert sand or soil particles, sea salts, burning biomass, pollen, industrial and traffic pollution. All-sky irradiance The amount of solar radiation reaching the Earth's surface is mainly determined by Earth-Sun geometry (the position of a point on the Earth's surface relative to the Sun which is determined by latitude, the time of year and the time of day) and the atmospheric conditions (the level of cloud cover and the optical transparency of atmosphere). All-sky irradiance is computed with all factors taken into account Bias Represents systematic deviation (over- or underestimation) and it is determined by systematic or seasonal issues in cloud identification algorithms, coarse resolution and regional imperfections of atmospheric data (aerosols, water vapor), terrain, sun position, satellite viewing angle, microclimate effects, high mountains, etc. Clear-sky irradiance The clear sky irradiance is calculated similarly to all-sky irradiance, but without considering the impact of cloud cover. Long-term average Average value of selected parameter (GHI, DNI, etc.) based on multiyear historical time series. Long-term averages provide a basic overview of solar resource availability and its seasonal variability. P50 value Best estimate or median value represents 50% probability of exceedance. For annual and monthly solar irradiation summaries it is close to average, since multiyear distribution of solar radiation resembles normal distribution. P90 value Conservative estimate, assuming 90% probability of exceedance (with the 90% probability the value should be exceeded). When assuming normal distribution, the P90 value is also a lower boundary of the 80% probability of occurrence. P90 value can be calculated by subtracting uncertainty from the P50 value. In this report, we apply a simplified assumption of normal distribution of yearly values. Root Mean Square Represents spread of deviations given by random discrepancies between measured and Deviation (RMSD) modelled data and is calculated according to this formula: 6 * ∑7 (* +,-./0,1 − +31,4,1 5 = & *89 On the modelling side, this could be low accuracy of cloud estimate (e.g. intermediate clouds), under/over estimation of atmospheric input data, terrain, microclimate and other effects, which are not captured by the model. Part of this discrepancy is natural - as satellite monitors large area (of approx. 3.3 x 3.4 km for MSG satellite), while sensor sees only micro area of approx. 1 sq. centimeter. On the measurement side, the discrepancy may be determined by accuracy/quality and errors of the instrument, pollution of the detector, misalignment, data loggers, insufficient quality control, etc. Solar irradiance Solar power (instantaneous energy) falling on a unit area per unit time [W/m2]. Solar resource or solar radiation is used when considering both irradiance and irradiation. Solar irradiation Amount of solar energy falling on a unit area over a stated time interval [Wh/m2 or kWh/m2]. © 2018 Solargis page 7 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Uncertainty Is a parameter characterizing the possible dispersion of the values attributed to an of estimate, Uest estimated irradiance/irradiation values. In this report, uncertainty assessment of the solar resource model estimate is based on a detailed understanding of the achievable accuracy of the solar radiation model and its data inputs (satellite, atmospheric and other data), which is confronted by an extensive data validation experience. The second source of uncertainty is ground measurements. Their quality depends on accuracy of instruments, their maintenance and data quality control. Third contribution to the uncertainty is from the site adaptation method where ground-measured and satellite- based data are correlated. © 2018 Solargis page 8 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 1 INTRODUCTION 1.1 Background This report is prepared within Phase 2 of the project Renewable Energy Resource Mapping for the Republic of Malawi. This project focuses on solar resource mapping and measurement services as part of a technical assistance in the renewable energy development implemented by the World Bank in Malawi. It is being undertaken in close coordination with the Ministry of Natural Resources, Energy and Mining (MoNREM) of Malawi, the World Bank’s primary country counterpart for this project, and Malawi Meteorological Services (MMS). This project is funded by the Energy Sector Management Assistance Program (ESMAP), administered by the World Bank and supported by bilateral donors. This report summarizes results of two years of the measuring campaign at three solar meteorological stations installed in Malawi as part of the World Bank’s ESMAP mission in Malawi. This report describes delivery of site-specific solar resource and meteorological data for three sites, where solar meteorological stations have been operated. As a result of high-quality operation of the meteorological sites and site adaptation of the Solargis model, reliable historical time series and TMY data is computed. The delivered time series and Typical Meteorological Year data is ready to use for bankable evaluation of solar energy projects. The report also describes the results quality control of site-specific measurements, methodology of site adaptation of the solar model, and related data uncertainties. The measurements at three sites are provided by GeoSUN Africa company (South Africa). The model data for the same sites and related works, together with this report are supplied by Solargis (Slovakia). 1.2 Delivered data sets The site-specific data, provided as part of this delivery, include: • Solar and meteorological ground measurements, after second level quality assessment (first level delivered by GeoSUN Africa), representing period 03/2016 to 03/2018 • Time series calculated by the solar model, representing last 24+ years (01/1994 to 03/2018) • Typical Meteorological Year (TMY) data, representing last 24 calendar years (01/1994 to 12/2017) The data is delivered in formats ready to use in energy simulation software. This report provides detailed insight of the methodologies and results. Table 1.1 Delivered data characteristics Feature Time coverage Primary time step Delivered files Measurements Mar 2016 to Mar 2018 1 minute Quality controlled measurements (GeoSUN Africa) 1- minute Model data – original Jan 1994 to Mar 2018 15 minutes Time series – hourly (Solargis) Time series – monthly Time series – yearly Model data – site adapted Jan 1994 to Mar 2018 15 minutes Time series – hourly (Solargis) Time series – monthly Time series – yearly Model data – site adapted Jan 1994 to Dec 2017 hourly Typical Meteorological Year P50 (Solargis) Typical Meteorological Year P90 © 2018 Solargis page 9 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 1.2 Parameters included in the site-adapted time series and TMY data (hourly time step) Parameter Acronym Unit TS TMY P50 TMY P90 Global horizontal irradiance GHI W/m2 X X X Direct normal irradiance DNI W/m2 X X X Diffuse horizontal irradiance DIF W/m2 X X X Global tilted irradiance (at optimum angle) GHI W/m2 X - - Solar azimuth SA ° X X X Solar elevation SE ° X X X Air temperature at 2 meters TEMP °C X X X Wind speed at 10 meters WS m/s X X X Wind direction at 10 meters WD ° X X X Relative humidity RH % X X X Air Pressure AP hPa X X X Precipitable Water PWAT kg/m2 X X X 1.3 Information included in this report This report presents: • Solar resource and meteorological measurements after 24 months of operation o Review and quality check of the measured data o Calibration procedures and results o List and explanation of the occurred disturbances and failures • Comparison of the measurements to the model time series and uncertainty analysis o Comparison of solar and meteo measurements with the model data o Site adaptation of satellite data using the ground measurements and uncertainty estimate o Estimate of data uncertainty • Data analysis (measured vs. modelled) o Monthly summaries of solar and meteorological parameters captured at the site o Variability of measured solar parameters o Frequency of occurrence of GHI and DNI 1-minute and 15-minute values o Frequency of occurrence of GHI and DNI 1-minute and 15-minute ramps 1.4 Benefits of the measurement campaign The solar meteorological stations helped to reveal strong solar weather patterns. The accurate data helped to identify weakness of solar models, especially of the part that relates to satellite-data processing. The data will be used for re-calibration of the models to better represent this type of solar tropical climate. The collected data also improves the knowledge of solar variability due to intermittent clouds. © 2018 Solargis page 10 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 2 POSITION OF SOLAR METEOROLOGICAL SITES In Malawi, three measuring stations have been installed within the World Bank’s ESMAP Solar Resource Mapping initiative. Two of them are located at the airports (airport Chileka and the airstrip Kasungu), and one at the University in Mzuzu (Figure 2.1, Table 2.1). Figure 2.1: Position of solar meteorological stations in Malawi Table 2.1 Overview information on the solar meteorological station locations Site location Nearest town Latitude [º] Longitude [º] Altitude [m a.s.l.] Measurement station host Chileka airport Blantyre -15.67984 34.97229 767 Malawi Meteorological Services Kasungu airport Kasungu -13.01530 33.46840 1065 Malawi Meteorological Services Uni Mzuzu Mzuzu -11.41990 33.99530 1285 Mzuzu University © 2018 Solargis page 11 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Besides the good geographical distribution within the territory of Malawi, the sites fit well to the population centres, where solar installations can be primarily deployed. In addition to geographical and socio-economic criteria, the sites fulfil the criteria for the operation and maintenance of the solar measuring stations: • Availability of free horizon (horizon is partially shaded in Mzuzu, Figures 2.2 to 2.4) • Availability of GSM networks, • Availability of local work force for maintenance, • Easy to access and high level of security During the two years of the measurements, the measured data was analyzed and harmonized with the objective to acquire reference solar radiation data for reducing the uncertainty of the model (Chapter 6). The quality data from Tier-1 and Tier-2 meteorological stations is available from this assessment (Chapter 3). Position and detailed information about measurement sites is available also on Global Solar Atlas website: http://globalsolaratlas.info/?c=-12.811801,34.541016,7&e=1 Figure 2.2: Horizon at Chileka meteorological station Figure 2.3: Horizon at Kasungu meteorological station Figure 2.4: Horizon at Mzuzu meteorological station © 2018 Solargis page 12 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 3 GROUND MEASUREMENTS 3.1 Instruments and measured parameters Basic information about the measurements sites is in Table 3.1. The solar parameters are measured by high accuracy instruments (CMP 10 for measurements of GHI and CHP1 for measurements of DNI) and by medium accuracy instruments (RSR for GHI, DNI and DIF) (Table 3.2 and 3.3). The measurement campaign in Malawi has being undertaken by company GeoSUN Africa from South Africa. Table 3.1 Overview information on measurement stations operated in the region ID Site name Station type Installation date 1 Chileka Tier 1 18 March 2016 2 Kasungu Tier 2 11 March 2016 3 Mzuzu Tier 2 14 March 2016 Table 3.2 Instruments installed at the solar meteorological stations Site name GHI DIF DNI WS WD TEMP RH AP PWAT Chileka CMP 10 CMP 10 CHP 1 05103 05103 HMP155 HMP155 PTB110 TB4 CMP 10 - - Kasungu 014A 024A HMP155 HMP155 PTB110 TB4 RSR 2 RSR 2 RSR 2 CMP 10 - - Mzuzu 014A 024A HMP155 HMP155 PTB110 TB4 RSR 2 RSR 2 RSR 2 Table 3.3 Technical parameters and accuracy class of the instruments at Tier 1 and Tier 2 stations Parameter Instrument Type Manufacturer Uncertainty GHI Secondary standard pyranometer CMP 10 Kipp & Zonen < ±2 % (daily) DIF Secondary standard pyranometer CMP 10 Kipp & Zonen < ±2 % (daily) DNI Pyrheliometer CHP 1 Kipp & Zonen < 1 % (daily) GHI 2 Rotating Shadowband Radiometer with LI200 RSR 2 Irradiance Inc. Indicatively ±5 % DIF 2 Rotating Shadowband Radiometer with LI200 RSR 2 Irradiance Inc. Indicatively ±8 % DNI 2 Rotating Shadowband Radiometer with LI200 RSR 2 Irradiance Inc. Indicatively ±5 % WS Tier 1 station wind speed sensor (at 10 m) 05103 R.M. Young ±0.3 m/s Tier 2 station wind speed sensor (at 3 m) 014A Met One ±1.5 % WD Tier 1 station wind direction sensor (at 10 m) 05103 R.M. Young ±3 ° Tier 2 station wind direction sensor (at 3 m) 024A Met One ±5° TEMP Temperature probe (at 2 m) HMP 155 Vaisala ±0.45°C RH Relative humidity probe HMP 155 Vaisala ±1.7% RH AP Barometric pressure sensor PTB110 Vaisala ±1.5 hPa PWAT Tipping-bucket rain gauge TB4 Hydrological services ±3% ± (0.06% of reading - Data logger CR1000 Campbell Scientific + offset) © 2018 Solargis page 13 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 3.2 Station operation and calibration of instruments In this report, the solar and meteorological measurements during the two-year measurement campaign are analyzed. As the measurement stations have been installed during March 2016, the period considered for the data analysis represents more than 24 months – until 31 March 2018 for all three stations. Overview of the data availability, the time step and measurement period is shown in Tables 3.4 and 3.5. Table 3.4 Overview information on solar meteorological stations operating in the region Site name Campaign measurement period Primary time step Chileka 19 March 2016 – 31 March 2018 1 minute Kasungu 18 March 2016 – 31 March 2018 1 minute Mzuzu 18 March 2016 – 31 March 2018 1 minute Table 3.5 Period of measurements analyzed in this report Year, month 2016 2017 2018 Station 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 Chileka Kasungu Mzuzu Table 3.6 shows data recovery statistics for the whole measurement period for each station. In this statistics, only serious issues (missing data for a longer period, erroneous data – tracking malfunction, serious soiling due to insufficient cleaning or missing data) are accounted. Short-term operational issues (shading by surrounding objects, morning dew on the instrument, etc.) or specific features of the individual sites (systematic shading at the Mzuzu station in the morning and afternoon) are not considered. All stations fulfilled criteria of 2-years availability of high-quality ground measurements. Periods of missing or erroneous data were shorter than required criteria. The data exceeding the two-year measurement period partly recover the data loss. Table 3.6 Data recovery statistics of the measurement campaign Malawi Data loss* Influenced days** Exceeding Acceptance criteria Measurement period Erroneous + Missing data Total Length of individual periods two years *** 95% 15+ days 03/2016-03/2018 (%) (days) Description (days) (days) (days) insufficient cleaning, tracker issues, Chileka 4.2 31 29 1,2,2,4,1,5,6,1,5,1,1 13 OK OK missing data Kasungu 0.6 4.5 missing data 4 1,3 14 OK OK Mzuzu 0 0 - 0 - 14 OK OK The column Data loss* of the table above represents amount of missing data or data excluded during quality control process. Percentage share is