TECHNICAL GUIDELINES FOR Small Island Mapping with UAVs ACP-EU Natural Disaster Risk ReducƟon Program An iniƟaƟve of the African, Caribbean and Pacific Group, funded by the European Union and managed by GFDRR RIGHTS & PERMISSIONS The World Bank and Humanitarian OpenStreetMap Team support the free online communication and exchange of knowledge as the most effective way of ensuring that the fruits of research, economic and sector work, and development practice are made widely available, read, and built upon. It is therefore committed to open access, which, for authors, enables the widest possible © 2019 International Bank for Reconstruction and Development dissemination of their findings and, for researchers, readers, and / International Development Association or The World Bank users, increases their ability to discover pertinent information. The 1818 H Street NW, Washington DC, 20433 www.worldbank.org material in this work is made available under a Creative Commons BY 4.0 License. You are encouraged to share and adapt this This work is a product of independent consultants working content for any purpose, including commercial use, as long as full in collaboration with The World Bank and Humanitarian attribution to this work is given. OpenStreetMap Team. Part of the funding that supported this work was provided by the Australian Department of Foreign Any queries on rights and licenses, including subsidiary rights, Affairs and Trade (DFAT) through the Pacific Humanitarian should be addressed to the Office of the Publisher, The World Challenge. The findings, interpretations, and conclusions Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202- expressed in this work do not necessarily reflect the views 522-2422; e-mail: pubrights@worldbank.org. of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The ATTRIBUTION boundaries, colors, denominations, and other information Please cite the work as follows: World Bank and Humanitarian shown on any map in this work do not imply any judgment on OpenStreetMap Team (2019). Technical Guidelines for Small Island the part of The World Bank concerning the legal status of any Mapping with UAVs. Washington, DC. License: Creative Commons territory or the endorsement or acceptance of such boundaries. Attribution CC BY 4.0. Contents 1. OBJECTIVE OF THIS GUIDANCE NOTE 9 2. UAVS AS A GAME CHANGER FOR PICS 13 3. SECURING UAV FLIGHT AUTHORIZATION AND PERMITS IN THE PACIFIC 15 4. DETERMINING WHETHER THE UAV IS THE RIGHT MAPPING TOOL 21 5. CHOOSING THE RIGHT UAV PLATFORM 25 6. CHOOSING THE SENSOR 31 Preferred camera specifications 31 Tagging images using GPS 33 Other important considerations 33 7. PREPARING FOR THE UAV MISSION 35 Time management 35 Takeoff and landing sites 37 Defining the area of interest and coordination 37 Flight planning 38 Target landscape limitations 40 Ground control point target placement 40 2 8. MANAGING WORK FLOW IN THE FIELD 43 9. AFTER THE FLIGHT: PROCESSING THE DATA 47 Processing software and hardware 47 Post-processing work flow to create orthomosaics 49 Data processing following 2018 TC Gita Response Survey 49 Sharing Data and Imagery 50 10. HANDLING PROJECT MANAGEMENT 53 11. CASE STUDY: POST–TROPICAL CYCLONE GITA MISSION IN TONGA 57 Collection of post-cyclone UAV data 57 Application of post-cyclone UAV data 59 12. CONCLUSIONS 63 REFERENCES AND OTHER RESOURCES 65 ANNEX 1. APPLYING FOR APPROVALS IN FIJI AND TONGA FOR OCTOBER 2017 UAV FIELD TESTING 67 Technical Guidelines for Small Island Mapping with UAVs 3 BOXES Box 1. New Zealand CAA Part 101 Rules FIGURES Figure 1. Decision tree showing the approvals required to fly under various conditions. Source: Airshare, n.d. Figure 2. Launching the Goshawk III Surveyor UAV by V-TOL Aerospace in Tonga. Figure 3. Advantages of rotary (left) and fixed-wing (right) small UAVs. Figure 4. Price, range, and ease-of-use comparison of various UAV platforms available on the market. Figure 5. Time needed for the steps of a UAV deployment as a share of the whole process. As indicated by the red arrow, the actual flying of the platform occupies only a small proportion of time. Figure 6. A large, open field selected to launch and land the fixed-wing UAV in Tongatapu, Tonga. Figure 7. Mission planning for the LiDAR survey in Fiji during the October 2017 field testing. Figure 8. Recommended forward and side overlaps to create a mosaic with good positional accuracy. Figure 9. Mission planning. Figure 10. Placing of a ground control point. The accurate coordinates of this GCP are surveyed on the spot using high-accuracy GPS. Figure 11. Checking conditions and following safety measures in the field. Figure 12. Launch of a multirotor UAV, which requires less open space than for a fixed wing. Figure 13. User interface of OpenAerialMap.org showing UAV images that are available for Tonga. Figure 14. Image showing location of the structures assessed as damaged (in orange) and destroyed (in red). Figure 15. Images showing pre- and post-event status of the same structures. Figure 16. A house destroyed by Tropical Cyclone Gita in Longolongo village, Tongatapu, Tonga. TABLES Table 1. Five Data Types Generated During UAV Field Testing in Tonga and Fiji Table 2. Comparison of Survey Mapping Methods Table 3. Comparison of Popular UAVs for Mapping Table 4. Comparison of RGB Cameras Used on Mapping UAVs Table 5. Comparison of Popular UAV Image Processing Software Programs Table 6. Sample Sensor Types and Experts Needed for Various UAV Data Applications Table 7. Time Needed for Obtaining Approvals in Fiji 4 ABBREVIATIONS HOT Humanitarian OpenStreetMap Team 2D two-dimensional IMU inertial measurement unit ADS-B Automatic Dependent Surveillance– LiDAR Light Detection and Ranging Broadcast LiPo lithium polymer AGL above ground level NIR near infrared AOI area of interest NOTAM Notice to Airmen AT automatic triangulation OAM OpenAerialMap ATC air traffic control OSM OpenStreetMap CAA Civil Aviation Authority PacDID Pacific Drone Imagery Dashboard CV computer vision PIC Pacific Island Country DSLR digital single-lens reflex PPK Post-Processed Kinematic DSM digital surface model RGB red-green-blue DTM digital terrain model RTK Real-Time Kinematic EXIF exchangeable image file format SfM Structure from Motion GCP ground control point SIDS Small Island Developing States GeoTIFF Georeferenced Tagged Image File Format SSD solid-state drive GIS geographic information system UAV unmanned aerial vehicle GPS Global Positioning System UAViators Humanitarian UAV Network GPU graphics processing unit ACKNOWLEDGMENTS This document was prepared by Cristiano Giovando, Patrick Meier, Joseph Mulhausen, and Denise Soesilo, with oversight from Keiko Saito and Michael Bonte- Grapentin. V-TOL Aerospace provided substantial inputs based on field work carried out in Tonga and Fiji. Illustrations by Elena Mirandola. Technical input on an earlier version by Christopher Lippitt. Report Design by Heidi Romano. 5 Executive Summary Image acquisition and surveying using unmanned • Platforms that are suitable for the mapping aerial vehicles (UAVs) is a very promising technology needs and local conditions in small island states for Small Island Developing States (SIDS). UAVs should be selected, if the budget permits, several can be a relatively low-cost data collection tool at platforms can be used for various specialized the surveying scales often needed in small island tasks (page 28). contexts. Further, UAVs can capture thousands of • Flying the UAV is only a very small portion of the images in a single flight and provide greater detail time needed. Most of the time required is taken up than satellites or even manned aircraft. The World by planning, obtaining permission to fly, and post- Bank and Humanitarian OpenStreetMap Team (HOT) processing of the data captured (page 36). compiled this guidance note to document experience and best practices in the use and operation of UAVs • To ensure safety during flying operations, for economic development in SIDS. Many of the regulations and air traffic control procedures need lessons presented in this guidance note stem from to be clear and understood and must be rehearsed the UAV4Resilience organized by the World Bank before emergency conditions arise (page 15). (World Bank 2017b) and from experiences with Pacific • To expedite licensing and regulatory processes Drone Imagery Dashboard (PacDID) deployments in and reduce costs, it helps to use local service the Pacific islands (HOT 2016). This report is intended providers that specialize in the mapping areas of for local technological agencies of island nations interest within the country. Capacity building of in- that work to operationalize UAVs as a standard data country actors ensures that needed skill sets are collection tool. available on short notice (page 54). The report offers the following key messages: • Power and battery management for UAVs and for ground control stations can be a challenge in • UAVs can be used to collect information for the field if not planned thoroughly in advance of disaster risk reduction and response. For both activities (page 53). these tasks, it is vitally important that high-quality baseline data be collected and made available before a disaster strikes as part of preparedness and capacity-building work (page 57). Technical Guidelines for Small Island Mapping with UAVs 7 1. Objective of this guidance note The objective of this guidance note is to establish key Lessons learned and best practices derived from principles for end use of unmanned aerial vehicles these exercises are offered throughout this note (UAVs) in a Pacific Island Country (PIC) context. to help the reader better understand how most The document addresses how, when, and for what effectively to use UAVs for survey mapping. Other applications UAVs should be used. For pilots who are applications of UAVs, such as for cargo delivery, are new to UAV mapping, it also provides instructions beyond the scope of this note. and recommendations to ensure that imagery of high Flight itself constitutes only a small portion of the quality is generated for accurate integration with entire UAV operation (as explained in chapter 7). Post- other geospatial layers. flight data processing and data sharing with end users The ultimate goal is to establish UAVs as the principal represent a much larger share of the operation, and data collection and survey mapping instrument for they present huge challenges given the sheer volume Small Island Developing States in the Pacific region of data that needs to be processed;2 challenges are and beyond. As the principles and best practices even greater if time is limited. Important lessons on described in this document are applicable globally, the data sharing, learned from the Pacific Drone Imagery plan is for this document to continue to evolve and for Dashboard (PacDID) project,3 have been integrated its online version (https://docs.openaerialmap.org/ in this guidance note. Efficient access to imagery can uav-guidelines) to reflect the latest in UAV technology. lead to improved mapping work flows for disaster risk management and other applications. To take stock of the current operating conditions in the Pacific region, field tests were conducted in Tonga and Fiji in October 2017. Five survey mapping challenges were carried out to generate five types of typical geospatial data products that are high in demand (shown in table 1).1 1. Details on the field tests carried out to generate the five types of data products are in Table1. 