ICTworks

Development programming is full of information gathering, such as farm size. In a program I once worked on, smallholder farmers had to have less than three hectares to participate. Other than community involvement and random house checks, we often have few ways of verifying this information.

Crop and farm boundaries can be very valuable information for many reasons:

• We can make better recommendations in crop management based on location (on a hill, in a valley, the soil type etc.).
• Farm size and crop type can also provide a more robust market planning and profitability analysis.
• We can also provide fertilizer and irrigation recommendations for many crops with lower-cost high-resolution satellite imagery.

However, farm and crop boundary information can be costly for development programs to gather and analyze due to several reoccurring issues:

• Farmer Location: they are often out working in the field where there isn’t cell connectivity;
• Farm Boundary: You have to walk the boundary with the farmer, which can take an hour;
• Precise Location: You have to have a GPS enabled device(s) that collects location offline;
• Multiple Staff: You will probably need multiple people to have this done in a timely manner.

Sometimes satellite imagery can an affordable option to find farm boundaries if you are not planning to be in the field, but there has been a lot of hype about drones as a better option. Drones provide very high-resolution near real-time imagery often at a low cost if you are already paying to be in the field, hence the question:

Can We Measure Farm Size Using Drone Imagery?

We took a fixed wing and a quadcopter drone to Benin to see what information we could acquire with cashew farmers. We wanted to see if we could collect boundary information by flying drones instead of walking farm boundaries.

We used ArcGIS collector to walk farm boundaries, this was considered our baseline for comparison. After flying the farmers’ fields, we asked farmers if they would draw their boundary on top of the drone imagery.  We had three levels of results:

• Fixed-wing drone: Our internet dongle was not strong enough to collect flight information from the fixed wing aircraft, which meant that the flight operators were flying blind – making much of the imagery unusable.
• Quadcopter drone: We could see flight paths for the quadcopter drone without internet, we had full imagery sets but for a much smaller area, and only a total of 12 farms where the quadcopter was flown
• Handheld GPS: We walked 50 farmers’ fields using traditional GPS survey methodologies.

When we compared the boundary information for the drawn vs walked fields, we had a standard deviation for drawing outwards of 0.05 hectares and a standard deviation for drawing inwards of 0.1 hectares per plot. This error could be further reduced with practice and repetition. A larger sample size would also be more indicative, but this is a good preliminary analysis.

Yes, We Can Measure Farms, But…

In looking at the two GIS data collection methodologies using unmanned aerial vehicles, here were our conclusions:

Fixed wing drones

Fixed wing drones are more expensive to operate; require a smooth flat landing area; have more regulations; consent is required from farmers that are not participating and happen to be in the flight path; and, they require high bandwidth/internet to see the flight path and if areas are missed.

Conclusion: not a good fit for small farms close together where monitoring staff will fly the drone.

Quadcopter drones

Quadcopter drones are more price accessible; have less regulations; fly less areas; require less training; can be flown only for participating farmer fields; can transmit imagery and flight areas without internet; and, they can take more time to fly.

Conclusion: a better fit for small farms close together with monitoring staff able to fly the drone.

Fixed Wind vs. Quadcopter Drones

So, to walk, to quadcopter, to fixed wing, or to satellite imagery – that is the question.  The answer to this in my opinion is . . . it depends.  Here are my recommendations:

• Small farmer fields that are close together where you already have frequent on the ground monitoring- chose a quadcopter drone
• Large farmer fields that are close together, where you need real-time imagery over a large area, choose a Fixed wing drone.
• When you need to understand what has been happening in a large area over the past 6 months and field teams were not covering it, choose medium resolution satellite imagery that surveys the entire earth every 5 days.
• When you need to see what is happening now in detail in an area greater than 20 sq km, choose high resolution imagery from a satellite you have tasked.

In many of the conversations I have had around drones, no one is talking about field travel costs or whether a time series is needed. I am happy to see more programs using imagery analysis whether drone or satellite based. However, please include full cost accounting: field travel, flight time, equipment costs over time, trained staff, and the number of areas to be flown.

In the pilot in Benin, the drone provided a lot of extra analytical benefits that can be a great add on if needed. We could see where farms needed leveling, weeding, and areas that could use more planting. There are many algorithm providers out there that can give you crop stress and, if you have grain crops, fertilizer recommendations.

However, before making a go / no-go decision on when to use a drone, make sure that operational costs are included, and field agents have the bandwidth to follow up with individual farmers.

Why Fly at 60m When You Can Fly at 600km?

Satellites used to be expensive, low resolution and provided poor coverage over Africa. They now survey the earth every day, images are available for free and analyzing them can be automated. There are extremely exciting innovations . Why should we smallholder watchers care? Here’s three reasons.

Satellites Can Scale

Setting up a remote sensing agricultural monitoring system can be somewhat complicated, requires technical skill and may be somewhat costly in the short term when compared with deploying drones or field teams with GPS. Soon after, economies of scales kick in and the unit cost of monitoring one more field is a fraction of the traditional methods.

At reasonable scale, Geo Gecko delineates a field using high resolution satellite imagery for $0.33 (all costs included). Compare this with $1.40 per field using field data collectors. Surveys using drones will be multiples of this figure given the additional hardware required and skilled technicians piloting. Cloud computing, higher resolution imagery and constantly refined algorithms mean this price will keep reducing and the quality increasing.

Satellites Take More Than Pretty Pictures

The various satellite sensors can tell you some surprising things about what is happening on the ground. Is a pixel in an image concrete or vegetation? If the latter, how healthy is it? How much moisture is in the soil? What temperature is the ground? How much rainfall does it get? What is the elevation and how steep is the slope? All this data is embedded within an image.

Satellites Can Time Travel

With satellites you can time travel with time series data. Historical data of the entire earth are available for the past 40 years. A drone can record a field in ultra-high resolution and a field data collector can tell you how many children the farmer has but they can’t chart the health of crops in that field each month for the past decade. Satellites can.

True, the data won’t be perfect but if there is one thing that developing countries need, it is reliable data. From this, patterns and trends can be discerned at a national scale so that evidence-based, transparent decisions can be made.

Coauthored by Kathryn Clifton of Catholic Relief Services and and Bernard Wright of Geo Gecko.