calculated from daytime values and days represent cumulative amount of missing data (one day may be composed from several shorter missing data periods). The column Influenced days** represents number of days with fully or partially missing data or days excluded by quality control process. The column Exceeding two years*** shows count of days of measurements exceeding the two years period. During the solar measuring campaign, local staff of Malawi Meteorological Services was trained by GeoSUN Africa, and they executed the instruments inspection and cleaning visits, typically in 1-7 days intervals. GeoSUN Africa performed detailed visits and station maintenance every six months of the measurement campaign (Table 3.7). © 2018 Solargis page 14 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 3.7 Meteorological stations maintenance Chileka Comments and issues Station type Tier 1 • Battery connection issue in July 2016 caused measured data loss in 11 - 14 July 2016. Instruments cleaning interval Average: 4.8 • Several events of pyrheliometer or Solys tracker [days] Longest: 31 misalignment occurred during campaign causing inaccurate DNI or DIF measurements (April 2016, July – Verification visits date 10 Oct 2016 October 2016, December 2016 – February 2017, June 3 Apr 2017 2017, July 2017). Incorrect values were replaced by 27 Oct 2017 calculated ones. 4 Sep 2018 • A solar eclipse on 1 September 2016 caused a dip in the solar radiation data. Instruments field verification GHI – reference CMP11 • No cleaning was performed in December 2016 DIF – reference CMP21 • Several events of morning dew caused distortion of DNI DNI – reference CHP1 measured values (March 2017, August – November 2017). • Inaccurate values of TEMP measured on 4 December 2017, probably due to storm. Kasungu Comments and issues Station type Tier 2 • Several events of morning dew caused distortion of GHI measured values (April – June 2016, February – Instruments cleaning interval Average: 1.8 November 2017). [days] Longest: 5 • Incorrect measurements of RSR due to cable issues (July and September 2016) Verification visits date 11 Oct 2016 • A solar eclipse on 1 September 2016 caused a dip in the 5 Apr 2017 solar radiation data. 22 Oct 2017 • Shading of the instruments by trees during the early 6 Sep 2018 morning and late evening Instruments field verification GHI – reference CMP11 Mzuzu Comments and issues Station type Tier 2 • Shading on the instruments during the early mornings and late afternoons caused by trees. Instruments cleaning interval Average: 1.6 • Events of morning dew caused distortion of GHI [days] Longest: 7 measured values. • A solar eclipse on 1 September 2016 caused a dip in the Verification visits date 13 Oct 2016 solar radiation data. 6 Apr 2017 • False readings of the rain gauge occurred throughout 24 Oct 2017 the month. 0 mm rain recorded is correct, but any value 7 Sep 2018 measured is assumed to be incorrect and replaced with "Empty" (June – October 2016). Instruments field verification GHI – reference CMP11 Instruments field verification, i.e. comparative measurements of solar radiation parameters and cross check with the reference instruments was performed by GeoSUN Africa after one year of operation. Instruments field verification proved that sensitivity (calibration constants) remained stable within the instrument specifications, At the end of measurement campaign, the verification, calibration and instrument replacement was performed on all installed sensors. These tasks were provided by GeoSUN Africa, procedures are described in detail in Chapter 3.5. © 2018 Solargis page 15 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 3.3 Quality control of measured solar resource data Prior to correlation with satellite-based solar data, the ground-measured solar radiation was quality-controlled by company Solargis. Quality control (QC) was based on methods defined in SERI QC procedures and Younes et al. [1, 2] and the in-house developed tests. The ground measurements were also inspected visually, mainly for identification of shading, effect of dew and other error patterns. Figures 3.1 to 3.10 show results of quality control for individual stations. The colors in Figures 3.1, 3.4 and 3.7 indicate the following flags: • Green: data passed all tests • Blue: data excluded by visual inspection - mainly shading • Grey: sun below horizon • White and brown strips: missing data • Red: GHI, DNI and DIF consistency problem • Violet and cyan: problems with physical limits The data records not passing the quality control test were flagged and excluded from the further processing. The results show various amount of excluded data readings (Tables 3.9, 3.12, 3.15) around 1% for Kasungu station, around 5.5% for Chileka station mainly due to some periods with insufficient cleaning, and around 14% for Mzuzu station, where increased shading early morning and late afternoon was identified. 3.3.1 Chileka Table 3.8 Occurrence of data readings for Chileka meteorological station Data availability DNI CHP1 GHI CMP10 Sun below horizon 529 254 49.5% 529 254 49.5% Sun above horizon 539 475 50.5% 539 475 50.5% Total data readings 1 068 729 100.0% 1 068 729 100.0% Table 3.9 Excluded ground measurements after quality control (Sun above horizon) in Chileka Occurrence of data samples (Sun above horizon) Type of test DNI CHP1 GHI CMP10 Physical limits test 0 0.0% 2 551 0.5% Consistency test (GHI – DNI – DIF) 14 998 2.8% 14 998 2.8% Visual test (incorrect data) 13 333 2.5% 12 852 2.4% Other (non valid data) 1 180 0.2% 101 0.0% Total excluded data samples 29 511 5.5% 30 502 5.7% Total samples 539 475 100.0% 539 475 100.0% © 2018 Solargis page 16 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.1 Results of GHI and DNI quality control in Chileka. Green – data passing all tests; grey – sun below horizon; red – consistency issue, violet – physical limit, blue excluded by visual inspection. Top: DNI (CHP1); bottom: GHI (CMP10) Figure 3.2 Insufficient cleaning – skewed DIF and DNI profiles. Blue: DNI CHP1; green: GHI CMP 10; red: GHI CMP10 © 2018 Solargis page 17 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.3 Comparison of solar irradiance: measured and calculated. Left: DNI; Right: GHI; X-axis: measured data; Y-axis: calculated data from complementary components Main findings: • Several periods of inconsistency between independent GHI-DNI and DIF measurements is present in the data (Figure 3.2) mainly in the first year of measurements. This might be a result of insufficient cleaning. The Malawi Meteorological Department took measures to improve the cleaning schedule. • Early morning small shading from surrounding objects. This cannot be changed. • Inconsistency between the measured and calculated DNI and GHI (DNI from GHI and DIF, GHI from DNI and DIF). Both comparisons (Figure 3.3) show increased spread of compared data values. The overall bias is acceptable. The measured DNI is in average lower by -0.7% than its calculated counterpart. On contrary the measured GHI is higher by +0.8% then its calculated counterpart. Table 3.10 Quality control summary - Chileka Description Station description, Installation report available metadata Instrument accuracy 2 x Secondary standard pyranometer CMP10 (GHI, DIF) 1 x First class pyrheliometer CHP1 (DNI) Instrument calibration Instruments were calibrated Data structure Clear Cleaning and maintenance Cleaning log available information Several periods with insufficient cleaning identified Time reference Correct and clear time reference Quality control complexity Full quality control tests applied including (GHI-DNI-DIF) consistency test Quality control results Several periods with degraded measurements due to insufficient cleaning Period More than 24 months from which ca 5.5% (DNI) and 5.7% (GHI) was excluded by QC Other issues Legend: Quality marker Very good Good Sufficient Problematic Insufficient Not specified © 2018 Solargis page 18 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 3.3.2 Kasungu Table 3.11 Occurrence of data readings for Kasungu meteorological station Data availability GHI CMP10 GHI, DNI RSR 2 Sun below horizon 530 140 49.5% 530 140 49.5% Sun above horizon 540 026 50.5% 540 026 50.5% Total data readings 1 070 166 100.0% 1 070 166 100.0% Table 3.12 Excluded ground measurements after quality control (Sun above horizon) in Kasungu Occurrence of data samples (Sun above horizon) Type of test GHI CMP10 DNI RSR 2 GHI RSR 2 Physical limits test 3 542 0.7% 5 0.0% 2 602 0.5% Consistency test (GHI – DNI – DIF) - - 0 0.0% 0 0.0% Visual test (incorrect data) 3 110 0.6% 0 0.0% 90 0.0% Other (non valid data) 17 0.0% 2 000 0.4% 2 000 0.4% Total excluded data samples 6 669 1.2% 2 005 0.4% 4 692 0.9% Total samples 540 026 100.0% 540 026 100.0% 540 026 100.0% © 2018 Solargis page 19 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.4 Results of GHI and DNI quality control in Kasungu. Green – data passing all tests; grey – sun below horizon; violet and cyan – physical limit issue, blue excluded by visual inspection; brown – missing data. Top: DNI (RSR 2); middle: GHI (RSR 2); bottom: GHI (CMP10) Main findings: • Effect of dew on CPM10 instrument in the morning hours (Figure 3.5). • Short periods of discrepancy between GHI measured by the CMP10 and RSR 2. Overall difference is 1.4% (data from CMP10 is higher, Figure 3.6), but occasionally the difference is at the level of 3 to 4% (excluding the effect of dew, when the effect is much higher). © 2018 Solargis page 20 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.5 Effect of dew – degraded GHI CMP10 (yellow) readings. Blue: DNI RSR 2; yellow: GHI CMP 10; red: GHI RSR 2; green: DIF RSR 2; dashed: cleaning Figure 3.6 Difference between GHI from CMP10 and RSR 2 - Kasungu. Table 3.13 Quality control summary - Kasungu Description Station description, metadata Installation report available Instrument accuracy Secondary standard pyranometer CMP10 (GHI) Rotating Shadowband Radiometer RSR 2 (GHI, DIF, DNI) Instrument calibration Instruments were calibrated Data structure Clear Cleaning and maintenance Cleaning log available information Diligent cleaning Time reference Correct and clear time reference Quality control complexity RSR 2 data, full QC CMP10 data, without (GHI-DNI-DIF) consistency test, compared to GHI from RSR 2 Quality control results Small issues with degradation of GHI from CMP10 due to morning dew Small issues with early morning shading Period More than 24 months from which less than 1.2% (GHI CMP10) and 0.4% (DNI RSR 2) was excluded by QC Other issues Legend: Quality marker Very good Good Sufficient Problematic Insufficient Not specified © 2018 Solargis page 21 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 3.3.3 Mzuzu Table 3.14 Occurrence of data readings for Mzuzu meteorological station Data availability GHI CMP10 GHI, DNI RSR 2 Sun below horizon 530 116 49.5% 530 116 49.5% Sun above horizon 540 173 50.5% 540 173 50.5% Total data readings 1 070 289 100.0% 1 070 289 100.0% Table 3.15 Excluded ground measurements after quality control (Sun above horizon) in Mzuzu Occurrence of data samples (Sun above horizon) Type of test GHI CMP10 DNI RSR 2 GHI RSR 2 Physical limits test 5 158 1.0% 0 0.0% 2 746 0.5% Consistency test (GHI – DNI – DIF) - - 0 0.0% 0 0.0% Visual test (incorrect data) 72 632 13.4% 71 927 13.3% 70 220 13.0% Other (non valid data) 0 0.0% 35 0.0% 35 0.0% Total excluded data samples 77 790 14.4% 71 962 13.3% 73 001 13.5% Total samples 540 173 100.0% 540 173 100.0% 540 173 100.0% © 2018 Solargis page 22 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.7 Results of GHI and DNI quality control − Mzuzu. Green – data passing all tests; grey – sun below horizon; violet and cyan – physical limit issue, blue excluded by visual inspection. Top: DNI (RSR 2); middle: GHI (RSR 2); bottom: GHI (CMP10) Main findings: • Systematic early morning and late afternoon shading from surrounding objects (Figure 3.8). • Effect of dew on CPM10 instrument in morning hours (Figure 3.9). • Short periods of discrepancy between GHI measured by the CMP10 and RSR 2. Overall difference is small (approaching 0.9%, Figure 3.10), but occasionally the difference is at the level of 4 to 5% (excluding the effect of dew, when the effect is much higher). There are trees around the plot that trigger local shading in the morning and evening hours. The affected data was removed from further analysis. The data for Mzuzu University has higher uncertainty, compared to Chileka and Kasungu. Shading can be resolved only by moving of the station to the location less affected. As shown in the report, the Mzuzu data is very helpful for understanding the tropical microclimate influenced by large lake and by mountains. © 2018 Solargis page 23 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.8 Systematic shading – drop of DNI in Mzuzu. Blue: shaded DNI; red: Unshaded DNI Figure 3.9 Effect of dew – degraded GHI CMP10 readings. Blue: DNI RSR 2; yellow: GHI CMP 10; red: GHI RSR 2; green: DIF RSR 2; dashed: cleaning Figure 3.10 Difference between GHI from CMP10 and RSR 2 − Mzuzu. © 2018 Solargis page 24 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 3.16 Quality control summary - Mzuzu Description Station description, Installation report available metadata Instrument accuracy Secondary standard pyranometer CMP10 (GHI) Rotating Shadowband Radiometer RSR 2 (GHI, DIF, DNI) Instrument calibration Instruments were calibrated Data structure Clear Cleaning and maintenance Cleaning log available information Diligent cleaning, lower frequency in November 2016 Time reference Correct and clear time reference Quality control complexity RSR 2 data, full QC CMP10 data, without (GHI-DNI-DIF) consistency test, compared to GHI from RSR 2 Quality control results Higher occurrence of degradation of GHI from CMP10 due to morning dew Systematic late afternoon and early morning shading from surrounding objects Period More than 24 months from which more than 14.4% (GHI CMP10) and 13.3 (DNI RSR 2) was excluded by QC Other issues Legend: Quality marker Very good Good Sufficient Problematic Insufficient Not specified 3.4 Recommendations on the operation and maintenance of the sites Based on the results of quality control (Tables 3.10, 3.13 and 3.16), we conclude that the solar radiation measurements come from the high accuracy (regarding the CMP10 and CHP1 instruments) and medium accuracy (RSR 2 instrument) equipment that is professionally operated and maintained. Some issues were identified during the data quality control: • All meteorological stations: Effect of morning dew condensation is seen on CMP10 instrument measurements (to some extent present at all sites). These data values were flagged and excluded from further processing. We do not recommend application of any measures to deal with dew for the existing sites. Yet, for future measuring campaigns, it is advised to consider installation of ventilation units for pyranometers to reduce data affected by morning dew. • Chileka: Insufficient cleaning results in incorrect GHI, DNI and DIF measurements during several periods. These data were flagged and excluded from further processing. We recommend improving cleaning frequency. • Mzuzu: Systematic early morning and late afternoon shading from surrounding objects (trees). The data were flagged and excluded from further