+27 (0)21 300 2098 | +27 (0)83 647 5777 | +256 792 998107

Inchanga, KZN, South Africa

Ggaba Road, Nsambya hill,  Kampala, Uganda


  • Charlie Wijnberg

NDVI Drone Survey – Nampula, Mozambique

The NDVI study was based in Nampula Province, Mozambique. The primary crop grown on this flat 3000ha farm is soya bean but maize and other crops have been tested historically. The land has only been farmed for 8 years and the major problem encountered has been the water retention of the soil in the area. Despite the soil appearing to be of good decomposed granite with small gravel with even water distribution, there are also clay particles which inhibit this. Water can be found pooling all over the flat 3000ha farm and despite all efforts to locate and plan draining or diversion of this standing water, it became apparent that this challenge could only be overcome by adopting very recent aerial imaging to isolate and prioritize these areas. The farm decided to utilize UAS to identify these areas requiring urgent attention and also to estimate the coming crop yields. 3DroneMapping was asked to assist in a trial capacity. If the results of the survey were of any benefit, further surveys could be budgeted for.

Day 1

Located at an average of 660m AMSL, the area appears flat and covers 3000ha, however only 1400ha had been planted this year that required survey. These areas had been mapped with a handheld GPS device and these were handed over to survey. The client asked for NGB imagery and then to be converted to pseudo NDVI for purposes of identifying those zones of good plant growth. The bodies of water would also be located as a by-product of the imagery given the relationship between near infrared and water being easily identifiable. Day 1 was spent traversing the entire farm placing control points for use in ortho rectifying the final model. The decision was made to place control was based on the fact that it would be resurveyed to compare the same areas in the coming year to see if the changes made would have any effect. Permanent control points were placed near an existing runstrip used for aerial spraying. Temporary control was evenly dispersed over the site and surveyed using a Leica GPS 1200 RTK survey grade GPS to centimetre accuracy. Those locations where RTK was not possible were surveyed using post processing methods with similar accuracies as the RTK points. Weighted down fertilizer sacks were used as makers however overnight 15% of these were repurposed by locals. It was realised that some of the locals had replaced the sacks after confrontation with a farm manager which was problematic as these were not replaced in the same positions. These points were later identified in the helmert transformation part of the model generation as being outliers and were rejected. The remainder of the day was spent identifying the areas to be flown and flight planning as well as a test flight to check the UAS radios and camera setup.

Day 2

The weather on day 2 was far from ideal. Low cloud from the evenings 50mm of rain began at about 400m AGL. It was decided to wait until 10am to begin the survey. At 10am the weather still had not improved and it was predicted for isolated showers in the afternoon. After consultation with the client, it was made clear that the survey would have to take place regardless of the conditions as the soya plants were on the verge of ripening and time was of the essence. With this in mind, the team reluctantly prepared to fly the first 400ha northern area. The initial setup of the 3DroneMapping “Inyoni” fixed wing UAS took 10min to prepare and check. A modified CanonS110 camera with filter to capture near-infrared, green and blue spectrumand a 10A battery being the only payload. The flight plan was set to climb to 200m AGL and then head north to 300m to begin the strip survey routine. The strip was set at 131m between lines with an image set to capture every 118m. This provided an overlap of 73% and a sidelap of 60%. The estimated flight time was 93 minutes. The take-off went well with 30-40 amp draw to climb to the first 200m AGL waypoint. The UAS landed without any intervention as planned after spending 99min airborne. It was noted that the end capacity of the battery would still allow for 35-40 min of safe flying.

The southern area was divided into 2 section. The southern area being 370ha and the northern 590ha. Both areas were flown successfully with autonomous take-off and landing sequences. It was noted that the barometer shifts upon land varied about 6-10m but given the topography, we had the luxury of open space to allow the safe landing. However of the course of the day, the partial clouds caused shadows over the ground. As the average flight lasted 100minutes, these shadows were unavoidable.

The resulting areas were then processed in Agisoft Photoscan to produce orthomosaics. Control previously survey was added and strengthed the model. It should be noted that raw IMU, GPS and barometer data added to the metadata to the images sped up the alignment process in Agisoft at least by 50%. The end results proved good with RMS values of less than 10mm. It was decided to combine the central and southern areas to produce a large NGB orthomosaic as there was sufficient overlap to merge the chunks.

The resulting NGB images were then put through FIJI imagining software to calculate pseudo NDVI values and then projected to give UTM coordinates.

Day 3

Overnight it was decided to to fly a new area with a RGB camera to provide continuity between the northern and southern locations. This 500ha area was flown in 95min and then processed to produce a RBG othoimage. All images were then placed in GIS software and handed to the client for analysis.


A total of over 2100ha was surveyed in a total of 6.5hours of flying. The end pixel size derived was averaged to 15cm/ pixel. It should be noted that high resolution was not required by the client for plant health purposes. This enabled us to fly at 300m AGL giving such good coverage. The end raster images were inserted into a GIS package suitable for the clients and their shareholders to view and analyse. This has given an enormous advantage to farm managers to identify problem areas that normally are not visible from the ground especially on such flat terrain. Initial analysis has shown that the previously cleared ant mounds have affect the soils conditions to perfectly round 17m radius where the plants are not growing well.

These average at about 6 per hectare. The original slash and burn lines promote good plant growth possibly due to being elevated or the added carbon. The flat areas that collect water have severely repressed plant growth and encourage the growth of more water resistant grass species. While the cloudy conditions have not been the best for NGB data collection, some information relating to plant health was derived. On a large scale area such as this, NGB information does have its place. It is impossible for farm managers to accurately identify those zones that are not producing well and this tool enables them to prioritize and positively identify those locations and their impact on the plant growth.

#GIS #NDVI #PrecisionAg #Mozambique #agriculture