OceanKnowledge for everyone!

Take off on KiritimatiAmong the many things which were keeping me enthralled on Kiritimati Island are all the amazing nerds coming through here – mind you, “nerd” in my books is one of the biggest compliments! I am thinking about Leslie, Paul and Alex from the NOAA El Nino rapid response program launching weather balloons by day and night, Todd and Todd who observe bubbles in the plasma of the ionosphere at the magnetic equator, Julia and Danielle who know every coral head in the atoll personally, or Kim and Alyssa hunting down El Nino traces in fossil corals up to 7.000 years of age. A lot of stories to be told. But today’s story is about Zulfikar and Hervé, the best bodyguards I could encounter on this island. But that is yet another story. Today’s story is about their work for the Geophysics division of the Secretariat of the Pacific Communities, the SPC, and their cool toy: a survey drone!


In the climate change debate sea level rise takes a front seat. We talk about current and future inundation of islands and continental coastal areas. But how can we find out which areas are threatened to which degree and how much the sea level actually rose? The answer are topographic maps and computer simulations of extreme wave events. To obtain the topographic maps, you can use tedious ground measurment of the GPS longitude, latitude and altitude coordinates for many, many points all over the survey area, so called Real Time Kinematic GPS (RTK). This method has the disadvantage that you can only measure along streets or other accessible areas with a rather low resolution – most streets are 10 m or more apart from each other. So the better (and definitely cooler!) method is to use aerial photographs taken by a drone. And that is what Z&H did, and I watched them for a few days and asked a trillion of questions.

Auatabu, Zulfikar Indiana and Hervé setting the GCPBefore we (ok, they) could fly the drone, we had to set up a reference grid for the drone flight. The drone has its own GPS, but it is not very exact and due to sudden movements by wind the coordinates given by the drone to each point on the pictures can be off by a few meters. Thus, not good enough for a topographic map, especially not when you use a technique giving a 2.5 cm resolution! To rectify this, five points of the survey area are measured with RTK and marked with a checkered marker visible on the aerial photographs. In the picture processing, the exact coordinates of each of those ground control points (GCPs) is entered and all other coordinates are automatically corrected accordingly.

Z&H’s plan of the day was to make two overlapping flights, which they wanted to stitch together to one big picture. So we had to lay out eight ground control points. To measure the exact coordinates of each of those GCPs, we needed to link them to one special reference point, which in the geophysics lingo is called a benchmark point. This is “just” a point, marked by a metal marker, for which the exact longitudinal, latitudinal and altitudinal location, the x, y and z, were established through a series of very exact and long term GPS measurements. Unfortunately the altitude given by GPS is no good, as it refers to some point on the calculated surface of earth (the reference ellipsoid), but not to sea level or anything meaningful. Luckily, H&Z had previously established a number of benchmark points, which they already reference to a local tide gauge, in order to obtain a valid altitude above mean sea level. We placed a GPS sender exactly on one of those benchmark points, announcing the exact coordinates. This GPS signal was picked up by a GPS receiver, which we carried with us. Actually, most of the time it was Auatabu carrying it, while we went to the center and corner points of the planned survey area, in order to establish GCPs. Points, which are visible on the photographs and for which we know the exact location in relation to the benchmark point. To make them visible, we spread out yellow-black checkered markers and secured them with fossil corals, the only stones available.

Hervé and Zulfikar during the pre-flight checkAll of this took more than an hour before we came to the really cool part, the preparation of the drone. The flight course had to be programmed, as well as start and landing points. Drone and camera had to be checked, the launch ramp assembled. And finally: Off she went! (The drone is obviously female: kickass and seriously cool!). The little black Styrofoam triangle with a load of microelectronics cut through the sky, 75 m high and approximate 80 km/h fast. At the turning points, the furthest points of her flight, she was often out of our field of sight. But then she came back, sometimes to be mistaken for a frigat bird. And after short 23 minutes, the flight was completed and she returned to land. The landing is a belly landing on a more or less flat field next to Tabwakea council. The landing was not the smoothest, but good enough for such an old lady - this drone is way past the official 100 flights the manufacturer gives her. With duct tape and glue she underwent surgical procedures by her handlers again and again, because the SPC isn’t the richest of organizations and new drone bodies are expensive.

After the flight H&Z performed a post-flight check and downloaded the hundreds of photographs to the control unit. Immediately afterwards they do the pre-flight procedures for the second flight, although the wind gets a bit stronger. Just when we were about launch, dark clouds rolled in from the lagoon side. Start or delay? It didn’t look so bad, so we launched and followed the drone by car, to be able to have a visual on the flight. We observed her from the beach just off the jetty, at the ocean side. In the South, over London, thick dark clouds came in and it started to rain on the remote end of the island. Abort! We jump in the car and the car jumps over the speed bumps with 70 km/h – probably a new record on this road! As soon as we arrived at the landing field, Hervé commanded the drone to abort and land. By now the wind had developed into a strong tropical squall and the rain was gushing down. Not really the best conditions to land (or fly) a drone! She came down properly, but then she bounced, her right wing caught in the tall grass, she flipped over and got to a stop belly up.