processing. This is not optimum configuration, but acceptable. Choice of locating the meteo station in the campus of the Mzuzu University has been made as it fits the education curriculum in renewable energy at the University and it also creates additional synergies. The position of the station helped to reveal strong solar weather patterns affected by the presence of mountains and large lake. The data measured contributed significantly to getting better knowledge of the patterns of tropical solar climate. © 2018 Solargis page 25 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 3.5 Instruments re-calibration at the end of measurement campaign At the end of measurement campaign, all three solar meteorological stations were visited by GeoSUN Africa. Solar measurement sensors (pyrheliometer, thermopile pyranometers and silicon pyranometers) were swopped with newly calibrated instruments. The thermopile and silicon pyranometers were replaced with calibrated thermopile pyranometers and new calibrated silicon pyranometers. The thermopile pyranometers were calibrated according to ISO 9847 standard by GeoSUN Africa. The pyrheliometer was replaced with pyrheliometer calibrated according to ISO 9059 standard. All solar measurement instruments were cleaned, levelled and aligned, and the new calibration multipliers were updated on the datalogger program. Summary of replaced instruments and works performed on all sites is available in Table 3.17. Barometer, thermometer and hygrometer were side-by-side compared with a calibrated reference instruments. The temperature and humidity sensors were cleaned and reinstalled. Verification was done for barometric pressure by comparing the field instrument reading with the reference reading at ambient conditions. The functionality of the wind instruments was verified using an anemometer drive (for wind speed). The cups or propellers (depending on the type of anemometer) were removed from the anemometers after which the shaft was coupled to an anemometer drive. The drive was set on six fixed speeds for respective periods of time. The wind direction was tested with a four point verification test (the wind vane was pointed in the four main wind directions by hand). Table 3.17 Summary of replaced solar instruments Manufacturer, Station Parameter Calibration date Comments, other works on site Instrument type Kasungu Kipp & Zonen CMP10 GHI 31 August 2018 • The gate entry switch was repaired LI-COR LI-200 GHI, DNI 8 May 2018 • Solar instrument cables were neatened with cable ties • 5 new microfiber cloths were left for the cleaner Mzuzu Kipp & Zonen CMP10 GHI 31 August 2018 • Station solar panel was cleaned • Solar instrument cables were neatened with LI-COR LI-200 GHI, DNI 8 May 2018 cable ties • 5 new microfiber cloths were left for the cleaner Chileka Kipp & Zonen CMP11 GHI 23 July 2018 • The sun tracker was cleaned and re-aligned • Solar instrument cables were neatened with Kipp & Zonen CMP11 DIF 23 July 2018 cable ties Kipp & Zonen CHP1 DNI 29 August 2018 • Datalogger enclosure battery was replaced with a new battery • 5 new microfiber cloths were left for the cleaner to be used to clean the pyranometer domes and pyrheliometer window © 2018 Solargis page 26 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 3.11 Pyranometers replaced at Chileka station. Figure 3.12 New LI-COR installed and levelled. © 2018 Solargis page 27 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 4 SOLAR RESOURCE MODEL DATA 4.1 Solar model Solar radiation is calculated by Solargis model, which are parameterized by a set of inputs characterizing the cloud transmittance, state of the atmosphere and terrain conditions. A comprehensive overview of the Solargis model is made available in the recent book publication [3]. The methodology is also described in [4, 5]. The related uncertainty and requirements for bankability are discussed in [6, 7]. In Solargis approach, the clear-sky irradiance is calculated by the simplified SOLIS model [8]. This model allows the fast calculation of clear-sky irradiance from the set of input parameters. Sun position is a deterministic parameter, and it is described by the algorithms with satisfactory accuracy. Stochastic variability of clear-sky atmospheric conditions is determined by changing concentrations of atmospheric constituents, namely aerosols, water vapor and ozone. Global atmospheric data, representing these constituents, are routinely calculated by world atmospheric data centres: • In Solargis, the new generation aerosol data set representing Atmospheric Optical Depth (AOD) is used. The calculation accuracy is strongly determined by quality of aerosols, especially for cloudless conditions. The aerosol data implemented by MACC-II/CAMS and MERRA-2 projects are used [9, 10]. • Water vapor is also highly variable in space and time, but it has lower impact on the values of solar radiation, compared to aerosols. The GFS and CFSR databases (NOAA NCEP) are used in Solargis, and the data represent the daily variability from 1994 to the present time [11, 12, 13]. • Ozone absorbs solar radiation at wavelengths shorter than 0.3 µm, thus having negligible influence on the broadband solar radiation. The clouds are the most influencing factor, modulating clear-sky irradiance. The effect of clouds is calculated from satellite data in the form of the cloud index (cloud transmittance). The cloud index is derived by relating radiance recorded by the satellite in spectral channels and surface albedo to the cloud optical properties. In Solargis, the modified calculation scheme of Cano has been adopted to retrieve cloud optical properties from the satellite data [14]. To calculate all-sky irradiance in each time step, the clear-sky global horizontal irradiance is coupled with cloud index. Direct Normal Irradiance (DNI) is calculated from Global Horizontal Irradiance (GHI) using modified Dirindex model [15]. Diffuse irradiance for tilted surfaces, which is calculated by Perez model [16]. The calculation procedure also includes terrain disaggregation; the spatial resolution is enhanced with use of the digital terrain model to 250 meters [17]. Solargis model version 2.1 has been used. Table 4.1 summarizes technical parameters of the model inputs and of the primary data outputs. Table 4.1 Input data used in the Solargis and related GHI and DNI outputs for Malawi Inputs into the Solargis Source Time Original Approx. grid model of input data representation time step resolution Cloud index Meteosat MFG and MSG 1994 to 2004 30 minutes 2.8 x 3.3 km satellites (EUMETSAT) 2005 to date 15 minutes 3.3 x 4.0 km Atmospheric optical depth MACC-II/CAMS* (ECMWF) 2003 to date 3 hours 75 km and 125 km (aerosols)* MERRA-2 (NASA) 1994 to 2002 1 hour 50 km Water vapour CFSR/GFS (NOAA) 1994 to date 1 hour 35 and 55 km Elevation and horizon SRTM-3 (SRTM) - - 250 m Solargis primary outputs - 1994 to date 15 minutes 250 m (GHI, DNI) * Aerosol data for 2003-2012 come from the reanalysis database; the data representing years 2013-present are derived from near- real time operational model © 2018 Solargis page 28 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 4.1 Solar meteorological stations in the context of global horizontal irradiation. Data source: Solargis model 4.2 Site adaptation of the solar model: Method The fundamental difference between a satellite observation and a ground measurement is that the signal received by the satellite radiometer integrates a large area, while a ground station represents a pinpoint measurement. This results in a mismatch when comparing instantaneous values from these two observation instruments, mainly during intermittent cloudy weather and changing aerosol load. Nearly half of the hourly Root Mean Square Deviation (RMSD) for GHI and DNI can be attributed to this mismatch (value at sub-pixel scale), which is also known as the “nugget effect” [18]. The satellite pixel is not capable describing the inter-pixel variability in complex regions, where within one pixel diverse natural conditions mix-up (e.g. fog in narrow valleys or along the coast or locations with high gradient). In © 2018 Solargis page 29 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 addition, the coarse spatial resolution of atmospheric databases such as aerosols or water vapor is not capable to describe local patterns of the state of atmosphere. These features can be seen in the satellite DNI data by increased bias due to imperfect description of the aerosol load. Satellite data have inherent inaccuracies, which have certain degree of geographical and time variability. Especially DNI is strongly sensitive to variability of cloud information, aerosols, water vapor, and terrain shading. The relation between uncertainty of global and direct irradiance is nonlinear. Often, a negligible error in global irradiance may have high counterpart in the direct irradiance component. The solar energy projects require representative and accurate GHI and DNI time series. The satellite-derived databases are used to describe long-term solar resource for a specific site. However, their problem when compared to the high-quality ground measurements is a slightly higher bias and partial disagreement of frequency distribution functions, which may limit their potential to record the occurrence of extreme situations (e.g. very low atmospheric turbidity resulting in a high DNI and GHI). A solution is to correlate satellite-derived data with ground measurements to understand the source of discrepancy and subsequently to improve the accuracy of the resulting time series. The Solargis satellite-derived data are correlated with ground measurement data with two objectives: • Improvement of the overall bias (removal of systematic deviations) • Improvement of the fit of the frequency distribution of values. Limited spatial and temporal resolution of the input data and simplified nature of the models results in the occurrence of systematic and random deviations of the model outputs when compared to the ground observations. The deviations in the satellite-computed data, which have systematic nature, can be reduced by site adaptation or regional adaptation methods. In this report we apply site-adaptation of the solar model, i.e. adapting the model for the local conditions represented by on-site measurements. The terminology related to the procedure improving accuracy of the satellite data is not harmonized, and various terms are used: • Correlation of ground measurements and satellite-based data; • Calibration of the satellite model (its inputs and parameters); • Site adaptation of satellite-based data. The term site adaptation is more general and best explains the concept of adapting the satellite-based model (by correlation, calibration, fitting and recalculation) to the ground measured data. Site adaptation aims to adapt the characteristics of the satellite-based time series to the site-specific conditions described by local measurements. Three conditions are important for successful adaptation of the satellite-based model: 1. High quality DNI and GHI ground measurements for at least 12 months must be available; optimally data for 2 or 3 years should be used; 2. High quality satellite data must be used, with consistent quality over the whole period of data; 3. There must be identified a systematic difference between both data sources. Systematic difference can be measured by two characteristics: • Bias (offset) • Systematic deviation in the distribution of hourly or daily values (in the histogram) Systematic difference can be stable over the year or it can slightly change seasonally for certain meteorological conditions (e.g. typical cloud formation during a day, seasonal air pollution). The data analysis should distinguish systematic differences from those arising at occasional events, such as extreme sand storms or forest fires. The episodically-occurring differences may mislead the results of adaptation, especially if short period of ground measurements is only available. If one of the three above-mentioned conditions is not fulfilled, site adaptation (regional adaptation) will not provide the expected results. On the contrary, such an attempt may provide worse results. For the quantitative assessment of the accuracy enhancement procedures, the following metrics is used: • Metrics based on the comparison of all pairs of the hourly daytime data values: Mean Bias, and Root Mean Square Deviation (RMSD), histogram, in an absolute and relative form (divided by the daytime mean DNI values); • Metrics based on the difference of the cumulative distribution functions: KSI (Kolmogorov-Smirnov test Integral) [19] © 2018 Solargis page 30 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 The normalized KSI is defined as an integral of absolute differences of two cumulative distribution functions D normalized by the integral of critical value acritical: , where critical value depends on the number of the data pairs N. As the KSI value is dependent on the size of the sample, the KSI measure may be used only for the relative comparison of fit of cumulative distribution of irradiance values. For the accuracy enhancement of solar resource parameters in this study, a combination of two methods was used. First, systematic deviations due to influence of aerosols were partially removed. Afterwards, to improve the distribution of values, the fitting of cumulative frequency distribution curves of ground measurements and satellite data was used. The site-adaptation procedure first identifies the sources of discrepancies by comparing the ground-measured data with Solargis model data, for the period of the overlap between both data sets. Based on this analysis, correction coefficients to improve the fit between the measured and the model Solargis data are developed. In the second step, these coefficients are used for the adaptation of the full history (24-years) time series. The satellite data is available in 15-minute time step and the ground measurements in 1-minute time step. To partially remove the conceptual difference of point and satellite pixel measurements, prior to site adaptation, all the measures are calculated using aggregated data in the hourly time step. The adaptation was based on measured DNI data from RSR 2 (Kasungu and Mzuzu) and CHP1 (Chileka) instruments and GHI data from the secondary standard CMP10 instrument. GHI measured by the RSR 2 was not used, however the difference to the secondary standard instrument in the measured sites is very small (Chapter 3.3). 