2. The data size often runs into hundreds of gigabytes. 3. PacDID is a platform that leverages the OpenAerialMap (OAM) concept, making open imagery collected by satellites and UAVs easily available and accessible, particularly for the target audience in Pacific Island Countries. More information is available at Pacific Humanitarian Challenge (n.d.). Technical Guidelines for Small Island Mapping with UAVs 9 TABLE 1. Summary of Disaster Effects by Sector (T$ millions) SURVEY TYPE/ AREA OUTPUT/FILE OUTPUT EXAMPLE OUTPUT USE/APPLICATION UAV PLATFORM TYPE Small area 10 km2 2D orthorectified General mapping, including delineating building (Challenge 1) mosaic footprints and road networks; location and geometric features of bridges and other key Fixed-wing RGB and RGB+ NIR infrastructure, such as electricity/telephone (GeoTIFF) poles, power lines; river embankments, bridges and land use and land cover. If an NIR channel is available, classification of vegetated areas, identification of permanent and temporary water bodies, etc. are possible. Medium area 50 km2 2D orthorectified Same as Challenge 1. (Challenge 2) mosaic Fixed-wing with RGB RGB camera (GeoTIFF) Elevation 0.5 km2 Surface model Inputs for flood modeling, drainage design, (Challenge 3) transportation design, coastal engineering, Point cloud landslide modeling (susceptibility), and erosion Multirotor with (GeoTIFF) monitoring for slope stability studies. LiDAR sensor Real-time, beyond 0.5 km2 Video, photos, Near-real-time information collected by flying line of sight (10 km orthomosaic the UAV beyond line of sight for post-disaster (Challenge 4) from reconnaissance, situational awareness, damage (MP4, JPEG take off assessment, post-disaster mapping, etc. Delta wing fixed- GeoTIFF) location) wing Bathymetry 0.5 km2 Point cloud and Tsunami and storm surge modeling, coastal (Challenge 5) DEM inundation/erosion modeling, reef monitoring, etc. Multirotor (GeoTIFF) Source: UAV4Resilience 2017 Note: RGB = red-green-blue; NIR = near infrared; DEM = digital elevation model. The field tests used various types of UAVs. All orthomosaic outputs are available from OpenAerialMap.org.   11 2. UAVs as a game changer for PICs The Pacific region is one of the most disaster-prone Given islands’ sparse distribution, moreover, PICs may regions of the world. It is subject to a variety of not be captured by satellite imagery unless operators natural hazards, including floods, tropical cyclones, specifically prioritize them. Thus PICs must employ earthquakes, tsunamis, droughts, and volcanic alternative means for capturing very high-resolution eruptions. Geospatial data can play a key role in imagery data. monitoring hazard conditions on the ground, but UAVs may solve this issue for PICs. Unlike aerial unique data collection challenges exist for PICs. surveys using manned aircrafts, UAVs can be flown Because these counties consist of multiple small at very low cost by qualified personnel and have the islands or atolls that are often very remote from one flexibility to handle PICs’ various requirements. Now another, it difficult to collect geospatial data at the a mature technology, UAVs are potentially a game appropriate scale for analysis. changer that will allow high-resolution images of these In most countries, mapping is conducted using remote islands to be regularly and affordably captured satellite, aerial, or ground-captured data, or a by local experts. combination of these. However, in an island country   context, satellite images do not have the necessary spatial resolution (pixel size) to show details, as the islands are so small relative to the pixel size. Technical Guidelines for Small Island Mapping with UAVs 13 3. Securing UAV flight authorization and permits in the Pacific Flying UAVs requires a thorough understanding of The Airshare website5 offers an excellent entry point local and national civil aviation regulations. Many to learn about regulations in New Zealand’s controlled governments now provide specific provisions for UAV airspace. In addition to an online learning module, operations, including permits and licenses that must Airshare also provides a decision tree (Figure 1) that be obtained before even entering the country. To learn allows UAV pilots to navigate through the different about countries’ specific requirements and rules for conditions under which the UAV flight is being planned flying UAVs, the Global Drone Regulations Database is and thus identify the approvals required to fly under a good starting point.4 those conditions. New Zealand regulations are the most prevalent In many countries, flight authorizations are granted in the Pacific. Many countries in the region base relatively easily if flights are conducted within the their regulations on the New Zealand Civil Aviation visual line of sight in uncontrolled airspace where Authority (CAA) UAV regulatory framework (see Box there are least restrictions, and below 400 ft (or 1) and modify them for local contexts. Some Pacific 120 m). For the simple flight authorization, UAVs countries, such as Fiji, take a hybrid approach and should weigh less than 15 kg; should stay clear of all reference aspects of Australian regulations along manned aircraft, persons, and property; and should with New Zealand’s. In general, familiarity with New remain outside of airspace restricted areas and the Zealand or Australian regulations will be beneficial for 4 km radius of any aerodrome – flying under these UAV pilots planning to fly in the Pacific region. conditions makes the flight qualify under the Part 101 rules under New Zealand regulations. In Tonga, pilots operating under these conditions can fly once the UAV is registered with the Tongan CAA without requesting authorizations for each flight. 4. Global Drone Regulations Database, https://www.droneregulations.info. 5. Airshare, https://www.airshare.co.nz. Technical Guidelines for Small Island Mapping with UAVs 15 If a flight goes beyond the Part 101 rules (e.g. fly BOX 1. above 400 ft), the pilot and their organization must apply for Part 102 exposition. For the field testing in NEW ZEALAND CAA PART 101 Tonga, a Part 102 exposition was obtained to fly above RULES FOR PERSONS OPERATING 400 ft and in controlled airspace. Details on how to GYRO GLIDERS AND PARASAILS, obtain Part 102 expositions can be found on Airways UNMANNED AIRCRAFT, KITES, website6. AND ROCKETS To supplement these official regulations, the 1. They may not operate an aircraft that is Humanitarian UAV Network (UAViators), a global 25 kg or larger, and they must ensure that volunteer organization of humanitarian UAV pilots, it is safe to operate. developed a code of conduct and a set of UAV mission best practices (uavcode.org) that should be 2. They must at all times take all practicable incorporated when planning UAV mapping activities, steps to minimize hazards to persons, particularly in the context of humanitarian projects.9 property, and other aircraft. The permit application process in Fiji is described in annex 1. 3. They may fly only in daylight. 4. They must give way to all crewed aircraft. 5. They must be able to see the UAV with their own eyes (i.e., not through binoculars, a monitor, or smartphone) to ensure separation from other aircraft (or they may use an observer to do this in certain cases). 6. They must not fly their aircraft higher than 120 m (400 ft) AGL (unless certain conditions are met). 7. They must have knowledge of airspace restrictions that apply in the area in which they will operate. 8. They may not fly closer than 4 km to any aerodrome7 (unless certain conditions are met). 9. When flying in controlled airspace, they must obtain an air traffic control (ATC) clearance issued by Airways (via Airshare8). 10. They may not fly in special-use airspace (e.g., military operating areas or restricted areas) without the permission of the area’s 6. https://www.caa.govt.nz/unmanned-aircraft/ controlling authority. intro-to-part-102/ 7. See Airshare, “4 Things You Need to Know about Flying Near 11. They must have consent from anyone they an Aerodrome,” https://www.airshare.co.nz/must-know/ wish to fly above. things-to-know-flying-near-an-aerodrome. 8. See Airshare, “My Flights,” 12. They must have the consent of the property https://www.airshare.co.nz/my-flights/plan-a-flight. owner or person in charge of the area they 9. See UAViators (n.d.) for the code of conduct and UAViators wish to fly above. (2015) for a guide to best practices. 