This is how you do not land a drone (from right to left)After quickly carrying her into the council meeting house she got a proper examination. Result: a deep scratch and bend in the lid of the camera compartment and a broken closing mechanism. And the camera didn’t show us any pictures! Shock! Luckily that part was solved quickly. The After the last flight - RIPbattery had been unsettled during the wild landing and removing and reinserting it did the trick. At least the flight pictures can be used. Until we are done with the post-flight check, the sun is shining again and all is good, typical tropical situation. After lunch, H&Z, masters of the drone, investigated the damage and decided this had been the last flight for this drone body. It was time for the new one to start its adventures. But before the first flight it gets thoroughly reinforced with meters of duct tape.

After duct taping the whole belly of the new drone body and transferring the electronics, we were ready to try the second flight again. The pre-flight check was quick. But again the wind picked up. Launch or no launch? It actually didn’t look really bad, the sky was clear, so we launched, hopped in the car, drove towards the jetty, saw darker clouds over London, saw the rain coming, turned around, hurried back and aborted the flight. The drone landed perfectly smooth, elegant really. And it didn’t start to rain, the wind quieted down before the drone hit the ground – we were too paranoid. But now it was too late for a new launch. Tomorrow would be another day. We packed up the drone and retraced our trip from the morning, in order to collect the checkered markers of the GCPs. But, the first one was gone. So was the second one! Were they stolen shortly after we put them out? Where they already gone during the first flight? If so, it was all in vain, no visible GCPs, no topographic mapping! Evaluation of the photos on the computer showed that the markers were there. Lucky us! At least one good flight and the second one also captured a good part of the ocean coast to be surveyed. So those data were usable as well

The drone clinic: duct tape is the drone's best friendNow we had all those drone pictures, but what happens next to convert them into a topographic map? First step was to upload all pictures and the flight plan in a specific. This program stitched all photos, which have an overlap of 90%, together. Each GCP marker had to be manually identified on one picture. The program then found all other 50 or more photos containing that specific GCP. Now the keyboard monkey work started: H needed to click on the center of the marker on 30+ pictures. This way, the exact GPS coordinates (derived from the benchmark point and RTK measurements from the morning) of this point in established on each of those photographs. 210 clicks later, H was done with identifying the centers of all GPCs for both flights. Afterwards the computer’s immense processing power was required to recalculate the GPS coordinates of each point of the roughly 1.5 km2 survey area. The outcome is one immense picture, like one huge photograph, with the additional layer of x, y and z coordinates for each point. Remember: the resolution is 2.5 cm, which means that roughly 2.4 billion points are displayed and mapped!

These points were now the start of the second coolest step in the process, after flying the drone. H created a three dimensional point cloud, where each point is displays according to its coordinates in a three dimensional room. Because this is not a flat photograph as before, but 3D, we could tilt and twirl it in space to see the real topography of the survey area. But that was not all, because this 3D depiction still contained houses and trees and all the other stuff on the actual land surface. Therefore it is also called a Digital Surface Model (DSM). But what we needed for the topography was just a model of the actual terrain, the land surface without trees and stuff. To obtain the so called Digital Terrain Model (DTM) Z&H needed to get rid of all the disturbing things in the point cloud, either by using color (remove all green: most shrubs and trees are gone), by using more or less right angles from the surface upwards (most houses are gone), and by manually removing everything else. Now a lot of areas were left white, without information on the actual altitude. Obviously the x and y coordinates are still there, as on a 2D map, only the third dimension is gone. This was now extrapolated from all the points surrounding the white spot. If the area around a house varies between 1.70 m and 1.84 m above sea level, it is very likely that the house itself stands on land with the same altitude. So the extrapolations are quite accurate and reasonable. With this last step Z&H got finally their topographic 3D DTM and subsequently their 2D map.

From photos towards a topographic map: a 3D point cloudBut this map is still just a map. You might have noticed that H&Z are the kings of cool technology, so just a map was not enough for them. No, they wanted to make it a really useful tool for the Government of Kiribati and especially for Kiritimati Island. So they had spent days strolling through the villages and noting the exact coordinates of each house, the height of its fundament, the material its build of, and if it’s commercial or residential. Now they added this layer to the map, to have a map with all houses exactly displayed on the topographic map, using a future-telling SPC software, also called PacSAFE. This tool allows to add variation of hazards like extreme wave events, to see inundation height and impact of inundation according to different climate change and thus sea level rise scenarios. Because just the elevated sea level, that the future will bring, is only half the truth. The real damage is done by waves, which might change the coast line, which might inundate parts of the island, which destroy buildings and natural landmarks. But this part of the work, entering wave data from normal “every day” wave height, direction and energy to events which occur statistically every 10 or 50 or 100 years and which is going to tell the real truth about threat levels, is still to be done for Kiritimati.

A few days after accompanying H&Z and seeing all the steps of their work, I checked in with them again to see, if they could finish their flights and survey. But what do they tell me? The steering mechanism of the new drone was broken after just three flights! Back to measuring the rest of the coast line via RTK.

F*ck the drone!