4.3 Results of the model adaptation at three sites The original Solargis data show a regional pattern of overestimation, compared to the ground measurements – for both GHI and DNI. The biggest difference was found at Mzuzu station where the mismatch between ground measurements and satellite data exceeded 12% and 20% for GHI and DNI, respectively. Such discrepancy is beyond usual uncertainty of Solargis satellite data for this region (typical expectation is ±6 to ±8% for GHI and ±12 to ±15% for DNI). The detailed inspection of the ground measurements and satellite data indicates three possible sources (or their combination) of this mismatch: • Specific geographical conditions of the site, where climate is influenced by the distance to the Lake Malawi and orographic features determining strongly changing microclimates. The gradient of solar radiation in the West-East direction at the distance of 20 km is 9% for GHI, and 21% for DNI, respectively. Considering size of the satellite pixel of ca 3.3 x 4.0 km, the strong microclimatic gradient can be only partially recorded in the satellite data. The limited resolution of the satellite data is one source of its mismatch with pinpoint ground measurements. • Quality control of ground measurements for Mzuzu site also indicates influence of local shading. The high frequency variability of clouds (small scattered clouds) makes it difficult to distinguish the shading from surrounding objects (drop of solar irradiance due to clouds). The shading due to nearby objects may partly reduce diffuse irradiance, thus affecting the values of GHI and DNI. • Performance of satellite models is in general lower in the conditions of high occurrence of scattered clouds and specific persistent cloudy situation in the tropical regions. Similar features (mainly microclimate gradient) can be found also in Chileka site, showing higher mismatch between model data and measurements. The model adaptation allowed removing a large part of the mismatch between the satellite-based data and the ground measurements. Tables 4.2 to 4.5 summarize validation of the site-adaptation results for all solar measuring stations. © 2018 Solargis page 31 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 4.2 Direct Normal Irradiance: bias and KSI before and after model site-adaptation Meteo station Original DNI data DNI after site adaptation Bias Bias KSI Bias Bias KSI 2 2 [kWh/m ] [%] [-] [kWh/m ] [%] [-] Chileka 42 10.5 188 0 0.0 54 Kasungu 41 9.6 192 0 0.0 68 Mzuzu 78 20.5 360 0 0.0 99 Mean 54 13.5 247 0 0.0 74 Standard deviation 21 6.0 0 0.0 Table 4.3 Global Horizontal Irradiance: bias and KSI before and after model site-adaptation Meteo station Original GHI data GHI after site adaptation Bias Bias KSI Bias Bias KSI [kWh/m2] [%] [-] [kWh/m2] [%] [-] Chileka 38 8.3 174 0 0.0 34 Kasungu 27 5.4 126 0 0.0 15 Mzuzu 59 12.7 272 0 0.0 44 Mean 41 8.8 191 0 0.0 31 Standard deviation 16 3.7 0 0.0 Table 4.4 Direct Normal Irradiance: RMSD before and after model site-adaptation Meteo station RMSD of original DNI data RMSD of DNI after site adaptation Hourly Daily Monthly Hourly Daily Monthly [%] [%] [%] [%] [%] [%] Chileka 40.4 21.8 13.9 39.2 18.3 4.6 Kasungu 37.4 20.2 12.6 36.3 17.4 6.2 Mzuzu 46.8 29.5 22.9 42.4 19.8 7.4 Mean 41.6 23.8 16.4 39.3 18.5 6.1 Table 4.5 Global Horizontal Irradiance: RMSD before and after model site-adaptation Meteo station RMSD of original GHI data RMSD of GHI after site adaptation Hourly Daily Monthly Hourly Daily Monthly [%] [%] [%] [%] [%] [%] Chileka 25.5 13.8 10.1 23.3 10.5 4.7 Kasungu 20.7 10.2 6.6 19.7 8.3 3.4 Mzuzu 27.0 16.5 13.6 22.8 10.2 4.3 Mean 24.4 13.5 10.1 21.9 9.7 4.1 As a result of the site adaptation at the level of individual measurement sites in Malawi, the mean bias of the adapted values was reduced to zero. The values of RMSD and KSI accuracy parameters are also reduced, for both GHI and DNI. Figures 4.2 to 4.4 present the results of the site adaptation for all three sites. © 2018 Solargis page 32 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Chileka: Original DNI Chileka: DNI after adaptation Chileka: Original GHI Chileka: GHI after adaptation Figure 4.2: Correction of DNI and GHI hourly values for Chileka. Left: original Solargis data, right: site-adapted Solargis data. The X-axis represents the measured data and the Y-axis represents the satellite-derived data. © 2018 Solargis page 33 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Kasungu: Original DNI Kasungu: DNI after adaptation Kasungu: Original GHI Kasungu: GHI after adaptation Figure 4.3: Correction of DNI and GHI hourly values for Kasungu Left: original Solargis data, right: site-adapted Solargis data. The X-axis represents the measured data and the Y-axis represents the satellite-derived data. © 2018 Solargis page 34 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Mzuzu: Original DNI Mzuzu: DNI after adaptation Mzuzu: Original GHI Mzuzu: GHI after adaptation Figure 4.4: Correction of DNI and GHI hourly values for Mzuzu. Left: original Solargis data, right: site-adapted Solargis data. The X-axis represents the measured data and the Y-axis represents the satellite-derived data. The change of model GHI and DNI for 24 year period after adaptation is presented on an example of Kasungu (Figure 4.5). Both GHI and DNI after adaptation are significantly lower. The other sites are very similar (Table 4.6). The site-adapted model values better represent the geographical variability of DNI and GHI solar resource and they also improve the distribution and match of hourly values. The measurements show that the model performs well in the region, and these results improve the confidence about the reliability of the measured and modelled solar resource data for Malawi. © 2018 Solargis page 35 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 4.5: Comparison of Solargis original and site-adapted data for Kasungu site. Left: DNI; Right: GHI; Data represent years 1994 to 2017. Table 4.6 Comparison of long-term average of yearly summaries of original and site-adapted values Meteo station DNI annual values GHI annual values Original Site-adapted Difference Original Site-adapted Difference [kWh/m2] [kWh/m2] [%] [kWh/m2] [kWh/m2] [%] Chileka 1779 1609 -9.6 2006 1854 -7.6 Kasungu 1861 1689 -9.3 2105 1995 -5.2 Mzuzu 1785 1457 -18.4 2062 1825 -11.5 © 2018 Solargis page 36 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 5 METEOROLOGICAL MODEL DATA 5.1 Meteorological model For the territory of Malawi, the last 24 years of the Solargis model-based meteorological data is derived from the regional models. The meteorological data in Solargis database is derived from the combination of several data sources: CFSR, CFSv2 from NOAA, and MERRA-2 from NASA. The original characteristics of the models are specified in Table 5.1. Table 5.1: Original source of Solargis meteorological data: models MERRA-2, CFSR and CFSv2. Climate Forecast System Climate Forecast System Climate Forecast System Reanalysis (CFSR) (MERRA-2) (CFSv2) Period 1994 to 2010 1994 to 2010 2011 to the present time Original spatial resolution 30 x 35 km 45 x 50 km 19 x 22 km Original time resolution 1 hour 1 hour 1 hour Table 5.2 shows meteorological parameters available in Solargis, their specifications, and it also indicates which of them have been delivered within this study. The original spatial resolution of the models is enhanced to 1 km for air temperature by spatial disaggregation and use of the Digital Elevation Model SRTM-3. The spatial resolution of other parameters is unchanged. Table 5.2 Solargis meteorological parameters delivered for 3 meteo sites Meteorological parameter Acronym Unit Time resolution Spatial Data Data representation delivered validated Air temperature at 2 meters TEMP °C 60 minute 1 km Yes Yes (dry bulb temperature) Relative humidity at 2 meters RH % 60 minute Original model Yes Yes 2 Wind speed at 10 meters WS m/s 60 minute Original model Yes Yes Wind direction at 10 meters WD ° 60 minute Original model Yes Yes Atmospheric pressure AP hPa 60 minute 1 km Yes Yes Precipitable water PWAT 60 minute Original model Yes No Important note: the meteorological parameters are derived from the numerical weather model outputs and they have lower spatial and temporal resolution. Thus, they do not represent the same accuracy as the solar resource data. Especially wind speed data has higher uncertainty, and it provides only overview information for solar energy projects. Thus, the local microclimate of the three meteorological sites may deviate from the values derived from the Solargis global database. 5.2 Validation of meteorological data The validation procedure was carried out to compare the modelled data with ground-measurements from the 3 meteorological stations installed within the ESMAP project: Chileka, Kasungu and Mzuzu. In general, the data from the meteorological models represent larger area, and it is not capable to represent accurately the local microclimate on the solar meteorological sites. © 2018 Solargis page 37 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 5.2.1 Air temperature at 2 meters Air temperature is derived from the model outputs of MERRA-2 and CSFv2 meteorological models, and the data is recalculated at the spatial resolution of 1 km (Table 5.3 and Figures 5.1 to 5.3). Considering spatial and time interpolation, the deviation of the modelled values to the ground observations for hourly values can occasionally reach several degrees of Celsius. Figures 5.1 to 5.3 show graphical representation of the model values accuracy at the meteorological stations. In general, the model matches the ground measurements quite well. The main issue identified is underestimation of air temperature by the model: the day-time temperature is represented with higher accuracy than night-time. To improve the data match between the model and ground measurements, model site adaptation was applied to reduce small systematic negative bias and to improve night-time temperature. Table 5.3 Air temperature at 2 m: Accuracy indicators of the model outputs [ºC]. CFSv2 model Meteorological station Bias Bias Bias Bias Bias RMSD RMSD RMSD mean min max nigh-time day-time hourly daily monthly Chileka -1.8 -2.7 0.3 -2.9 -0.8 3.1 2.3 2.0 Chileka, adapted 0.0 -0.7 1.9 -0.9 1.0 2.3 1.0 0.0 Mzuzu -1.7 -1.8 -1.1 -2.3 -1.2 2.5 2.0 1.7 Mzuzu, adapted 0.0 0.1 0.4 -0.5 0.5 1.8 0.9 0.0 Kasungu -1.7 -1.9 -0.8 -2.1 -1.3 2.3 1.9 1.7 Kasungu, adapted 0.0 -0.2 0.8 -0.4 0.4 1.6 0.7 0.1 Figure 5.1: Scatterplots of air temperature at 2 m at Chileka meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis) Blue: day-time, Black: night-time measurements, Red: model data before site adaptation © 2018 Solargis page 38 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 5.2: Scatterplots of air temperature at 2 m at Mzuzu meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis) Blue: day-time, Black: night-time measurements, Red: model data before site adaptation Figure 5.3: Scatterplots of air temperature at 2 m at Kasungu meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis) Blue: day-time, Black: night-time measurements, Red: model data before site adaptation © 2018 Solargis page 39 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 5.2.2 Relative humidity Relative humidity is calculated from the specific humidity, air pressure and the site-adapted air temperature. The comparison of the model values with the ground measurements at all 3 meteorological stations is shown in Table 5.4 and Figures 5.4 to 5.6. In general, the model matches the ground measurements quite well, representing both daily and yearly profiles. Table 5.4 Relative humidity: Accuracy indicators of the model outputs [%]. CFSv2 model Meteorological station Bias Bias Bias Bias Bias RMSD RMSD RMSD mean min max nigh-time day-time hourly daily monthly Chileka 8 3 6 11 5 15 11 10 Mzuzu 2 2 1 3 1 9 5 3 Kasungu 4 2 4 5 3 10 7 5 Figure 5.4: Scatterplots of relative humidity at 2 m at Chileka meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis). Blue: day-time, black: night-time measurements. © 2018 Solargis page 40 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 5.5: Scatterplots of relative humidity at 2 m at Mzuzu meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis). Blue: day-time, black: night-time measurements. Figure 5.6: Scatterplots of relative humidity at 2 m at Kasungu meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis). Blue: day-time, black: night-time measurements. © 2018 Solargis page 41 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 5.2.3 Wind speed and wind direction at 10 meters Wind speed and direction values delivered within Solargis data represent the height at 10 meters, and they are calculated from the MERRA-2 and CFSv2 models, from 10 m wind u- and v- components. The spatial resolution is kept as in the original data. Wind measurements take place at the height of 3 m for Mzuzu and Kasungu stations and 10 m for Chileka. Different height of ground measurements and model data can be the important source of systematic deviation . Comparison of the modelled wind speed with ground measurements is shown in Table 5.5 and Figures 5.7 to 5.9. The model values underestimate the wind conditions measured at Chileka and overestimate at Mzuzu. This can be attributed to different height of wind sensors above ground, and the model. Solar measuring station at Mzuzu shows conditions with low wind, compared to other locations. The data representation for wind speed and wind direction strongly depends on the local conditions; therefore the model values are only indicative; they better characterize a larger region rather than the local microclimate. In order to slightly improve the match of data between the models and ground measurements and to reduce the difference between the models, a site adaptation for Chileka and Mzuzu was applied. Table 5.5 Wind speed: accuracy indicators of the model outputs [m/s]. CFSv2 model Meteorological station Bias Bias Bias Bias Bias RMSD RMSD RMSD mean min max nigh-time day-time hourly daily monthly Chileka -1.7 -0.8 -2.7 -1.6 -1.8 2.1 1.8 1.7 Chileka, adapted 0.0 -0.2 0.3 -0.1 0.0 1.7 0.8 0.3 Mzuzu 1.5 0.9 2.0 1.3 1.8 1.7 1.6 1.6 Mzuzu, adapted 0.6 0.5 0.4 0.7 0.6 0.8 0.7 0.6 Kasungu 0.6 1.1 0.0 0.8 0.3 1.1 0.7 0.6 Figure 5.7: Scatterplots of wind speed at Chileka meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis). Blue: day-time, black: night-time, Red: model data before site adaptation. (observations and model data both at 10 m height) © 2018 Solargis page 42 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 5.8: Scatterplots of wind speed at Mzuzu meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis). Blue: day-time, black: night-time, Red: model data before site adaptation. (observations at 3 m height and model data at 10 m) Figure 5.9: Scatterplots of wind speed at Kasungu meteorological station. Measured values (horizontal axis) and meteorological model values (vertical axis). Blue: day-time, black: night-time. (observations at 3 m height and model data at 10 m) © 2018 Solargis page 43 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 5.3 Uncertainty of meteorological model data The meteorological parameters are derived from a combination of numerical meteorological models (MERRA-2, CFSv2 and CFSR). Considering the comparison results, the uncertainty of the estimate for three main meteorological parameters is summarized in Table 5.6. It was found that the modelled air temperature fits reasonably well the measured data, although due to the spatial resolution there are issues such as underestimation of night-time temperature and systematic small negative bias. Site adaptation was applied to partially mitigate these small discrepancies for air temperature. Similarly to air temperature, the model values of relative humidity fit quite well the measured data, representing both daily and yearly amplitude. Wind speed data, obtained from the meteorological model, represents an area of larger region, and this data is smoothed, in comparison to the point measurements collected at the meteorological sites. For Kasungu and Mzuzu, there is a discrepancy