16 start here preparation Civil Aviation Rules Wh we you <15kg 15 weight Part 101 1 airworthiness Operate Authority person or (see Par knowledge airspace App Air kno req Wh the aero aerodrome A <4km You need to be the holder of, be supervised by the holder o pilot - A Part 61 Pilot’s Licence - A Part 149 Certificate - An approved organisation qualification - (Rule 101.205(a)(3)) - NOTE: Not required if sh Is your operation shielded? Yes No Is the aerodrome controlled? ANR 1.1 Yes No Air Traffic Aerodrome Control operator authorisation agreement required required Observer required airspace Are you flying >400ft AGL? Yes No CAA approval required airshare Are you flying during the day? Yes You must be time of operation within visual Iine of sight - 101.200 Do not conduct Give way create a to all manned hazard to aircraft persons or property Disclaimer: This diagram is provided for information purposes only and is not to be relied on as a substitute for a comprehensive knowledge of the relevant rules and regulations that apply to the operation of UAVs. It is the UAV operator’s responsibility to read, understand and operate any UAVs in accordance with the Civil Aviation Rules. e Use this quick guide to get started when thinking about purchasing or using a UAV V2.2 CAA Updates hat is the eight of ur UAV? 5-25kg >25kg Part 101.215 CAA e under the approval y of an official required r organisation rt 101.202) propriate rspace owledge quired here is nearest odrome? ANR 5.1 >4km , or to of: n pilot hielded Are you clear of a low flying zone? ANR 1.3 No Yes Are you clear CAA of restricted/ approval military required operating areas? ANR 1.5/1.13 No Yes Approval required from Are you administering flying <400ft authority AGL? No Yes Keep clear of class C &D Are you flying airspace in controlled airspace Part 71 No Yes Are you flying within a UAV designated danger area? Air Traffic Control authorisation required ANR 1.10 Yes No Ensure a NOTAM is submitted >24hrs to flight g No Are you flying outdoors? Yes No Fly within 100m of a structure Fly below the top of structure Remain clear Avoid operating of all manned above persons or aircraft property without prior consent go fly! POWERED BY AIRWAYS 4. Determining whether the UAV is the right mapping tool UAVs may not always be the best mapping tool for On the other hand, a UAV is preferable for mapping a a project. Before deciding whether to use a UAV, small footprint (e.g., small pockets of high-flood-risk a satellite, or other tool, the project’s data needs, areas, or a small and remote island community). More budget, and time frame must be understood. Table 2 often than not, a single method is not used exclusively; lists some key factors to consider when deciding on rather, the various survey methods are used in the tool to use. complement to one another. For example, a group of small islands can be surveyed using a satellite or full- When selecting a survey method, the scale and extent size plane every five years, complemented with local of the area of interest must be considered, along with UAV survey updates every six months. The important the technical constraints of the project, availability thing is to establish a strategy that best captures the of surveying equipment (to establish GCPs), and the necessary data in the most cost-effective way. method’s cost-effectiveness. For instance, to acquire baseline imagery of large areas at a resolution of 50 cm/pixel with a capture window of one year, satellites are most practical. Technical Guidelines for Small Island Mapping with UAVs 21 Figure 2. Launching the Goshawk III Surveyor UAV by V-TOL Aerospace in Tonga. Credit: UAV4Resilience 2017 22 TABLE 2. Comparison of Survey Mapping Methods SATELLITE AIRPLANE UAV Approximate area covered 10,000 km2 750 km2 10–25 km2 (for a UAV in a day equivalent to an eBee)a Detail level 30–50 cm/pixel > 6–30 cm/pixel 3–10 cm/pixel Cost per 10 km2 $$ $$$ $ Cost per 1 million km2 $ $$ $$$$ Time to deploy 24 hours–1 week 3 days 24 hours (provided flight permits have been granted) Ease of deployment Easy (once the satellite Medium Easy is in orbit) Blocked by clouds Yes Depends on altitude No (though may be blocked by fog and rain) Blocked by wind No Yes Yes Regulatory burden Low Medium–high High Note: The number of $ denote relative costs: least costly ($) to most costly ($$$$) The use of a high-end UAV will allow a larger area to be captured. In Tonga, after Cyclone Gita in February 2018, a. approximately 40 km2 was captured per day using a Goshawk by V-TOL Aerospace Australia; see Figure 2. Technical Guidelines for Small Island Mapping with UAVs 23 5. Choosing the Right UAV Platform If a UAV is the right tool, the next decision concerns The ultimate choice of platform (fixed-wing, rotary which platform to use. There are multiple factors to wing, or hybrid, size; propulsion system; etc.) really consider in making this choice: the purpose of the depends on one question: which platform can carry data capture, the weight of platform allowed by local the sensor necessary to collect the data required regulations under the conditions of the intended (that is, data of the desired coverage, resolution, and survey location, the need for weatherproofing, and type)? To ensure that the platform can perform as the project’s technical requirements (e.g., the type needed under the given conditions, any decision must of sensor/camera needed for the survey). be based on thorough research. The remainder of the chapter discusses key considerations in deciding The choice of platform has a number which UAV platform to use, including the type of of consequences: sensor (camera) needed for the survey. 1. It affects the area that can be surveyed (because In deciding on a UAV for mapping, a major different platforms fly at different altitudes and consideration is whether to use a rotary or fixed- have different ranges). wing platform. Figure 3 compares the advantages 2. It affects the potential downtime of the survey of the two. Rotary UAVs (multi-rotor drones10) are (because some platforms are more weatherproof popular and relatively cheap for mapping; this than others); category includes quadcopters (drones with four 3. It affects the areas where it can be operated propellers) that can be carried in a backpack-size (specifically the landing and takeoff areas); case. Their portability, price, and ease of use make them the obvious choice for those starting out in UAV 4. It affects the spatial resolution of the data, which mapping on very small survey sites. One downside can be adjusted using various sensors, lenses or of using small rotary multicopters is their limited by adjusting flight altitude above ground level. flight time, which is on average a maximum of 30 minutes, covering an area less than 0.5 km2 per flight. Small fixed-wing UAVs can usually fly longer by taking advantage of their gliding capability, and they fly significantly longer distances than rotary UAVs. A relatively new type of platform is the hybrid UAV combining both rotors and fixed wings. This design allows the UAV to take off and land vertically and make use of wings for extended range. Table 3 compares some popular portable UAVs for mapping. 10. The words Drones and UAV can be used interchangeably. Technical Guidelines for Small Island Mapping with UAVs 25 Figure 3. Advantages of rotary (left) and fixed-wing (right) small UAVs. Source: PacDID. 2017 26 TABLE 3. Comparison of Popular UAVs for Mapping UAV MODEL TYPE ENDURANCE MTOW RANGE COST WIND (KG) (KM2)A (US$) RESISTANCE (M/S)B DJI Mavic Pro Rotary 20 minutes 0.7 0.3 1,000 10 DJI Phantom Pro 3 Rotary 23 minutes 1.4 0.4 1,500 10 DJI S 900 Rotary 18 minutes 8.2 0.3 1,200 10 DJI Matrice 100 Rotary 23–40 3.6 0.6 3,300 10 minutesc SenseFly eBee Classic Fixed-wing 50 minutes 0.7 1.5 15,000 12 SenseFly eBee X Fixed-wing 90 minutesd 1.5 5 17,000 12.8 Delair DT18 Fixed-wing 120 minutes 2 20 30,000 13.8 WingtraOne Hybrid 55 minutes 4.5 10 20,000 12 Alti UAS Transition Hybrid 12 hours 16 45 105,000e 13 UAV Factory Penguin B Fixed-wing 20 hours 21.5 36 100,000 Unknown V-TOL Goshawk III Fixed-wing 150 minutes 5 20 38,000 12 Surveyor Wingcopter 178 Hybrid 90 minutes 15 5 80,000 15 Note: MTOW = maximum takeoff weight. Where specs were unavailable, this number was calculated from the combination of maximum speed and endurance values. a. These are claimed levels, but actual levels are likely lower for quality image acquisition. b. Time depends on battery configuration. c. Time indicates the endurance when the endurance extension is used. d. Value includes Real-Time Kinematic/ Post-Processed Kinematic positioning capability. e. Source: Information on UAVs compiled form manufacturers’ materials Technical Guidelines for Small Island Mapping with UAVs 27 Other issues to consider when deciding on a UAV for In the context of small islands, or to cover areas of 10 mapping include the following: km2 and more, a high-endurance multirotor, a fixed- wing or a hybrid platform with these characteristics is • What is the area of the average site to be recommended: surveyed? If the area is larger than 1 km2 (~0.4 square mile), a fixed-wing UAV is preferred. Rotary • Proven reliability,11 for instance as measured in multicopter UAVs could be used for mapping flight hours. When in doubt, inquiries regarding larger sites, but they require several flights and reliability in the setting of its intended use should good ground access in order to take off and land be made towards the manufacturer before within each flight distance range. purchasing the platform. • What is the typical terrain type around the • Foam construction or easily replaceable parts for survey site? Rugged terrain and obstacles on the ease of repair can be helpful in maintaining the ground may not allow the use of fixed-wing UAVs, craft over the long term. Alternatively, craft that which normally require open spaces (e.g., soccer can be serviced close to the location of their use fields) for takeoff and landing maneuvers. A should be preferred. fixed-wing hybrid with vertical takeoff and landing • Total weight of more than 1.5 kg with a profile that capabilities would be preferred in this case. will handle wind and weather conditions on small • What is the budget available? Typically, small islands will minimize necessary downtime due to rotary quadcopter UAVs are cheaper than fixed- inclement weather. Yet keeping in mind that few to wing or hybrid models. They also may require less no battery-powered UAVs on the market in early advanced piloting skills. 2019 are rain-proof, the technology is evolving quickly.12 Wind resistance should be 10m/s or • What type of data needs to be collected? higher when possible. The most basic onboard imaging sensors are small optical digital cameras. These allow • In addition, a platform with a flight time longer capturing images in the visible spectrum range than 45 minutes is preferable in order to minimize for producing typical red-green-blue (RGB) landings, the part of the flight when damage to the orthomosaics. Other sensors such as LiDAR platform is most likely to happen. and multispectral cameras can range greatly in • Reliable and trialed UAV systems that have size and may be used on board larger UAVs for streamlined workflows are best suited to high- collecting other types of data. stress situations such as following disasters. Figure 4 shows the various UAV platforms available on the market, plotted according to the range, price, and ease of use. If budget is not an issue, use of several platforms could be considered according to the objectives and nature of the survey, where the user would select the optimal platform for the different objectives. Small multi-rotor UAVs will be useful when quick deployments for aerial images over small areas are needed. Large fixed-wing UAVs lend themselves to large- area mapping. 