Ag-orithms: using free satellite imagery for agricultural
64% of agricultural activity was identified using free satellite imagery.
Question: how do you understand what is happening in agriculture in a region of the DR Congo? The catch is, you can't go there (think bad roads or insecurity) and limited data currently exists.
We recently faced this challenge when we were tasked with identifying the level of agricultural activity (incl. smallholder farming) in a 14,000 sq km area of DRC. Satellite imagery was the obvious data source. We had a few options: new to the scene Planet Labs network of nano-satellites or Digital Globe's comprehensive archives. Both were attractive but with a limited budget, it was hard to compete with the free EU Sentinel imagery.
So we build our algorithms, ran them for the AOI, refined them, ran them, refined them... you get the idea. We got this nice image which identified fields according to their spatial and spectral characteristics.
So now we can tell the number of fields, their size and density. We can tell if they are ploughed (bare soil) or have dense vegetation. We can start doing more interesting analysis like linking it to market data to study if farm size is impacted on market access.
Oooh, 'very pretty' you say, but how accurate is it? Great question, dear reader! Well we decided to dig into that and concluded that we identified 64% of all agricultural activity visible at 2.5m resolution.
Geo Gecko is using satellite imagery to identify crop types and monitor them over time.