between the height of wind sensors at the meteorological station and the model. In order to slightly improve the data match between the models and ground measurements, and to reduce existing difference between MERRA-2 and CFSv2 models, at these two sites, a site adaptation was applied. Table 5.6 Expected uncertainty of modelled meteorological parameters at the project sites. Unit Annual Monthly Hourly Air temperature at 2 m °C ±1.5 ±1.5 ±2.5 Relative Humidity at 2 m % ±10 ±10 ±15 Average wind speed at 10 m m/s ±1.0 ±1.5 ±2.0 © 2018 Solargis page 44 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 6 SOLAR RESOURCE: UNCERTAINTY OF LONG-TERM ESTIMATES 6.1 Uncertainty of solar resource yearly estimate The uncertainty of site-adapted satellite-based GHI and DNI is determined by the uncertainty of the model and of the ground measurements [7], more specifically it depends on: 1. Parameterization and adaptation of numerical models integrated in Solargis for the given data inputs and their ability to generate accurate results for various geographical and time-variable conditions: • Data inputs into Solargis model (accuracy of satellite data, aerosols and water vapour and Digital Terrain Model). • Clear-sky model and its capability to properly characterize various states of the atmosphere • Simulation accuracy of the satellite model and cloud transmittance algorithms, being able to properly distinguish different types of desert surface, clouds, fog, but also snow and ice. • Diffuse and direct decomposition models • Site adaptation methods. • Ability of model to describe in sufficient detail specific microclimatic features such as strong gradient of solar irradiance (Chapter 4.3) 2. Uncertainty of the ground-measurements, which is determined by: • Accuracy of the instruments • Maintenance practices, including sensor cleaning, calibration • Data post-processing and quality control procedures. The statistics, such as bias and RMSD (Chapter 4.3) characterize accuracy of Solargis model in the given validation points, relative to the ground measurements. The validation statistics is affected by local geography and by quality and reliability of the ground-measured data. Therefore the validation statistics only indicates performance of the model in the region. At the same time, it can be strongly affected by specific local conditions. Solargis model uncertainty is compared to the data measured by the solar meteorological instruments. Representativeness of such data comparison (satellite and ground-measured) is determined by the precision of the measuring instruments, the maintenance and operational practices, and by quality control of the measured data – in other words, by the measurement accuracy achieved at each meteorological station. From the user’s perspective, the information about the model uncertainty has probabilistic nature. It generalizes the validation accuracy and it has to be considered at different confidence levels. The expert estimate of the calculation uncertainty in this report assumes 80% probability of occurrence of values. The solar model uncertainty is discussed in Chapters 4 and 6.1. The main findings are summarized in Table 6.1. The site-adaptation procedure reduced uncertainty of estimate of all parameters. Chapter 6.3 evaluates combined uncertainty, in which interannual variability is included as well (Chapter 6.2). The physical reduction of the model uncertainty is visible (Table 6.1); in addition the site adaptation increases confidence in the model data values. © 2018 Solargis page 45 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 6.1 Uncertainty of the model estimates for original and site-adapted annual long-term values (Considers 80% probability of occurrence; in the neighbourhood of three meteorological sites) Uncertainty of long-term Acronym Uncertainty of the original Uncertainty of the Solargis model after site annual values at 3 solar Solargis model adaptation meteorological sites After 1st year After 2nd year Global Horizontal Irradiation GHI ±9.0% (up to 13.0%*) ±4.5% ±4.0% Direct Normal Irradiance DNI ±12.0% (up to 21.0%*) ±6.0% ±5.5% * In complex landscape and specific microclimate with strong spatial gradient of weather patterns 6.2 Uncertainty due to interannual variability of solar radiation Weather changes in cycles and also has stochastic nature. Therefore annual solar radiation in each year can deviate from the long-term average in the range of few percent. The estimation of the interannual variability below shows the magnitude of this change. The uncertainty of GHI and DNI prediction is the highest if only one single year is considered, but when averaged for a longer period, weather oscillations even out and approximate to the long-term average. In this report, the interannual variability is calculated from the unbiased standard deviation stdev of GHI and DNI over 24 years, considering, in the long-term, the normal distribution of the annual sums for n years, where xi is any particular year and ̅ is long-term yearly average. Due to the limited number of years of available data, for the calculation we apply simplified assumption of normal distribution of yearly values: 9 = B7C9 ∑7 E89(E − ̅ ) 6 Tables 6.2 and 6.3 show GHI and DNI values that are to be exceeded at P90 for a consecutive number of years. The variability (varn) for a number of years (n) is calculated from the unbiased standard deviation (stdev): .I1,J 7 = √7 The uncertainty, which characterises 80% probability of occurrence (Uvar), is calculated from the variability (varn), multiplying it with 1.28155: = 1.28155 The lower boundary (negative value) of uncertainty represents 90% probability of exceedance, and it is used for calculating the P90 value. Table 6.2 Annual GHI that should be exceeded with 90% probability in the period of 1 to 10 (25) years Chileka│Years 1 2 3 4 5 6 7 8 9 10 25 Variability [±%] 4.1 2.9 2.3 2.0 1.8 1.7 1.5 1.4 1.4 1.3 0.8 Uncertainty P90 [±%] 5.2 3.7 3.0 2.6 2.3 2.1 2.0 1.8 1.7 1.6 1.0 Minimum GHI P90 1758 1786 1798 1806 1811 1815 1818 1820 1822 1824 1835 Kasungu│Years 1 2 3 4 5 6 7 8 9 10 25 Variability [±%] 3.6 2.5 2.1 1.8 1.6 1.5 1.4 1.3 1.2 1.1 0.7 Uncertainty P90 [±%] 4.6 3.3 2.7 2.3 2.1 1.9 1.7 1.6 1.5 1.5 0.9 Minimum GHI P90 1903 1930 1942 1949 1954 1957 1960 1963 1964 1966 1977 Mzuzu│Years 1 2 3 4 5 6 7 8 9 10 25 Variability [±%] 3.8 2.7 2.2 1.9 1.7 1.6 1.5 1.4 1.3 1.2 0.8 Uncertainty P90 [±%] 4.9 3.5 2.8 2.5 2.2 2.0 1.9 1.7 1.6 1.6 1.0 Minimum GHI P90 1735 1762 1773 1780 1785 1789 1791 1793 1795 1797 1807 © 2018 Solargis page 46 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 6.3 Annual DNI that should be exceeded with 90% probability in the period of 1 to 10 (25) years. Chileka│Years 1 2 3 4 5 6 7 8 9 10 25 Variability [±%] 9.1 6.4 5.2 4.5 4.1 3.7 3.4 3.2 3.0 2.9 1.8 Uncertainty P90 [±%] 11.6 8.2 6.7 5.8 5.2 4.7 4.4 4.1 3.9 3.7 2.3 Minimum DNI P90 1422 1476 1501 1515 1525 1532 1538 1543 1546 1550 1571 Kasungu│Years 1 2 3 4 5 6 7 8 9 10 25 Variability [±%] 8.3 5.8 4.8 4.1 3.7 3.4 3.1 2.9 2.8 2.6 1.7 Uncertainty P90 [±%] 10.6 7.5 6.1 5.3 4.7 4.3 4.0 3.7 3.5 3.4 2.1 Minimum DNI P90 1510 1562 1585 1599 1609 1616 1621 1625 1629 1632 1653 Mzuzu│Years 1 2 3 4 5 6 7 8 9 10 25 Variability [±%] 8.6 6.1 5.0 4.3 3.9 3.5 3.3 3.0 2.9 2.7 1.7 Uncertainty P90 [±%] 11.0 7.8 6.4 5.5 4.9 4.5 4.2 3.9 3.7 3.5 2.2 Minimum DNI P90 1296 1343 1364 1377 1385 1391 1396 1400 1403 1406 1425 One can interpret the above Tables 6.2 and 6.3 on the example of Chileka site as follows: i. GHI interannual variability at P90 of 5.2% has to be considered for any single year in Chileka. In other words, assuming that the long-term average is 1854 kWh/m2, it is expected (with 90% probability) that annual GHI exceeds, at any single year, the value of 1758 kWh/m2. ii. Within a period of three consecutive years, it is expected at P90 that annual average of GHI in Chileka exceeds value of 1798 kWh/m2; iii. For a period of 25 years, it is expected at 90% probability that due to interannual variability the estimate of the long-term annual DNI average will deviate within the range of ±2.3% in Chileka. Thus assuming that the estimate of the long-term average is 1609 kWh/m2, it can be expected at P90 that due to variability of weather, it should be at least 1571 kWh/m2. It is to be underlined that prediction of the future irradiation is based on the analysis of the recent historical data (period 1994 to 2017). Future weather changes may include man-induced climate change or natural events such as volcano eruptions, which may have impact on this prediction. Based on the existing scientific knowledge [20, 21], an effect of extreme volcano eruptions, with an emission of large amount of stratospheric aerosols, can be estimated on the example of Pinatubo event in 1991 (the second largest volcano eruption in 20th century). It can be expected that in such a case, the annual DNI in the affected year can decrease by approx. 16% or more, compared to the long-term average, still influencing another two consecutive years. In the same way, the volcano eruption of the comparable size may reduce long-term average estimate of DNI by about 4%. The decrease of GHI is much lower; the annual value in the particular year of eruption could be reduced by about 2% compared to the long-term average. 6.3 Combined uncertainty In this Chapter, the combined uncertainty of the annual GHI and DNI values is quantified. Taking into account uncertainties of both types of data (satellite and ground measured), the combined effect of two components of the uncertainty of the site-adapted GHI and DNI values has to be considered. 1. Uncertainty of the estimate (Uest) of the annual solar resource values, after site-adaptation, which is ±4.0% for GHI and ±5.5% for DNI (Chapter 6.1); 2. Interannual variability (Uvar) in any particular year, due to changing weather. At three Malawian sites, it varies from ±4.6% to ±5.2% for GHI and from ±10.6% to ±11.6% for DNI. The uncertainty due to weather variability decreases over the time with square root of the number of years (Chapter 6.2). The two above-mentioned uncertainties combine in Uc (see Glossary), which represents a conservative expectation of the minimum GHI and DNI assuming various number of years N (Tables 6.4 and 6.5). Considering a simplified assumption of normal distribution of the annual values, probability of exceedance can be calculated at different confidence levels. GHI and DNI minimum annual values expected for combined uncertainty in any single year are shown on Figure 6.1 and 6.2. © 2018 Solargis page 47 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 6.4 Combined probability of exceedance of annual GHI for uncertainty of the estimate ±4.0%. Nr. of Uncertainty Interanual Combined Expected minimum│Chileka 2 years of estimate variability uncertainty [kWh/m ] N [±%] N years [±%] P90 [±%] P01 P05 P10 P25 P50 P75 P90 P95 P99 1 4.0 5.2 6.6 2075 2010 1976 1918 1854 1790 1732 1698 1633 5 4.0 2.3 4.6 2010 1964 1940 1899 1854 1809 1768 1744 1698 10 4.0 1.6 4.3 2000 1957 1934 1896 1854 1812 1774 1751 1709 25 4.0 1.0 4.1 1993 1952 1931 1894 1854 1814 1777 1756 1715 Nr. of Uncertainty Interanual Combined Expected minimum│Kasungu 2 years of estimate variability uncertainty [kWh/m ] N [±%] N years [±%] P90 [±%] P01 P05 P10 P25 P50 P75 P90 P95 P99 1 4.0 4.6 6.1 2216 2151 2117 2059 1995 1931 1873 1839 1774 5 4.0 2.1 4.5 2158 2110 2085 2042 1995 1948 1905 1880 1832 10 4.0 1.5 4.3 2149 2104 2080 2040 1995 1950 1910 1886 1841 25 4.0 0.9 4.1 2144 2100 2077 2038 1995 1952 1913 1890 1846 Nr. of Uncertainty Interanual Combined Expected minimum│Mzuzu 2 years of estimate variability uncertainty [kWh/m ] N [±%] N years [±%] P90 [±%] P01 P05 P10 P25 P50 P75 P90 P95 P99 1 4.0 4.9 6.3 2035 1974 1941 1886 1825 1764 1709 1677 1615 5 4.0 2.2 4.6 1976 1932 1908 1869 1825 1781 1742 1718 1674 10 4.0 1.6 4.3 1967 1926 1904 1866 1825 1784 1747 1725 1683 25 4.0 1.0 4.1 1962 1922 1900 1865 1825 1786 1750 1729 1689 Table 6.5 Combined probability of exceedance of annual DNI for uncertainty of the estimate ±5.5%. Nr. of Uncertainty Interanual Combined Expected minimum│Chileka 2 years of estimate variability uncertainty [kWh/m ] N [±%] N years [±%] P90 [±%] P01 P05 P10 P25 P50 P75 P90 P95 P99 1 5.5 11.6 12.9 1984 1874 1816 1718 1609 1500 1402 1343 1233 5 5.5 5.2 7.6 1830 1765 1730 1673 1609 1545 1487 1452 1388 10 5.5 3.7 6.6 1802 1745 1715 1665 1609 1553 1502 1472 1416 25 5.5 2.3 6.0 1783 1732 1705 1659 1609 1558 1513 1485 1434 Nr. of Uncertainty Interanual Combined Expected minimum│Kasungu 2 years of estimate variability uncertainty [kWh/m ] N [±%] N years [±%] P90 [±%] P01 P05 P10 P25 P50 P75 P90 P95 P99 1 5.5 10.6 11.9 2055 1947 1890 1795 1689 1583 1487 1430 1323 5 5.5 4.7 7.3 1911 1846 1811 1753 1689 1624 1566 1531 1466 10 5.5 3.4 6.4 1886 1828 1797 1746 1689 1631 1580 1549 1491 25 5.5 2.1 5.9 1869 1816 1788 1741 1689 1636 1589 1561 1508 Nr. of Uncertainty Interanual Combined Expected minimum│Mzuzu 2 years of estimate variability uncertainty [kWh/m ] N [±%] N years [±%] P90 [±%] P01 P05 P10 P25 P50 P75 P90 P95 P99 1 5.5 11.0 12.3 1783 1688 1637 1552 1457 1362 1277 1226 1131 5 5.5 4.9 7.4 1653 1595 1565 1514 1457 1400 1349 1319 1262 10 5.5 3.5 6.5 1629 1579 1552 1507 1457 1407 1362 1335 1285 25 5.5 2.2 5.9 1614 1568 1543 1503 1457 1412 1371 1346 1300 This analysis is based on the data representing a history of years 1994 to 2017, and on the expert extrapolation of the related weather variability. This report may not reflect possible man-induced climate change or occurrence of extreme events such as large volcano eruptions in the future (see the last paragraph in Chapter 6.2). © 2018 Solargis page 48 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Graphical visualisation of Tables 6.4 and 6.5 on the example of Chileka is shown in Figures 6.1 and 6.2, where the expected probabilities of exceedance (different Pxx scenarios) are drawn on the cumulative distribution curve showing yearly GHI and DNI values. 