11. Choosing a system that has been in use for many years and has a good track record—rather than a prototype or new system—is recommended. 12. Heavy platforms are not recommended for new or inexperienced pilots due to the higher risk these platforms present. 28 FIGURE 4. Price, range, and ease-of-use comparison of various UAV platforms available on the market. RANGE UAV Factory Penguin B Hawkeye System GS1000 ALTi UAS Transition Hawkeye System GS500 Wingcopter Delair DT-18 NOVA F7200 EASE OF USE Skywalker Wingtra 2000 Sensefly eBee+ Vapor Helicopter Smartplanes UX series 250,000 US$ DJI S900 DJI Phantom DJI M100 DJI Mavic 1,500 US$ VTOL Hybrid Fixed-Wing Multirotor Source: Based on experience from the UAV4Resilience Challenge, aerial surveys after cyclone Gita in Tonga, and informational interviews with professionals and manufacturers. Note: VTOL = Vertical Takeoff and Landing. Technical Guidelines for Small Island Mapping with UAVs 29 6. Choosing the sensor Many survey mapping UAV platforms come with • Uncompressed image formats. Many cameras built-in or recommended cameras, which are already record only compressed formats (JPG), which configured for optimal mapping outcomes. However, reduces the radiometric quality of the imagery. built-in cameras may also limit the user’s ability to Sensors that are capable of recording in RAW, customize optimal sensor configurations. TIFF, or other uncompressed formats produce higher-quality images. • Trigger control. Ideally, the camera shutter is PREFERRED CAMERA triggered by an external system that records the SPECIFICATIONS location and time of each shot. This is typically the flight control system (autopilot) or a separate The UAV’s flight altitude and sensor determine the GPS sensor. Many professional-grade cameras spatial resolution of the imagery. To put this another have software that enables remote control, but way, the focal length, sensor size (size of the chip that consumer cameras often lack remote trigger detects light), and density of detectors on that chip ability unless custom modifications are made. For all influence the altitude required to collect data of these consumer cameras, the shutter must be a given spatial resolution (i.e., minimum resolvable triggered continuously by time interval (though not object size). Most survey mapping UAVs use digital all cameras have even this capability). If triggering RGB cameras that capture imagery in the visible by time interval is the only option, the sensor range of the light spectrum. When the camera can be must have a buffer sufficient to enable continuous selected separately, the following characteristics may acquisition at the rate required to collect the be desirable: forward overlap selected. • Large detector size. A larger detector size means • High-quality prime lenses. When possible, high- that the imagery collected will be of higher quality quality prime lenses (i.e., lenses of fixed focal or fidelity. This is in part because larger detectors length) should be used. Many smaller sensors are typically found in larger, higher-quality do not allow for the use of prime lenses. Where cameras, but also because larger detectors they are absent, the camera should be operated can collect more energy from a given pixel and only at the extremes (minimum or maximum of therefore typically have a much higher signal-to- the focal range) for digitally controlled “zoom”; noise ratio. or in the case of variable focal length lenses controlled manually, tape can be physically applied to the focus wheel to ensure that the focal length remains consistent throughout and between flights, which is critical for processing. Wide-angle lenses (~ < 24 mm in 35 mm equivalent) increase view angle distortions and can severely affect the geometric and radiometric quality of the resulting orthomosaic. Technical Guidelines for Small Island Mapping with UAVs 31 • Calibrated camera model. Images collected with Increasingly, mapping-UAV manufacturers provide digital cameras are subject to a certain degree of cameras that are already optimized for mapping, geometric distortion. A fundamental prerequisite minimizing the effort that end users must make to of automatic triangulation (AT) and mosaicking ensure appropriate cameras and calibrations. is the accurate calibration of the sensor to Ultimately, the payload capacity of available platforms undistort captured images correctly. The set of will determine which cameras can be carried. Some parameters describing sensor and lens geometric systems have embedded cameras that cannot be properties is called a camera model. For optimal swapped, while others allow for different sensor photogrammetric results, each sensor should be configurations. Digital single-lens reflex (DSLR) calibrated by a specialized laboratory for each cameras with full-frame sensors can be carried only by lens combination. Several UAV photogrammetric larger multirotor UAVs. Most portable small UAVs have software programs automatically calculate camera either embedded sensors or compact cameras that models as part of the AT processing routine. typically weigh less than 500 g. Table 4 compares three • Global shutter. Most consumer cameras employ common RGB cameras used on mapping UAVs. a “rolling shutter,” which can greatly affect image quality when placed on fast-moving UAVs. In recent years, manufacturers have developed “global shutter” sensors that greatly reduce distortion and image geometry issues. TABLE 4. Comparison of Popular UAVs for Mapping CAMERA SENSOR SIZE RESOLUTION WEIGHT (G) PRICE (US$) (MP) Mapir Survey3W 6.17 x 4.55 mm 12 76 400 (1/2.3 in.) Sony A6000 25.1×16.7 mm 24 468 800 (APS-C) Canon EOS 5D Mark IV 36 x 24 mm 30 890 3,000 (full frame) Source: Information on cameras compiled from manufacturers’ materials 32 TAGGING IMAGES USING GPS OTHER IMPORTANT CONSIDERATIONS GPS tags matter! For post-processing, each frame There are other important considerations for choosing taken by the camera needs to be tagged with its a UAV sensor in addition to the camera specifications GPS location. This is an important process to ensure and ability to tag images: positional accuracy of the output. If the camera is • Obstacle avoidance. Even UAVs that are small linked to an onboard GPS, the tagging can be achieved and relatively inexpensive (less than US$1,000) automatically by triggering the sensor through the now are equipped with basic “sense and avoid” computer managing the flight (e.g., as waypoints or systems that allow the aircraft to detect any by time) on the ground. This approach insures that obstacle in its path and automatically change the trigger is included as part of the flight record course to avoid it. along with the location and orientation information, permitting extraction of location information for each • RTK/PPK GPS correction technology. For frame from the flight log. accurate data capture, an alternative to using ground control points (described in chapter 7) is When operating a sensor system that is not linked to to use UAVs equipped with high-precision GPS, a GPS by design, the only option is post-flight syncing which is able to record the position of the aircraft of GPS data with the camera time exchangeable image much more precisely than standard GPS receivers. file format (EXIF) tags. Even when done correctly, Real-Time Kinematic (RTK) allows high-precision this introduces significant error into the location measurements of locations by using a base station estimates for each frame and is not recommended. with known coordinates that are accurately Most cameras record frames at only 1 hz (once per measured and a radio link or other means to second) frequency, which means that syncing to a 1–5 send and receive the correction data from the hz GPS record can introduce up to two seconds of base station, thus performing “live” triangulation error in the syncing process. When compounded with corrections while the UAV is flying. Post-Processed GPS error, this can result in location errors of tens of Kinematic (PPK) is similar to RTK, except that the meters, depending on the flight speed. To achieve the corrections to the GPS positions are calculated best mapping results, the sensors and systems used not during but after the flight. Both PPK and RTK should be linked to a GPS and able to tag image systems can be purchased as additional features frames automatically. on many professional mapping UAVs. Technical Guidelines for Small Island Mapping with UAVs 33 7. Preparing for the UAV Mission When on site, the following three tasks will be most TIME MANAGEMENT time-consuming. The actual flying of UAVs is the least time-consuming 1. Finding an area of operation for takeoff and part of the process. Figure 5 shows the relative landing. The operator should identify potential proportion of time needed for each step of the entire sites remotely (e.g., through existing satellite UAV mapping process, starting from preparation imagery) and dedicate a day or more to visually and including flying and post-processing. If the UAV assessing whether the operation sites are mission will fly under circumstances that require acceptable (Figure 6). special approvals, it can take anywhere from a few weeks to a few months to obtain these approvals; in 2. Reorganizing priorities due to weather or other addition, the approvals may be expensive, especially unforeseen issues. Although weather-related in cases where approval to fly higher than 400 ft is changes may be difficult to foresee, budgeting being sought. The decision tree shown inFigure 1— extra days for them can be helpful. taken from the Airshare (n.d.) website—identifies the 3. Processing and delivering the data. This task types of approvals required under different scenarios. is particularly important and should be given ample consideration. Based on experience during The preparation of flights and post-processing of the Tropical Cyclone Gita response in Tonga, data consume far more time than the actual flying the recommended approach is to complete the and data collection. Before any disaster occurs, it is delivery of data on site with at least one person important to have a streamlined approval process and working solely on post-processing. This approach work flow in place for emergency mapping. Having a is critical under tight deadlines, such as in post- good relationship with the local CAA is also important disaster situations. to enable smooth operations