100 P99: 1633 P95: 1698 GHI Value at Pxx P50 P90: 1732 P75 P90 90 P95 P99 80 P75: 1790 70 60 P50: 1854 Pxx 50 40 30 20 10 0 1550 1650 1750 1850 1950 2050 2150 GHI [kWh/m2] Figure 6.1: Expected Pxx values for GHI at Chileka site 100 P99: 1233 P95: 1343 DNI Value at Pxx P50 P90: 1402 P75 P90 90 P95 P99 80 P75: 1500 70 60 P50: 1609 Pxx 50 40 30 20 10 0 1050 1150 1250 1350 1450 1550 1650 1750 1850 1950 2050 2150 DNI [kWh/m2] Figure 6.2: Expected Pxx values for DNI at Chileka site © 2018 Solargis page 49 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 7 TIME SERIES AND TYPICAL METEOROLOGICAL YEAR DATA 7.1 Delivered data sets This report is accompanied by data sets delivered individually for position of each of three solar meterological stations in Malawi. The data include (Tables 7.1 and 7.2): • Solar and meteorological measurements, after second-stage quality assessment (first stage was executed by GeoSUN Africa) representing 24 months of the measuring campaign; • Time series, representing last 24+ years; • Typical Meteorological Year data, representing last 24 years. The data is delivered in formats ready to use in energy simulation software. This report provides detailed insight of the methodologies and results. Table 7.1 Delivered data, and its key characteristics Feature Time coverage Primary time step Delivered files Ground measurements Mar 2016 to Mar 2018 1 minute Quality controlled measurements – 1- minute (GeoSUN Africa) Analytical tasks consider only a period from Apr 2016 to Mar 2018 Model data – original Jan 1994 to March 2018 15 minutes Time series – hourly (Solargis) Time series – monthly Time series – yearly Model data – site adapted Jan 1994 to March 2018 15 minutes Time series – hourly (Solargis) Time series – monthly Time series – yearly Model data – site adapted Jan 1994 to Dec 2017 hourly Typical Meteorological Year P50 – hourly (Solargis) Typical Meteorological Year P90 – hourly Table 7.2 Parameters in the delivered site-adapted time series and TMY data (hourly time step) Parameter Acronym Unit TS TMY P50 TMY P90 Global horizontal irradiance GHI W/m2 X X X Direct normal irradiance DNI W/m2 X X X Diffuse horizontal irradiance DIF W/m2 X X X Global tilted irradiance (at optimum angle) GTI W/m2 X - - Solar azimuth SA ° X X X Solar elevation SE ° X X X Air temperature at 2 metres TEMP °C X X X Wind speed at 10 metres WS m/s X X X Wind direction at 10 metres WD ° X X X Relative humidity RH % X X X Air pressure AP hPa X X X Precipitable water PWAT kg/m2 X X X © 2018 Solargis page 50 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 7.2 TMY method The Typical Meteorological Year (TMY) data sets are delivered, together with Solargis time series data and this report. TMY contains hourly data derived from the time series covering complete 24 years (1994 to 2017). The data history of 24 years is compressed into one year (Figure 7.3 to 7.6) following two criteria: • Minimum difference between statistical characteristics (annual average, monthly averages) of TMY and long-term time series. This criterion is given about 80% weighting. • Maximum similarity of monthly Cumulative Distribution Functions (CDF) of TMY and full-time series, so that occurrence of typical hourly values is well represented for each month. This criterion is given about 20% weighting. TMY P50 data set is constructed on the monthly basis. For each month the long-term average monthly value and cumulative distribution for each parameters is calculated: Direct Normal Irradiance (DNI), Global Horizontal Irradiance (GHI), Diffuse Horizontal Irradiance (DIF) and Air Temperature (TEMP). Following the monthly data for each individual year from the set of 24 years are compared to the long-term parameters. The monthly data from the year, which resembles the long-term parameters more closely, is selected. The procedure is repeated for all 12 months, and the TMY is constructed by concatenating the selected months into one artificial (but representative) year. The method for calculation P90 data set is based on the TMY P50 method. It has been modified in a way of how a candidate month is selected. The search for set of twelve candidates is repeated in iteration until a condition of minimization of difference between annual P90 value and annual average of new TMY is reached (instead of minimization of differences in monthly means and CDFs, as applied in P50 case). Once the selection converges to minimum difference, the TMY is created by concatenation of selected months. The P90 annual values are calculated for each confidence limit − from the combined uncertainty of estimate and inter-annual variability, which can occur in any year (Chapter 6.3). To derive TMY that fits specific needs of the selected energy application the different weights are given to individual parameters – thus highlighting important properties. In solar energy applications, the higher importance is given to GHI and DNI. In assembling TMY P50, the values of DNI, GHI, DIF and TEMP are only considered, where the weights are set as follows: 0.9 is given to DNI, 0.3 to GHI, 0.02 to diffuse horizontal irradiance, and 0.07 to air temperature (divided by the total of 1.29). To derive solar resource parameters with an hourly time step, the original satellite data with time resolution of 15- minutes were aggregated by time integration. The meteorological parameters are available in the original 1-hourly time step. The TMY datasets were constructed from solar radiation and meteorological data (Chapters 4 and 5). Time zone was adjusted to UTC +02:00 time. More about the Solargis TMY method in [22]. 7.3 Results Two data sets are derived from the Solargis historical time series for the three sites: P50 and P90. In graphs and tables below we show the values for Chileka meteorological site, in order to present the methodology of TMY data calculation. Important note: Due to the inherent features of the underlying methods, monthly values in the TMY data sets do not fit to the values generated from full time series (Figures 7.1 to 7.4; Tables 7.3 to 7.6). Table 7.3 Monthly and yearly long-term GHI averages as calculated from time series and from TMY representing P50, and P90 cases at Chileka site Global Horizontal 1 2 3 4 5 6 7 8 9 10 11 12 Year Irradiation [kWh/m2] Time series (24 years) 161 152 160 145 139 118 121 149 172 186 179 173 1854 TMY for P50 case 159 149 161 143 142 117 120 152 176 181 179 173 1854 TMY for P90 case 143 118 152 139 132 112 119 146 165 181 171 153 1732 © 2018 Solargis page 51 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 7.4 Monthly and yearly long-term DNI averages as calculated from time series and from TMY representing P50, and P90 cases at Chileka site Direct Normal 1 2 3 4 5 6 7 8 9 10 11 12 Year Irradiation [kWh/m2] Time series (24 years) 99 105 126 144 166 143 130 148 146 147 136 118 1609 TMY for P50 case 99 100 128 143 165 145 134 151 146 144 137 117 1609 TMY for P90 case 68 58 113 131 159 127 121 143 133 138 122 91 1402 Table 7.5 Monthly and yearly long-term DIF averages as calculated from time series and from TMY representing P50, and P90 cases at Chileka site Diffused Horizontal 1 2 3 4 5 6 7 8 9 10 11 12 Year Irradiation [kWh/m2] Time series (24 years) 83 70 68 51 39 37 44 51 65 74 75 81 739 TMY for P50 case 82 69 66 49 41 35 42 51 68 70 74 81 730 TMY for P90 case 88 71 71 52 36 42 48 52 69 78 76 84 768 Table 7.6 Monthly and yearly long-term TEMP averages as calculated from time series and from TMY representing P50, and P90 cases at Chileka site Air temperature [°C] 1 2 3 4 5 6 7 8 9 10 11 12 Year Time series (24 years) 24.7 23.7 23.6 22.2 21.1 19.5 19.0 20.9 23.7 25.7 25.7 24.5 22.9 TMY for P50 case 23.9 23.1 23.6 21.8 21.9 19.4 19.3 20.9 23.8 25.6 25.5 23.9 22.7 TMY for P90 case 23.9 23.6 23.5 21.9 20.9 18.3 19.0 20.8 23.5 25.6 24.5 23.8 22.4 As an example of interpretation of the tVsables above, the TMY data sets for P50 and P90 for the Chileka site can be described as: 1. P50 TMY data set represents, for each month, the average climate conditions and the most representative cumulative distribution function, therefore extreme situations (e.g. extremely cloudy weather) are not represented in this dataset. The long-term annual summary of GHI and DNI are considered as the most critical parameters to consider, and in this data set P50 GHI value is 1854 kWh/m2 and DNI value is 1609 kWh/m2. 2. P90 TMY data set represents for each month the climate conditions, which after summarization GHI and DNI for the whole year results in the value equal or close to P90 derived by the analysis of uncertainty of the estimate and of the interannual variability for any single year (Chapter 6.3). Thus TMY for P90 represents generally a conservative estimate, i.e. a year with the long-term value of GHI of 1732 kWh/m2 and DNI of 1402 kWh/m2. © 2018 Solargis page 52 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 200 150 GHI [kWh/m 2 ] 100 50 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Time series TMY P50 TMY P90 Figure 7.1: GHI monthly values derived from time series and TMY P50 and P90 at Chileka site. 200 150 DNI [kWh/m 2 ] 100 50 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Time series TMY P50 TMY P90 Figure 7.2: DNI monthly values derived from time series and TMY P50 and P90 at Chileka site. 200 150 DIF [kWh/m 2 ] 100 50 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Time series TMY P50 TMY P90 Figure 7.3: DIF monthly values derived from time series and TMY P50 and P90 at Chileka site. © 2018 Solargis page 53 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 30.0 25.0 TEMP [°C] 20.0 15.0 10.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Time series TMY P50 TMY P90 Figure 7.4: TEMP monthly values derived from time series and TMY P50 and P90 at Chileka site It is important to note that the data reduction in the TMY data set is not possible without loss of information contained in the original multiyear time series. Therefore time series data are considered as the most accurate reference suitable for the statistical analysis of solar resource and meteorological parameters of the site. Figure 7.5: Seasonal profile of GHI, DNI and DIF for Typical Meteorological Year P50 Chileka site: X-axis – day of the year; Y-axis – solar irradiance W/m2 © 2018 Solargis page 54 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 7.6: Snapshot of Typical Meteorological Year for P50 for Chileka © 2018 Solargis page 55 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 8 CONCLUSIONS This report accompanies delivery of solar resource and meteorological site-specific measurements and modelled data for three sites, where solar meteorological stations have been installed and operated during two years of measurement campaign (from March 2016 to March 2018). Comparing to the results obtained after period of first twelve months of the measurement campaign, the data uncertainty has been reduced for both GHI and DNI parameters. The measurements reduced the uncertainty of the solar model: for yearly GHI estimates, the uncertainty has been reduced from approx. ±9.0% (13.0% in extreme geographies) to ±4.0%. For DNI, this reduction is from ±12.0% (21.0% in difficult geographies) to ±5.5%. Similarly, the uncertainty of other meteorological modelled data is better understood. Especially for air temperature, which is also very important for solar energy modelling. The historical time series and TMY data computed for three sites are provided in formats ready to use in standard photovoltaic energy simulation software. © 2018 Solargis page 56 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 ANNEX 1: SITE RELATED DATA STATISTICS Yearly summaries of solar and meteorological model outputs Statistics for site adapted model yearly values representing 24 years (1994 to 2017). 6.0 2192 Average annual sum of Global Horizontal Irradiation [kWh/m2] Average daily sum of Global Horizontal Irradiation [kWh/m2] 5.5 2009 5.0 1826 4.5 1644 4.0 1461 3.5 1278 3.0 1096 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Year Chileka 5.3% Kasungu 4.7% Mzuzu 5.0% Figure I: Interannual variability of site-adapted yearly GHI [kWh/m2]. Annual average (avg, solid line) and standard deviation (value behind the names of sites). 6.0 2192 Average annual sum of Direct Normal Irradiation [kWh/m2] Average daily sum of Direct Normal Irradiation [kWh/m2] 5.5 2009 5.0 1826 4.5 1644 4.0 1461 3.5 1278 3.0 1096 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Year Chileka 11.6% Kasungu 10.6% Mzuzu 11.0% Figure II: Interannual variability of site-adapted yearly DNI [kWh/m2]. Annual average (avg, solid line) and standard deviation (value behind the names of sites). © 2018 Solargis page 57 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 26.0 24.0 Average anual air temperature [°C] 22.0 20.0 18.0 16.0 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Chileka Kasungu Mzuzu Figure III: Interannual variability of yearly TEMP [°C]. Annual average (avg, solid line). © 2018 Solargis page 58 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Monthly summaries of solar and meteorological model outputs The graphs compare (site-adapted) monthly model time series compared to long-term statistics. 9.0 9.0 8.0 Chileka 8.0 Kasungu Daily sums of GHI [kWh/m2] Daily sums of GHI [kWh/m2] 7.0 7.0 6.0 6.0 5.0 5.0 4.0 4.0 3.0 3.0 2.0 2.0 1.0 1.0 0.0 0.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec max-min LTA Last Year max-min LTA Last Year 9.0 8.0 Mzuzu Daily sums of GHI [kWh/m2] 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec max-min LTA Last Year Figure IV: GHI monthly averages [kWh/m2]. Monthly average shown as solid line; min/max monthly values as boundary lines; last 12 months in red. 9.0 9.0 8.0 Chileka 8.0 Kasungu 7.0 7.0 6.0 Daily sums of DNI [kWh/m2] 6.0 Daily sums of DNI [kWh/m2] 5.0 5.0 4.0 4.0 3.0 3.0 2.0 2.0 1.0 1.0 0.0 0.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec max-min LTA Last Year max-min LTA Last Year 9.0 8.0 Mzuzu 7.0 6.0 Daily sums of DNI [kWh/m2] 5.0 4.0 3.0 2.0 1.0 0.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec max-min LTA Last Year Figure V: DNI monthly averages [kWh/m2]. Monthly average shown as solid line; min/max monthly values as boundary lines; last 12 months shown in red. © 2018 Solargis page 59 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 35.0 35.0 30.0 Chileka 30.0 Kasungu Monthly air temperature [°C] Monthly air temperature [°C] 25.0 25.0 20.0 20.0 15.0 15.0 10.0 10.0 5.0 5.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec max-min LTA Last Year max-min LTA Last Year 35.0 30.0 Mzuzu Monthly air temperature [°C] 25.0 20.0 15.0 10.0 5.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec max-min LTA Last Year Figure VI: TEMP monthly averages [°C]. Monthly average shown as solid line; min/max monthly values as boundary lines; last 12 months shown in red. © 2018 Solargis page 60 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Frequency of occurrence of GHI and DNI daily model values for a period 1994 to 2017 The histograms below show occurrence statistics of daily values derived from the satellite-based time series for GHI and DNI. The time covered in the graphs below is 24 complete calendar years (1994 to 2017). The occurrence is calculated separately for each month. 12 20 20 12 12 20 January February March Percentage of days 10 10 10 15 15 15 8 8 8 6 10 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 12 12 20 20 12 April May June Percentage of days 10 10 10 15 15 15 8 8 8 10 6 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 12 12 20 12 20 July August September Percentage of days 10 10 10 15 15 15 8 8 8 6 10 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 12 12 20 20 12 October November December Percentage of days 10 10 10 15 15 15 8 8 8 10 6 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Daily sum of GHI [kWh/m2] Daily sum of GHI [kWh/m2] Daily sum of GHI [kWh/m2] Figure VII: Histograms of daily summaries of Global Horizontal Irradiation in Chileka. 12 20 12 20 20 12 January February March Percentage of days 10 10 10 15 15 15 8 8 8 6 10 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 20 12 12 20 20 12 April May June Percentage of days 10 10 10 15 15 15 8 8 8 10 6 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 12 20 20 12 12 20 July August September Percentage of days 10 10 10 15 15 15 8 8 8 6 10 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 20 12 12 20 20 12 October November December Percentage of days 10 10 10 15 15 15 8 8 8 10 6 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Daily sum of GHI [kWh/m2] Daily sum of GHI [kWh/m2] Daily sum of GHI [kWh/m2] Figure VIII: Histograms of daily summaries of Global Horizontal Irradiation in Kasungu. © 2018 Solargis page 61 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 12 20 12 20 20 12 January February March Percentage of days 10 10 10 15 15 15 8 8 8 6 10 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 12 12 20 12 20 April May June Percentage of days 10 10 10 15 15 15 8 8 8 10 6 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 12 12 20 12 20 July August September Percentage of days 10 10 10 15 15 15 8 8 8 6 10 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 12 20 12 12 20 October November December Percentage of days 10 10 10 15 15 15 8 8 8 10 6 6 10 6 10 4 4 4 5 5 5 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Daily sum of GHI [kWh/m2] Daily sum of GHI [kWh/m2] Daily sum of GHI [kWh/m2] Figure IX: Histograms of daily summaries of Global Horizontal Irradiation in Mzuzu. Figures VII to IX show histograms of daily GHI summaries for each month as calculated from Solargis time series representing the years 1994 to 2017. The distribution of daily values is not symmetric: median is drawn by the vertical line, and percentiles P10, P25, and P75, and P90 are displayed with dark grey and light grey color bands, respectively. The percentiles P10 and P90 show 80% occurrence of daily values within each month and percentiles P25 and P75 show 50% occurrence. 8 8 8 January February March Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 April May June Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 July August September Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 October November December Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Daily sum of DNI [kWh/m2] Daily sum of DNI [kWh/m2] Daily sum of DNI [kWh/m2] Figure X: Histograms of daily summaries of Direct Normal Irradiation in Chileka. © 2018 Solargis page 62 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 8 8 8 January February March Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 April May June Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 July August September Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 October November December Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Daily sum of DNI [kWh/m2] Daily sum of DNI [kWh/m2] Daily sum of DNI [kWh/m2] Figure XI: Histograms of daily summaries of Direct Normal Irradiation in Kasungu. 8 8 8 January February March Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 April May June Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 July August September Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 8 8 October November December Percentage of days 6 6 6 4 4 4 2 2 2 0 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Daily sum of DNI [kWh/m2] Daily sum of DNI [kWh/m2] Daily sum of DNI [kWh/m2] Figure XII: Histograms of daily summaries of Direct Normal Irradiation in Mzuzu. Figures X to XII show histograms of daily DNI summaries for each month as calculated from Solargis time series representing the years 1994 to 2017. The distribution of daily values is not symmetric: median is drawn by the vertical line, and percentiles P10, P25, and P75, and P90 are displayed with dark grey and light grey color bands, respectively. The percentiles P10 and P90 show 80% occurrence of daily values within each month and percentiles P25 and P75 show 50% occurrence. © 2018 Solargis page 63 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Frequency of occurrence of GHI and DNI 15-minute model values for a period 1994 to 2017 The histograms below show occurrence statistics of 15-minute values derived from the satellite-based site- adapted time series for GHI and DNI. The time covered in the graphs below is 24 complete calendar years (1994 to 2017). The occurrence is calculated separately for each month. Figure XIII: Histograms and cumulative distribution function of 15-minute GHI in Chileka Figure XIV: Histograms and cumulative distribution function of 15-minute GHI in Kasungu © 2018 Solargis page 64 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure XV: Histograms and cumulative distribution function of 15-minute GHI in Mzuzu Figures XIII to XV show monthly histograms (bars) and cumulative distribution (line) of 15-minute GHI values, calculated from site-adapted Solargis time series. The values represent the occurrence of GHI values within 50 W/m2 bins, ranging from 0 to 1200 W/m2. Figure XVI: Histograms and cumulative distribution function of 15-minute DNI in Chileka © 2018 Solargis page 65 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure XVII: Histograms and cumulative distribution function of 15-minute DNI in Kasungu Figure XVIII: Histograms and cumulative distribution function of 15-minute DNI in Mzuzu Figures XVI to XVIII show monthly histograms (bars) and cumulative distribution (line) of 15-minute DNI values, calculated from Solargis time series. The values represent the occurrence of DNI values within 50 W/m2 bins, ranging from 0 to 1200 W/m2. © 2018 Solargis page 66 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Frequency of occurrence of measured and model GHI and DNI during two years of measurements Figures XIX to XXIV show histograms comparing the measured values with the model GHI and DNI data. The period covered in these histogram is last 24-months (two full years of data, i.e. from 1 Apr 2016 to 31 Mar 2018): • 1-minute measured vs. 15-min satellite-based model values • 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values • Daily measured (aggregated from 1-min) vs. daily satellite-based model values Aggregation process deals with the missing values in the ground measurement in three steps: 1. Only those 1-minute measured data values that passed through quality control (Chapter 3.3) is taken into account (satellite time series data does not have gaps.); 2. Aggregation of 1-minute measured data values into 15-minute slots (equivalent to satellite time slots) is applied if more than 15 valid data-points is available, otherwise the 15-minute data slot is ignored in further statistical comparison; 3. Daily aggregation of measured data represents the same 15-minute time slots in a day (passing through the two steps above), as those in the satellite-based data. Incorrect data slots found in the measurements are excluded in both the measured and model data. Figure XIX: Measured vs. satellite-based GHI values in Chileka 1-minute measured vs. 15-min satellite-based values. 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values Daily measured (aggregated from 1-min) vs. daily satellite-based values Figure XX: Measured vs. satellite-based GHI values in Kasungu 1-minute measured vs. 15-min satellite-based values. 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values Daily measured (aggregated from 1-min) vs. daily satellite-based values © 2018 Solargis page 67 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure XXI: Measured vs. satellite-based GHI values in Mzuzu 1-minute measured vs. 15-min satellite-based values. 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values Daily measured (aggregated from 1-min) vs. daily satellite-based values Figure XXII: Measured vs. satellite-based DNI values in Chileka 1-minute measured vs. 15-min satellite-based values. 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values Daily measured (aggregated from 1-min) vs. daily satellite-based values Figure XXIII: Measured vs. satellite-based DNI values in Kasungu 1-minute measured vs. 15-min satellite-based values. 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values Daily measured (aggregated from 1-min) vs. daily satellite-based values Figure XXIV: Measured vs. satellite-based DNI values in Mzuzu 1-minute measured vs. 15-min satellite-based values. 15-minute measured (aggregated from 1-min) vs. 15-min satellite-based values Daily measured (aggregated from 1-min) vs. daily satellite-based values © 2018 Solargis page 68 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Frequency of occurrence of GHI and DNI ramps Figures XXV to XXXVIII show histograms of instantaneous changes (ramps) calculated from the measurements and compared to the instantaneous changes calculated from the model data. Figures show both negative (-) and positive (+) changes. Two versions for GHI and DNI are shown: • Ramps calculated from 1-minute measured values compared to ramps calculated from 15-minute satellite-based data (figure on the left) • Ramps calculated from 15-minute aggregated valid measurement compared to ramps calculated from 15-minute satellite-based data (figure on the right). Occurrence of gaps in the measurements is managed in the same way as described about in this Chapter: 1. For measurements, only those 1-minute data values (measurements) that passed through quality control (Chapter 3.3) is taken into account (satellite time series does not have gaps.); 2. For measurements, the aggregation (averaging) of 1-minute measured data values into 15-minute slots (equivalent to satellite time slots) is applied if more than 15 valid data-points is available, otherwise the 15-minute data slot is ignored in further statistical comparison; Figure XXV: 1-minute and 15-minute GHI ramps (measured and satellite data) at Chileka. Figure XXVI: 1-minute and 15-minute GHI ramps (measured and satellite data) at Kasungu © 2018 Solargis page 69 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure XXVII: 1-minute and 15-minute GHI ramps (measured and satellite data) at Mzuzu Figure XXVIII: 1-minute and 15-minute DNI ramps (measured and satellite data) at Chileka Figure XXIX: 1-minute and 15-minute DNI ramps (measured and satellite data) at Kasungu © 2018 Solargis page 70 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure XXX: 1-minute and 15-minute DNI ramps (measured and satellite data) at Mzuzu © 2018 Solargis page 71 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 LIST OF FIGURES Figure 2.1: Position of solar meteorological stations in Malawi ............................................................................... 11 Figure 2.2: Horizon at Chileka meteorological station............................................................................................... 12 Figure 2.3: Horizon at Kasungu meteorological station ............................................................................................ 12 Figure 2.4: Horizon at Mzuzu meteorological station ................................................................................................ 12 Figure 3.1 Results of GHI and DNI quality control in Chileka. ................................................................................... 17 Figure 3.2 Insufficient cleaning – skewed DIF and DNI profiles. .............................................................................. 17 Figure 3.3 Comparison of solar irradiance: measured and calculated. .................................................................... 18 Figure 3.4 Results of GHI and DNI quality control in Kasungu. ................................................................................. 20 Figure 3.5 Effect of dew – degraded GHI CMP10 (yellow) readings. ....................................................................... 21 Figure 3.6 Difference between GHI from CMP10 and RSR 2 - Kasungu. .................................................................. 21 Figure 3.7 Results of GHI and DNI quality control − Mzuzu. ..................................................................................... 23 Figure 3.8 Systematic shading – drop of DNI in Mzuzu. ........................................................................................... 24 Figure 3.9 Effect of dew – degraded GHI CMP10 readings. ..................................................................................... 24 Figure 3.10 Difference between GHI from CMP10 and RSR 2 − Mzuzu. .................................................................. 24 Figure 3.11 Pyranometers replaced at Chileka station. ............................................................................................. 27 Figure 3.12 New LI-COR installed and levelled. .......................................................................................................... 27 Figure 4.1 Solar meteorological stations in the context of global horizontal irradiation. ....................................... 29 Figure 4.2: Correction of DNI and GHI hourly values for Chileka. ............................................................................. 33 Figure 4.3: Correction of DNI and GHI hourly values for Kasungu ............................................................................ 34 Figure 4.4: Correction of DNI and GHI hourly values for Mzuzu................................................................................ 35 Figure 4.5: Comparison of Solargis original and site-adapted data for Kasungu site. ............................................ 36 Figure 5.1: Scatterplots of air temperature at 2 m at Chileka meteorological station. ........................................... 38 Figure 5.2: Scatterplots of air temperature at 2 m at Mzuzu meteorological station.............................................. 39 Figure 5.3: Scatterplots of air temperature at 2 m at Kasungu meteorological station. ......................................... 39 Figure 5.4: Scatterplots of relative humidity at 2 m at Chileka meteorological station........................................... 40 Figure 5.5: Scatterplots of relative humidity at 2 m at Mzuzu meteorological station. ........................................... 41 Figure 5.6: Scatterplots of relative humidity at 2 m at Kasungu meteorological station. ....................................... 41 Figure 5.7: Scatterplots of wind speed at Chileka meteorological station. .............................................................. 42 Figure 5.8: Scatterplots of wind speed at Mzuzu meteorological station. ............................................................... 43 Figure 5.9: Scatterplots of wind speed at Kasungu meteorological station. ........................................................... 43 Figure 6.1: Expected Pxx values for GHI at Chileka site ............................................................................................ 49 Figure 6.2: Expected Pxx values for DNI at Chileka site ............................................................................................ 49 Figure 7.1: GHI monthly values derived from time series and TMY P50 and P90 ................................................... 53 Figure 7.2: DNI monthly values derived from time series and TMY P50 and P90 ................................................... 53 Figure 7.3: DIF monthly values derived from time series and TMY P50 and P90 .................................................... 53 Figure 7.4: TEMP monthly values derived from time series and TMY P50 and P90 ............................................... 54 Figure 7.5: Seasonal profile of GHI, DNI and DIF for Typical Meteorological Year P50 .......................................... 54 © 2018 Solargis page 72 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Figure 7.6: Snapshot of Typical Meteorological Year for P50 for Chileka ................................................................ 55 Figure I: Interannual variability of site-adapted yearly GHI [kWh/m2]........................................................................ 