and communications to avoid risks. Technical Guidelines for Small Island Mapping with UAVs 35 FIGURE 5. Time needed for the steps of a UAV deployment as a share of the whole process. As indicated by the red arrow, the actual flying of the platform occupies only a small proportion of time. FLIGHT FLIGHT MISSION APPROVAL PROCESSING ANALYSIS PLANNING TIME FIGURE 6. A large, open field selected to launch and land the fixed-wing UAV in Tongatapu, Tonga. Credit: UAV4Resilience 2017 36 TAKEOFF AND LANDING SITES DEFINING THE AREA OF INTEREST AND COORDINATION The selection of takeoff and landing sites is particularly important if a fixed-wing UAV is being operated. The The area of interest (AOI) where images will be type and size of platform will determine the area/ collected is determined by the mapping needs and space needed for landing procedures. An open space environmental conditions. It is useful to plan the with an even surface (such as a sports field) is often mission using preexisting imagery in a geographic ideal. Site selection should begin with a desktop information system (GIS) as a backdrop; this step assessment using maps and images, and then include allows checking for obstacles on the ground that could an in-person visit to ensure that the location is suitable impede takeoff or landing and also offers a view of the for takeoff and landing. topography, land cover, and road accessibility of the operational base (takeoff and landing site). Airspace designation charts published by CAAs show the classification of airspace and should always be consulted when determining if a UAV mission in the AOI is feasible. Once these factors are reviewed and AOI defined, the AOI polygon along with basic metadata on desired image specifications (e.g., target ground sample distance) can be exported from a GIS and used as input in the flight-planning and coordination software. FIGURE 7. Mission planning for the LiDAR survey in Fiji during the October 2017 field testing. Credit: UAV4Resilience 2017 Technical Guidelines for Small Island Mapping with UAVs 37 With increasingly widespread use of UAVs, the sky When creating a flight plan, it is important to include can quickly become crowded with flying imaging extra flight lines and frames outside of the AOI to platforms, both manned and unmanned. This scenario cover all perimeter zones with enough frames. As a is particularly likely when multiple UAV teams come rule of thumb, two extra frames at the end of each together after a natural disaster event to map the flight line and one extra flight line on each side of the impacted areas for rapid damage assessment and AOI are normally enough to ensure proper coverage. support to response operations. Coordination with the Most professional flight-planning software will already local CAA is paramount to ensure safety. account for the need of additional overlap. To optimize mapping activities, online tools are being Flight altitude should be set at a fixed value above developed to allow easy sharing and coordination mean sea level for areas with homogeneous ground of “who’s flying where.” The Imagery Coordination elevation and should be adjusted above ground Service,13 for example, offers an open source platform level when elevation changes significantly (e.g., for requesting imagery collections and facilitates mountainous areas). This ensures a consistent overlap coordination between UAV pilots and other traditional ratio between frames even when the distance between imagery providers (e.g., satellite-imaging companies). platform and target ground changes. Several UAV vendors provide software to design and create flight plans based on input AOI files, and UAV FLIGHT PLANNING and camera specifications. These programs can be When flying, UAV sensors collect images by frames. installed on laptop computers, tablet devices, and After the flight, the frames need to be stitched smartphones. These applications provide interactive together to create a mosaic image that shows the methods for selecting the desired ground sample entire area seamlessly. The image collection for distance and overlap, as well as for optimizing UAVs assumes a pattern of parallel flight lines and a flight patterns based on the maximum altitude fairly stable aircraft. Aircraft that are less stable — a (ceiling) allowed by law, aircraft speed, and camera category that includes many UAVs — necessitate some specifications (sensor size and focal length). changes in the collection strategy. For example, AT processing requires significantly greater forward and side overlaps to achieve accurate results. Further, images should be collected at a vertical (nadir) or near-vertical angle. Slightly oblique images may still be processed, but results will vary greatly because features on the ground will appear from different perspectives and make it more challenging for the mosaicking software to find matching patterns. Forward and side overlap — the amount of overlap between frames in the forward and lateral direction from the perspective of the platform’s direction of movement — must be properly handled to create seamless mosaics that represent the location of the features in the image. The highest overlap possible should be collected, with a minimum forward overlap of 60 percent and minimum side overlap of 30 percent to create a mosaic with good positional accuracy (Figure 8). During operations, there is a tradeoff between the time available for the survey and the overlap. The more overlap, the more time is required to complete the flying. To produce accurate terrain models, a minimum forward overlap of 80 percent and a minimum side overlap of 75 percent are recommended to maximize the number of 13. Imagery Coordination Service, observations of landscape features. https://coordination.openaerialmap.org. 38 FIGURE 8. Recommended forward and side overlaps to create a mosaic with good positional accuracy. Source: PacDID. 2017 FIGURE 9. Working with the Ground Control Station for mission planning and live monitoring of the flights Credit: UAV4Resilience 2017 Technical Guidelines for Small Island Mapping with UAVs 39 For countries that are serious about establishing UAVs TARGET LANDSCAPE LIMITATIONS as an accessible platform, investment in robust and AT and mosaicking rely on automatic extraction durable control points will be worthwhile. If known of point features from input images. In the case of control points are available, creating a denser network imagery collected over visually homogenous pattern of benchmarks could be recommended. The resolution areas such as water, bare desert, or snow and ice, it of UAV imagery is such that the positioning accuracy is almost impossible for AT software to discriminate of GCPs is often less than a few centimeters. To match unique points and match frames correctly. Accurate existing data sets, to compare pre- and post-disaster IMU (inertial measurement unit) information may damages with some degrees of automation, or to sometimes compensate for the lack of feature conduct engineering surveys, a local geodetic system points in these areas and provide enough positional is required. information for correct orthomosaicking. When the timeliness of the data is important, setting up ground control points to achieve high absolute accuracy may not be cost- or time-effective. In these GROUND CONTROL POINT TARGET situations, a UAV equipped with an RTK or PPK PLACEMENT system should be used instead. Ground control points (GCPs) are used to ensure high positional accuracy of the final UAV image in case PPK or RTK tools are not available. GCPs are a set of identifiable features in the collected images with known spatial coordinate information. GCPs are normally collected with survey-grade GPS devices that provide centimeter-level precision. These features can be either existing physical objects (e.g., corner of a road intersection) or custom targets manually positioned in advance of a UAV survey across the target AOI (Figure 10). The GCPs should be well distributed across the AOI; otherwise, they could end up skewing the positional accuracy of the final images. The manual positioning of custom targets can be very time-consuming, as it requires identifying access routes to the areas where targets should be placed, then recording each location with an accurate GPS device. Targets should be large enough to be seen in multiple (at least three) overlapping aerial shots and should be anchored to the ground so they are not accidentally moved by people or wildlife. Depending on the environment, it is good practice to include a note next to each target that explain its purpose; this should minimize interference from residents. 40 FIGURE 10. Placing of a ground control point. The accurate coordinates of this GCP are surveyed on the spot using high-accuracy GPS. Credit: UAV4Resilience Challenge 2017 Technical Guidelines for Small Island Mapping with UAVs 41 8. Managing work flow in the field UAVs are often small enough to be carried in A generic pre-flight safety checklist can include the backpacks, but usually a motor vehicle such as a following steps: small pickup truck is used to carry the associated • Verify proper distance from residential areas, equipment, including the survey equipment (base people, overhead power lines, trees, etc. station) if GCPs are being used. A vehicle also provides shelter in case of inclement weather, and the • Inspect takeoff site; identify alternative options for vehicle battery or cigarette lighter plug can be used to landing recharge UAV batteries and laptops in the field during • Communicate with air traffic control and provide long survey missions. updated flight plans Teams in the field should include at least two people: • Check weather conditions, particularly wind speed a pilot who monitors the UAV’s flight on the computer and direction screen, and a pilot who maintains sight of the UAV • Inspect UAV, batteries, and radios for any anomaly and surrounding airspace at all times while flying within line of sight. • Ensure that batteries are fully charged At the takeoff location, the team’s designated • Test telemetry link and (if available) failsafe safety officer ensures that all precautions are taken sensors/mechanisms before, during, and after flights.14 At this point, it • Acquire GPS lock and verify home location is assumed that all legal prerequisites to fly a UAV recording have been obtained, including special permissions • Ensure safe distance between aircraft and any and communications instructions, especially those bystander applicable during an active disaster response scenario. To identify cases where New Zealand–based • Verify flight plan perimeters. UAV regulations require special permissions, please refer to the Airshare website (Airshare n.d.). 