57 Figure II: Interannual variability of site-adapted yearly DNI [kWh/m2]....................................................................... 57 Figure III: Interannual variability of yearly TEMP [°C]. ................................................................................................ 58 Figure IV: GHI monthly averages [kWh/m2]. ............................................................................................................... 59 Figure V: DNI monthly averages [kWh/m2].................................................................................................................. 59 Figure VI: TEMP monthly averages [°C]. ..................................................................................................................... 60 Figure VII: Histograms of daily summaries of Global Horizontal Irradiation in Chileka. ......................................... 61 Figure VIII: Histograms of daily summaries of Global Horizontal Irradiation in Kasungu. ...................................... 61 Figure IX: Histograms of daily summaries of Global Horizontal Irradiation in Mzuzu............................................. 62 Figure X: Histograms of daily summaries of Direct Normal Irradiation in Chileka. ................................................. 62 Figure XI: Histograms of daily summaries of Direct Normal Irradiation in Kasungu. .............................................. 63 Figure XII: Histograms of daily summaries of Direct Normal Irradiation in Mzuzu. ................................................ 63 Figure XIII: Histograms and cumulative distribution function of 15-minute GHI in Chileka .................................... 64 Figure XIV: Histograms and cumulative distribution function of 15-minute GHI in Kasungu ................................. 64 Figure XV: Histograms and cumulative distribution function of 15-minute GHI in Mzuzu ...................................... 65 Figure XVI: Histograms and cumulative distribution function of 15-minute DNI in Chileka.................................... 65 Figure XVII: Histograms and cumulative distribution function of 15-minute DNI in Kasungu ................................ 66 Figure XVIII: Histograms and cumulative distribution function of 15-minute DNI in Mzuzu .................................. 66 Figure XIX: Measured vs. satellite-based GHI values in Chileka ............................................................................... 67 Figure XX: Measured vs. satellite-based GHI values in Kasungu .............................................................................. 67 Figure XXI: Measured vs. satellite-based GHI values in Mzuzu ................................................................................ 68 Figure XXII: Measured vs. satellite-based DNI values in Chileka .............................................................................. 68 Figure XXIII: Measured vs. satellite-based DNI values in Kasungu ........................................................................... 68 Figure XXIV: Measured vs. satellite-based DNI values in Mzuzu .............................................................................. 68 Figure XXV: 1-minute and 15-minute GHI ramps (measured and satellite data) at Chileka. .................................. 69 Figure XXVI: 1-minute and 15-minute GHI ramps (measured and satellite data) at Kasungu ................................ 69 Figure XXVII: 1-minute and 15-minute GHI ramps (measured and satellite data) at Mzuzu .................................. 70 Figure XXVIII: 1-minute and 15-minute DNI ramps (measured and satellite data) at Chileka ................................ 70 Figure XXIX: 1-minute and 15-minute DNI ramps (measured and satellite data) at Kasungu ................................ 70 Figure XXX: 1-minute and 15-minute DNI ramps (measured and satellite data) at Mzuzu ..................................... 71 © 2018 Solargis page 73 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 LIST OF TABLES Table 1.1 Delivered data characteristics ............................................................................................................... 9 Table 1.2 Parameters included in the site-adapted time series and TMY data (hourly time step) .................. 10 Table 2.1 Overview information on the solar meteorological station locations ............................................... 11 Table 3.1 Overview information on measurement stations operated in the region .......................................... 13 Table 3.2 Instruments installed at the solar meteorological stations ............................................................... 13 Table 3.3 Technical parameters and accuracy class of the instruments at Tier 1 and Tier 2 stations........... 13 Table 3.4 Overview information on solar meteorological stations operating in the region ............................. 14 Table 3.5 Period of measurements analyzed in this report ................................................................................ 14 Table 3.6 Data recovery statistics of the measurement campaign ................................................................... 14 Table 3.7 Meteorological stations maintenance ................................................................................................. 15 Table 3.8 Occurrence of data readings for Chileka meteorological station ...................................................... 16 Table 3.9 Excluded ground measurements after quality control (Sun above horizon) in Chileka ................... 16 Table 3.10 Quality control summary - Chileka ....................................................................................................... 18 Table 3.11 Occurrence of data readings for Kasungu meteorological station ................................................... 19 Table 3.12 Excluded ground measurements after quality control (Sun above horizon) in Kasungu ................. 19 Table 3.13 Quality control summary - Kasungu..................................................................................................... 21 Table 3.14 Occurrence of data readings for Mzuzu meteorological station ....................................................... 22 Table 3.15 Excluded ground measurements after quality control (Sun above horizon) in Mzuzu .................... 22 Table 3.16 Quality control summary - Mzuzu ........................................................................................................ 25 Table 3.17 Summary of replaced solar instruments ............................................................................................. 26 Table 4.1 Input data used in the Solargis and related GHI and DNI outputs for Malawi .................................. 28 Table 4.2 Direct Normal Irradiance: bias and KSI before and after model site-adaptation .............................. 32 Table 4.3 Global Horizontal Irradiance: bias and KSI before and after model site-adaptation ........................ 32 Table 4.4 Direct Normal Irradiance: RMSD before and after model site-adaptation ........................................ 32 Table 4.5 Global Horizontal Irradiance: RMSD before and after model site-adaptation .................................. 32 Table 4.6 Comparison of long-term average of yearly summaries of original and site-adapted values ......... 36 Table 5.1: Original source of Solargis meteorological data: models MERRA-2, CFSR and CFSv2. .................. 37 Table 5.2 Solargis meteorological parameters delivered for 3 meteo sites ..................................................... 37 Table 5.3 Air temperature at 2 m: Accuracy indicators of the model outputs [ºC]. .......................................... 38 Table 5.4 Relative humidity: Accuracy indicators of the model outputs [%]...................................................... 40 Table 5.5 Wind speed: accuracy indicators of the model outputs [m/s]. .......................................................... 42 Table 5.6 Expected uncertainty of modelled meteorological parameters at the project sites. ....................... 44 Table 6.1 Uncertainty of the model estimates for original and site-adapted annual long-term values ........... 46 Table 6.2 Annual GHI that should be exceeded with 90% probability in the period of 1 to 10 (25) years ....... 46 Table 6.3 Annual DNI that should be exceeded with 90% probability in the period of 1 to 10 (25) years. ...... 47 Table 6.4 Combined probability of exceedance of annual GHI for uncertainty of the estimate ±4.0%. .......... 48 Table 6.5 Combined probability of exceedance of annual DNI for uncertainty of the estimate ±5.5%. .......... 48 © 2018 Solargis page 74 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 Table 7.1 Delivered data, and its key characteristics .......................................................................................... 50 Table 7.2 Parameters in the delivered site-adapted time series and TMY data (hourly time step) ................. 50 Table 7.3 Monthly and yearly long-term GHI averages as calculated from time series and from TMY .......... 51 Table 7.4 Monthly and yearly long-term DNI averages as calculated from time series and from TMY .......... 52 Table 7.5 Monthly and yearly long-term DIF averages as calculated from time series and from TMY ........... 52 Table 7.6 Monthly and yearly long-term TEMP averages as calculated from time series and from TMY ...... 52 © 2018 Solargis page 75 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 REFERENCES [1] NREL, 1993. User’s Manual for SERI_QC Software-Assessing the Quality of Solar Radiation Data. NREL/TP- 463-5608. Golden, CO: National Renewable Energy Laboratory. [2] Younes S., Claywell R. and Muneer T, 2005. Quality control of solar radiation data: Present status and proposed new approaches. Solar Energy 30, 1533-1549. [3] Perez R., Cebecauer T., Šúri M., 2013. Semi-Empirical Satellite Models. In Kleissl J. (ed.) Solar Energy Forecasting and Resource Assessment. Academic press. http://www.sciencedirect.com/science/article/pii/B9780123971777000024 [4] Cebecauer T., Šúri M., Perez R., High performance MSG satellite model for operational solar energy applications. ASES National Solar Conference, Phoenix, USA, 2010. http://solargis.com/assets/publication/2010/Cebecauer-Suri-Perez--ASES2010--High-performance-MSG- satellite-model-for-operational-solar-energy-applications.pdf [5] Šúri M., Cebecauer T., Perez P., 2010. Quality Procedures of Solargis for Provision Site-Specific Solar Resource Information. Conference SolarPACES 2010, September 2010, Perpignan, France. http://solargis.com/assets/publication/2010//Suri-Cebecaue--Perez-SolarPACES2010--Quality-procedures- of-solargis-for-provision-site-specific-solar-resource-information.pdf [6] Cebecauer T., Suri M., Gueymard C., Uncertainty sources in satellite-derived Direct Normal Irradiance: How can prediction accuracy be improved globally? Proceedings of the SolarPACES Conference, Granada, Spain, 20-23 Sept 2011. http://solargis.com/assets/publication/2011/Cebecauer-Suri-Gueymard--SolarPACES2011--Uncertainty- sources-in-satellite-derived-direct-normal-irradiance--How-can-prediction-accuracy-be-improved-globally.pdf [7] Suri M., Cebecauer T., 2014. Satellite-based solar resource data: Model validation statistics versus user’s uncertainty. ASES SOLAR 2014 Conference, San Francisco, 7-9 July 2014. http://solargis.com/assets/publication/2014//Suri-Cebecauer--ASES-Solar2014--Satellite-Based-Solar- Resource-Data--Model-Validation-Statistics-Versus-User-Uncertainty.pdf [8] Ineichen P., A broadband simplified version of the Solis clear sky model, 2008. Solar Energy, 82, 8, 758-762. [9] Benedictow A. et al. 2012. Validation report of the MACC reanalysis of global atmospheric composition: Period 2003-2010, MACC-II Deliverable D83.1. [10] Molod, A., Takacs, L., Suarez, M., and Bacmeister, J., 2015: Development of the GEOS-5 atmospheric general circulation model: evolution from MERRA to MERRA2, Geosci. Model Dev., 8, 1339-1356, doi:10.5194/gmd-8- 1339-2015 [11] GFS model. http://www.nco.ncep.noaa.gov/pmb/products/gfs/ [12] CFSR model. https://climatedataguide.ucar.edu/climate-data/climate-forecast-system-reanalysis-cfsr/ [13] CFSv2 model http://www.cpc.ncep.noaa.gov/products/CFSv2/CFSv2seasonal.shtml [14] Cano D., Monget J.M., Albuisson M., Guillard H., Regas N., Wald L., 1986. A method for the determination of the global solar radiation from meteorological satellite data. Solar Energy, 37, 1, 31–39. [15] Perez R., Ineichen P., Maxwell E., Seals R. and Zelenka A., 1992. Dynamic global-to-direct irradiance conversion models. ASHRAE Transactions-Research Series, pp. 354-369. [16] Perez, R., Seals R., Ineichen P., Stewart R., Menicucci D., 1987. A new simplified version of the Perez diffuse irradiance model for tilted surfaces. Solar Energy, 39, 221-232. [17] Ruiz-Arias J.A., Cebecauer T., Tovar-Pescador J., Šúri M., 2010. Spatial disaggregation of satellite-derived irradiance using a high-resolution digital elevation model. Solar Energy, 84,1644-57. http://www.sciencedirect.com/science/article/pii/S0038092X10002136 [18] Zelenka A., Perez R., Seals R., Renne D., 1997. Effective Accuracy of Satellite-Derived Hourly Irradiances. Theoretical Appl. Climatology, 62, 199-207. [19] Espinar B., Ramírez L., Drews A., Beyer H. G., Zarzalejo L. F., Polo J., Martín L., Analysis of different comparison parameters applied to solar radiation data from satellite and German radiometric stations. Solar Energy, 83, 1, 118-125, 2009. © 2018 Solargis page 76 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 [20] Lohmann S., Schillings C., Mayer B., Meyer R., 2006. Long-term variability of solar direct and global radiation derived from ISCCP data and comparison with reanalysis data, Solar Energy, 80, 11, 1390-1401. [21] Gueymard C., Solar resource assessment for CSP and CPV. Leonardo Energy webinar, 2010. http://www.leonardo-energy.org/webfm_send/4601 [22] Cebecauer T., Šúri M., 2015. Typical Meteorological Year Data: Solargis Approach. Energy Procedia, Volume 69, 1958-1969. http://dx.doi.org/10.1016/j.egypro.2015.03.195 © 2018 Solargis page 77 of 79 Annual Solar Resource Report for Solar Meteorological Stations in Malawi after completion of 24 months of measurements Solargis reference No. 141-07/2018 SUPPORT INFORMATION Background on Solargis Solargis is a technology company offering energy-related meteorological data, software and consultancy services to solar energy. We support industry in the site qualification, planning, financing and operation of solar energy systems for more than 18 years. We develop and operate a new generation high-resolution global database and applications integrated within Solargis® information system. Accurate, standardised and validated data help to reduce the weather-related risks and costs in system planning, performance assessment, forecasting and management of distributed solar power. Solargis is ISO 9001:2015 certified company for quality management. This report has been prepared by Tomas Cebecauer, Marcel Suri, Branislav Schnierer, Nada Suriova, Juraj Betak, Artur Skoczek and Daniel Chrkavy from Solargis All maps in this report are prepared by Solargis Solargis s.r.o., Mytna 48, 811 07 Bratislava, Slovakia Reference No. (Solargis): 141-07/2018 http://solargis.com © 2018 Solargis page 78 of 79