14. The safety officer may be a third person, or one of the two pilots who takes on this additional role. The safety officer is responsible for ensuring all pre-flight checks are followed and that conditions to fly are appropriate (i.e., wind and cloud ceiling at minimum thresholds, clear skies). Depending on the procedures that have been agreed, the safety officer may also be responsible for communications with ATC for airspace deconfliction. Technical Guidelines for Small Island Mapping with UAVs 43 Most commercial UAV platforms come with a safety It is always important to modify settings to suit local pre-flight and post-flight checklist that should be lighting conditions prior to each flight. Preferred followed. At every step from launch to landing, the sensor settings are listed below (in order of preference designated safety officer should coordinate actions where more than one setting appears). and communications (with loud voice signals). While in • Focal length: fixed flight, the aircraft should be kept within visible range unless special permission has been obtained and • Operating mode: manual, shutter priority (shutter tracking equipment is available for flying beyond the 1/1000–1/1500) visual line of sight. After each flight the aircraft should • Format: RAW, TIFF, low-compression JPEG be inspected for any anomaly. • Focus: infinite Light and weather conditions are important. To • Flash: off minimize shadows in the image, it is best to fly during the day when the sun angle is high. The color depth • Auto-rotate: off and tone of imagery is also best when collected in full • Optical/digital stabilization: off sun. Clouds, even light haze, obscure the signal in the • Metering mode: full-frame metering rather than scene and make use of the imagery challenging, if not “spot” metering (to reduce exposure issues) impossible. High or even moderate winds may likewise pose difficulties, as many light UAVs do not operate The most important step before starting a UAV well in such conditions. The day and time chosen to mission is to upload the flight plan to the aircraft collect the imagery is therefore critical. Data collection onboard computer. This can be done in different should ideally be done between 10 a.m. and 2 p.m. to ways, but usually from a laptop or mobile device that minimize shadows and on a clear day to minimize haze provides a screen visualization of flight patterns over and cloud effects; it should also take into account the preexisting imagery or maps. This visual check allows limitations of the platform with respect to wind and for verification of the home (takeoff) setting and its operating range (Figure 11). relative position to the flight extent. FIGURE 11. Checking conditions and following safety measures in the field. Source: PacDID. 2017 44 FIGURE 12. Launch of a multirotor UAV, which requires less open space than for a fixed wing. Credit: UAV4Resilience 2017 Once the flight plan has been uploaded into the aircraft Once the UAV has landed, images and GPS logs are computer and final checks have been performed transferred to a local device either through the radio (and all necessary approvals have been obtained and link or by physically removing the storage medium communicated), the UAV is cleared for takeoff. (typically an SD card) from the aircraft for copying files to a laptop. When multiple flights are planned, data It is important to note that all mapping UAVs must are transferred at the end of the mission and only the be flown in autonomous mode following a predefined battery is swapped between flights. It is still important flight plan in order to obtain images suitable for to sample image quality by transferring a few full creating a mosaic. In general, manual flights where resolution files from the camera to a device used for pilots maneuver the aircraft remotely cannot ensure inspection. In case of any issue, the camera settings that the correct pattern is being followed and enough should be checked and adjusted. overlap is maintained between frames and flight lines. After takeoff, the aircraft should be checked for any anomaly (through visual and telemetry feedback) before starting the autonomous mission. The pilot should always be ready to take over in manual control mode if needed. Technical Guidelines for Small Island Mapping with UAVs 45 9. After the flight: processing the data Flying and collecting imagery is often the easiest In environments with limited Internet connectivity, this step in UAV mapping, while preparing and processing option is often not suitable for UAV mapping projects, imagery often requires significant computing and which require uploading massive amounts of data human efforts. Photogrammetric software that (tens to hundreds of gigabytes) to the cloud before employs processes such as Structure from Motion processing can start. (SfM) is actively and continuously improved to reduce As of 2018, the open source options for UAV imagery the level of effort required and to automate the processing software are quite limited. The most processing work flow. popular and mature programs are OpenDroneMap and its sister project WebODM. Table 5 lists some of the most popular UAV processing software available and PROCESSING SOFTWARE AND its pricing. HARDWARE Most software used to process UAV data is developed by commercial firms and offered through pricey subscription options or permanent desktop licenses. Some of the same companies also offer cloud-based processing services that allow for uploading imagery to scalable cloud computing infrastructure. Technical Guidelines for Small Island Mapping with UAVs 47 TABLE 5. Comparison of Popular UAV Image Processing Software Programs SOFTWARE OPTIONS LICENSE MONTHLY / PERMANENT PRICE (US$) DroneDeploy Cloud Commercial 399 (monthly only) Pix4Dmapper Desktop + cloud Commercial 350 / 8,700 Photoscan Desktop Commercial 3,499 (permanent only) Correlator3D Desktop Commercial 295 / 5,900 OpenDroneMap Desktop Open Source Free Source: Information on software compiled form manufacturers’ materials FIGURE 13. User interface of OpenAerialMap.org showing UAV images that are available for Tonga. Source: Open Imagery Network contributors. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/legalcode. 48 Several UAV manufacturers offer processing 8. Import GCPs and carry out manual matching to software along with their UAVs, which can result in a corresponding image features. significantly lower final cost than if UAV hardware and 9. Conduct color balancing. software are bought separately. It is worth inquiring before purchasing a mapping platform what software 10. Generate dense point cloud and digital surface is included in the purchasing price. Using software that model (DSM). is provided with the mapping platform is often the best 11. Extract digital terrain model (DTM) from DSM (with option. optional manual editing). The average amount of data processed in UAV 12. Carry out seamline generation and make manual mapping is large, often larger than in traditional GIS adjustments. projects. Like many remote sensing projects, UAV 13. Carry out orthomosaic generation (with optional mapping requires intensive computations and large overviews generation). storage and memory capacity for data processing. This type of processing is normally done on large 14. Issue output of final products in GIS-compatible workstations set up with multicore CPUs, at least formats (e.g., GeoTIFF). 16GB of RAM, and fast solid-state drive (SSD) disks. Processing 200 image frames locally on a gaming Some key processing steps employ computer vision laptop to produce an orthomosaic typically requires (CV) algorithms, which require advanced graphics one to two hours to complete. Some of the manual processing units (GPUs) that are typically found only in steps include importing input files (images, GPS log, expensive scientific and gaming video cards. etc.) and matching GCPs. Everything else can follow Desktop workstations for UAV image processing predefined processing parameters and eventually be can cost anywhere from US$2,000 (low end, for scripted to run fully autonomously. projects up to ~1,000 image frames) to US$10,000. A processing report is usually generated by the As processing often needs to be done in the field or software, including quality control and assurance while deployed in a disconnected environment, a bulky indicators about the processing work flow and the workstation is not the ideal solution. A valid alternative final data products. For example, it’s possible to use a for small to medium-size projects (100–5,000 frames) number of GCPs as “checkpoints” to verify the internal is offered by gaming laptops, which are often already and absolute positional accuracy of the output model equipped with configurations similar to those needed compared to known reference locations. Root-mean- for UAV image processing (advanced CPU, SSDs and square error tables and charts are often available large storage, sizable RAM, a dedicated video card). In in the report to assess each point’s horizontal and addition, if processing is to be conducted in the field, it vertical accuracy against predicted values. is advisable to use ruggedized equipment and to have several charged battery packs as backup options. DATA PROCESSING FOLLOWING 2018 TC GITA RESPONSE SURVEY POST-PROCESSING WORK FLOW TO CREATE ORTHOMOSAICS During the field surveys conducted in February 2018 (see chapter 11 for more details), the imagery data Whether on a desktop, on laptop, or in a cloud were processed by individual flight. As a result, the computing environment, the processing work flow island of Tongatapu was divided into strips of smaller to obtain orthomosaics from UAV-collected imagery orthomosaics that each corresponded to a UAV is very similar. The main steps involved, mostly flight. To improve the accuracy of the results and performed automatically by the software, are the radiometric quality across the whole island, all the following: imagery was subsequently processed into one large 1. Import image files (and optionally create block, generating a seamless mosaic that covered the overviews). entire island area (more than 260 km2). To make this large orthomosaic easy to share with end users, the 2. Import GPS log and match with images (not large GeoTIFF file was then divided into 1 km2 tiles. necessary if GPS information is already available Figure 13 shows some of the strips from the original through image EXIF metadata). UAV flights (on the left of the image), while Figure 14 3. Select processing parameters and algorithm types shows the boundaries of the reprocessed 1 km2 tiles. (if choices are available). The orthomosaic of the whole island was uploaded to 4. Extract features/points from each frame. OpenAerialMap.org. 5. Create camera model. 6. Carry out feature matching and automatic triangulation. 7. Carry out bundle block adjustment and AT model refinement. Technical Guidelines for Small Island Mapping with UAVs 49 FIGURE 14. Image showing location of the structures assessed as damaged (in orange) and destroyed (in red). Source: Open Imagery Network contributors. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/legalcode. SHARING DATA AND IMAGERY Other options for portability and sharing include lossless (e.g., LZW) and lossy (e.g., JPEG 2000) All geospatial data produced by UAV data processing compression algorithms that can drastically software should be outputed into formats that can be reduce the final file size. readily consumed in common GIS programs such as • Privacy issues. Compared to typical satellite ESRI ArcGIS and Quantum GIS. Data created using images, UAV imagery offers an incredible level of UAVs can be shared in various ways. For instance, detail. This feature allows for novel applications, DTMs, DSMs, and point clouds can be shared on such as identification of damaged structures opentopography.org as open data. Orthomosaics after a hurricane or earthquake. But it also can be shared by uploading onto the OpenAerialMap raises privacy issues, especially when such high- platform (Figure 13). These orthomosaic datasets resolution imagery is shared with others outside are foundational layers of information used in many of the project for which UAV flying was authorized. humanitarian and surveying project applications, The UAViators’ code of conduct (UAViators, n.d.) from baseline mapping (tracing) to remote damage and summary of best practices (UAViators 2015) assessment. can help guide decisions on how to handle and Sharing of UAV data and imagery must take into share collected imagery. If necessary, measures account both product formats and privacy issues: such as downsampling or blurring can be taken to protect people’s privacy or respect the sensitivity • Product formats. The generated output imagery of specific areas. mosaic is often a very large raster file (from tens to hundreds of gigabytes), with three bands of RGB value combinations representing colors in the visible range of the electromagnetic spectrum. The most common format for output is GeoTIFF. In order to facilitate handling and opening of such large files, the full mosaic may be split into same- size tiles, labeled in a grid sequence, and indexed by a shapefile or other schema file (e.g., VRT). 50 10. Handling project management Managing a UAV project requires attention to a • Food, water, shade. UAV missions should plan for number of concerns: food, water, and a source of shade in the field. In some locations, and particularly after a disaster • Lithium batteries. Lithium polymer (LiPo) has occurred, it may be difficult to purchase food batteries used in UAVs are highly flammable and and water on site. Because UAV operations require must be transported in LiPo safety bags when outdoor work during the hotter hours of the day, traveling. On commercial flights, passengers are it is also important to have an umbrella or tent to allowed a maximum of two 100–160 Wh batteries, protect people and equipment from overheating. which must be kept in carry-on luggage.15 Quantities of smaller batteries are not usually • Daily planning. The hours available for UAV limited, but some airlines or airport security staff flights can be limited by weather and sunlight. It may take away what they perceive as excessive is advisable to plan all aspects of the day’s work, quantities of batteries. Before traveling with including worst-case scenarios, on the preceding batteries related to a UAV mission, it is advisable evening. This approach ensures that valuable to obtain the airline’s transport policy in writing so daylight hours are used for flying rather than there will be no question of permissibility. planning and coordination. • Power supply. When working in the field, power • Mission reports. During the mission, the operator can be limited. A good power supply should be should keep a daily log that records flights, any ensured both on site and at the facility where issues encountered, and recommendations for processing of the data will take place (e.g., future missions. This gathering of small but key solar charger or car inverter in the field; reliable details of the day-to-day operations—such as electricity or on-site generator at the hotel or turnaround time, number of failures, potential office). for schedule delays—can be helpful in estimating potential extra costs and improving budgeting, planning, and execution of future missions. 15. For more information, see IATA (International Air Transport Association) guidance at http://www.iata.org/whatwedo/ cargo/dgr/Pages/lithium-batteries.aspx. Technical Guidelines for Small Island Mapping with UAVs 53 • Collaborators and service providers. In many • Maintaining skills. Flying UAVs during and after situations, hiring local professional service emergencies requires a high level of skill. To providers can be an efficient way to implement a ensure readiness in these high-stress situations, survey. Local professional firms are likely to have training should be ongoing. More specifically, already met relevant regulatory requirements operators should regularly practice and train in the (i.e., are likely certified or registered), so this piloting skills that are needed during emergencies. approach can save time. It can also simplify All flights during emergency operations must logistics because local service providers know be premeditated and based on training with the the operational context, the terrain, and the responsible emergency personnel and authorities; population. Finally, it can save money, since it otherwise well-meaning operators risk seriously obviates shipping of equipment and hiring of interfering with emergency response activities. international experts. Of course, any local service   providers hired must have the adequate expertise and experience for the project. Table 6 offers some guidance on skills profiles and equipment that may be needed for various types of mapping. TABLE 6. Sample Sensor Types and Experts Needed for Various UAV Data Applications. Table created by the UAV4Resilience team. PROJECT TYPE SENSOR TYPE RESOLUTION ACCURACY Experts Crop monitoring Red edge or As low as 30 cm/ 5m Remote sensing multispectral pixel specialist; agronomist Urban mapping Standard RGB 2–10 cm/pixel < 10 cm; requires Urban planner; camera use of GCP or machine learning RTK/PPK specialist Disaster mapping Standard RGB 10 cm/pixela 1m Disaster recovery camera management specialist; machine learning specialist for damage assessment (if crowdsourcing is not used) a. UNICEF Malawi uses 7 cm/pixel resolution for flood mapping. 54 11. Case Study: Post– Tropical Cyclone Gita Mission in Tonga On February 12, 2018, Tropical Cyclone Gita hit the COLLECTION OF POST-CYCLONE island nation of Tonga as a Category 4 hurricane, UAV DATA causing major damage across Tongatapu, the main island, as well as the neighboring island As is often the case with emergency response flights of Eua. The total economic loss was estimated when time is of the essence, a balance had to be at approximately 38 percent of gross domestic struck between high absolute positional accuracy and product, triggering international support for speed of flying and post-processing. The UAV4R team long-term recovery. At the request of Tonga’s compromised on positional accuracy in favor of faster National Emergency Management Office, the flying and data processing. Having flown UAVs in Tonga UAV4Resilience (UAV4R) team deployed to Tonga just four months prior to the cyclone, the team already to capture post-disaster UAV aerial images. had a valid Part 102 New Zealand certificate to fly and The main challenge was to fly as quickly as also knew all the takeoff and landing sites, which saved possible while still meeting all regulatory time. requirements. Because the plan was to fly above Moreover, the collection of baseline data only a few 400 ft, the team had to seek additional permission months before Tropical Cyclone Gita hit Tonga proved from the Civil Aviation Authority, just as for the to be invaluable. It not only provided usable data for October 2017 field-test missions. The approval post-disaster comparison, thus facilitating accurate process took three weeks to complete and damage assessments; it also acted as a trial run for modifying any flight plans for complex airspace the actual disaster, thus enabling the development operations took between one day and one week of procedures, location of suitable flight areas, and (more detail is in annex 1). The logistics involved rectification of problems. Figure 15 shows an example in reaching the island were also challenging. of the same structure before and after the Cyclone. Eventually, the Australian Department of Foreign Affairs and Trade (DFAT) emergency response team offered the UAV4R team transport from Brisbane to Tonga on a C17 military aircraft. Technical Guidelines for Small Island Mapping with UAVs 57 From the two data collection processes — one before 2. Additional technology may be necessary to and one after the disaster — two main lessons improve situational awareness in the airspace. emerged: When operating in post-disaster airspace, air traffic controllers will be hesitant to allow 1. A three-person crew works well in a post- atypical operations that they do not have clear disaster setting. In a post-disaster scenario, situational awareness of. They may find that use where rapid data are critical, UAV operations are of a dedicated real-time positioning system such typically carried out for as long as the weather and as UTM (Unmanned Traffic Management), or an light permit. The full day of flying means that hours existing technology such as ADS-B (Automatic are spent at night ensuring batteries are charged, Dependent Surveillance–Broadcast), provides data backed up, missions planned, and equipment a layer of comfort and routine to the nonroutine ready for the next day. Experience suggests that operations. Smoother ATC in turn can facilitate it is useful to operate with a three-person crew, extended UAV operations. with two in the field collecting data and operating the aircraft, and the third serving as ATC liaison The two missions combined represent a total and processing data from the day before. Rotating budget of about US$0.5 million, which includes the third person through the data collection and many experimental elements during the field test. flight crew to serve as ATC liaison provides crew The amount of information collected from these two members with a rest, while also ensuring progress missions represents millions of dollars in field work on data collection/processing. and data interpretation. Not only did the UAV collect images of urban damage, it also collected images of cropland and shoreline. In one two-hour flight, the UAV covered more than 12 km2, or 24,000 acres, acquiring more detailed imagery than could be produced by a satellite — and doing so more than 10 times faster than a ground crew conducting the same work. FIGURE 15. Images showing pre- and post-event status of the same structures. Credit: Open Imagery Network contributors, UAV4R Team (World Bank and VTOL Aerospace) 58 This classification scheme is rather different from APPLICATION OF POST-CYCLONE the three-class system (destroyed, damaged, UAV DATA not damaged) that engineers on the ground The UAV data collected in Tonga following Tropical used, so it is difficult to compare the two sets Cyclone Gita was used for several purposes: of results. The ground survey by the National Emergency Management Office initially identified • Crowdsourced building damage assessment by approximately 3,000 houses as damaged, whereas the Humanitarian OpenStreetMap Team (HOT). HOT estimated the number at around 2,500. The pre- and post-event UAV images were used to Both sets of numbers must be taken with a pinch carry out a crowdsourced damage assessment. A of salt, as both have continued to fluctuate. small subset of volunteers from HOT were tasked However, in terms of order of magnitude, the two with doing the visual interpretation of the damage are aligned. To improve the accuracy of housing to housing. A simple classification scheme was damage assessments, and also to make the used: houses that appeared to be more than 50 two approaches comparable, a unified damage percent damaged were labeled “destroyed,” and classification scheme must be developed. those that appeared to be less than 50 percent damaged were labeled “damaged” (Figure 16 ). FIGURE 16. A house destroyed by Tropical Cyclone Gita in Longolongo village, Tongatapu, Tonga. Credit: World Bank Technical Guidelines for Small Island Mapping with UAVs 59 • Supporting the claims validation process. The • Quantifying the damage and recovery needs of housing sector bore the brunt of the damage school buildings. The pre- and post-event images caused by the cyclone. The rapid assessment were used in the detailed first-order estimates of conducted by the World Bank, Asian Development the reconstruction and repair needs of schools, Bank, and other partners indicated that which suffered significant damage in the cyclone. approximately 25 percent of the housing stock Some 75 percent of the schools on Tongatapu, for had been damaged or destroyed. In May 2018, example, reportedly were affected by the cyclone. the government decided to provide support to The UAV images were sufficiently detailed to households whose housing had been affected allow quantification of the number of classrooms, by the cyclone. The cash transfer beneficiary list staff quarters, toilets, etc. that were damaged was developed based on the damage assessment or destroyed; the images were also used to conducted by the government; and in keeping develop the repair and reconstruction plan for the with common practice globally, grievance redress education sector. systems were put in place to allow homeowners to appeal the government’s decision on the damage to their house. Problems arose when homeowners started repairing their houses after the event without waiting for the government assessment to take place. Thus by the time the government damage survey was carried out, some of the houses had already been repaired — meaning that the beneficiary list failed to include some homeowners whose house had indeed been damaged by Tropical Cyclone Gita. The UAV images, however, showed the condition of structures two to three weeks after the event, and thus were instrumental in supporting the claims validation effort. 60 12. Conclusions The agility and versatility of UAVs as data collection Last but not least: while the innovation around UAVs platforms stand to significantly improve the availability is surely exciting and experimentation is encouraged, of timely spatial data to small island countries. UAV the safety of airspace users and those on the ground technology should be harnessed to further improve is of utmost importance and should always take resilience, risk reduction, and disaster response work precedence over other objectives. Any serious UAV in small island contexts. To ensure that this work operator or user of UAV technology should seek can go forward, adequate capacity building, training, professional training in how to fly responsibly in the and preparation are needed. This Guidance Note has relevant airspace. Exemplary and careful use of the discussed the most important factors to take into technology will help build trust among the public, account in using UAVs and UAV technology effectively. governments, and private sector and help ensure that UAV technology and related innovations can be Given the myriad of options available for platforms, used for beneficial applications going forward. Every software, and sensors, there exists no one-size-fits- operator has a role to play in shaping a responsible all approach to UAVs. Instead, interested parties are community of UAV operators. encouraged to experiment with various systems and configurations while also building on best practices. The field is rapidly and constantly evolving; as new technology becomes available, it opens up new possibilities and optimizations in this work. Technical Guidelines for Small Island Mapping with UAVs 63 References and other resources Airshare. https://www.airshare.co.nz. Pacific Humanitarian Challenge. n.d. “Meet the Innovators.” ———. n.d. “I Want to Fly a Drone in NZ. http://pacifichumanitarianchallenge.org/ Where Do I Start.” winners/#PacDID. https://www.airshare.co.nz/where-do-i-start. World Bank. “UAV State of Play in Development.” FSD (Swiss Foundation for Mine Action). 2016. https://uav-development.github.io/. Drones in Humanitarian Action: A Guide to the Use of Airborne Systems in Humanitarian Crises. ———. 2016. UAV State of Play for Development: http://drones.fsd.ch/en/homepage/. Innovations in Program and Humanitarian Contexts. Washington, DC: World Bank Global Drone Regulations Database. https://www.scribd.com/document/320974919/ https://www.droneregulations.info. UAV-State-of-Play-for-Development. Accessible from World Bank intranet. HOT (Humanitarian OpenStreetMap Team). 2016. “Improving Resilience with Aerial Imagery.” ———. 2017a. Guidance Note: Managing the Risks of August 15. Unmanned Aircraft Operations in Development Projects. https://www.hotosm.org/updates/2016-08-15_ Washington, DC: World Bank. improving_resilience_with_aerial_imagery. http://documents.worldbank.org/curated/ en/895861507912703096/Guidance-note-managing- the-risks-of-unmanned-aircraft-operations-in- UAViators (Humanitarian UAV Network). n.d. development-projects. “Humanitarian Code of Conduct.” https://uavcode.org/. ———. 2017b. “Lessons from Mapping Geeks: How Aerial Technology is Helping Pacific Island Countries ———. 2015. “Humanitarian UAV/Drone Missions: Recover from Natural Disasters.” East Asia and Towards Best Practices.” Pacific on the Rise blog. November 20. http://uaviators.org/docs. http://blogs.worldbank.org/eastasiapacific/ lessons-mapping-geeks-how-aerial-technology- Imagery Coordination Service. helping-pacific-island-countries-recover-natural. https://coordination.openaerialmap.org. New America. 2015. Drones and Aerial Observation: New Technologies for Property Rights, Human Rights, and Global Development. A Primer. http://drones.newamerica.org/primer/. Technical Guidelines for Small Island Mapping with UAVs 65 Annex 1. Applying for Approvals in Fiji and Tonga for October 2017 UAV Field Testing After processing submitted materials, CAA Fiji granted FIJI the approval to operate up to 200 ft AGL. However, for CAA Fiji outlined the following path to obtaining permission to operate up to 400 ft AGL, Airports Fiji necessary approvals for UAV operation: Ltd. required submission of a NOTAM and provided the standard NOTAM request form to be filled out. 1. Submit completed Form OP138 to CAA, accompanied by the following documents: Table 7 presents the time needed for obtaining approvals in Fiji. The durations are based on the • Proposed pilots’ remote pilot licenses experience of an Australia-based firm with experience • Letter of intent from the sponsoring working and training with CAA Fiji and should thus be organization, such as the World Bank considered best-case scenarios. • UAV operations manual • Third-party liability insurance cover letter 2. Obtain permission from the operator of the airspace—i.e., Nadi Airport Tower (the air services provider at Nadi Airport for Airports Fiji Ltd.)—to operate up to 400 ft AGL via a NOTAM (Notice to Airmen) request. 3. File a TAF2000 form to obtain an import permit from the Telecommunications Authority of Fiji for all transmitting devices. Technical Guidelines for Small Island Mapping with UAVs 67 TONGA The following list provides a breakdown of the documents that are required to be completed and submitted to obtain the Part 102 Exposition under the New Zealand CAA regulations which is valid for Tonga; • RePL’s and Police checks of the proposed pilot’s • Letter of intent from the World Bank • Insurance cover letter • Part 102 Exposition* • Part 101 Compliance Matrix* • Part 102 Compliance Matrix outlining the element by element compliance with the Part 102 Rules* • UAV Operators Safety Management System • UAV Operators Maintenance System Manual • Aircraft/ Platform Flight Manuals • Aircraft/ Platform Maintenance Manuals • Fit and Proper Persons tests for key personnel. TABLE 7. Time Needed for Obtaining Approvals in Fiji APPROVAL DOCUMENT TYPE DURATION (FROM REQUEST TO APPROVAL) OP 138 5 Weeks NOTAM 4 Days TAF 2000 2 Days TOTAL DURATION 6 WEEKS 68