Genetic algorithm for assimilating remotely sensed evapotranspiration data using a soil-water-atmosphere-plant

Agricultural monitoring is necessary for efficient food security management at country level. Typically, monitoring requirement from the point of view of an agricultural/irrigation manager would be to “see” each field at a regular interval to which 15 days is reasonable. Evapotranspiration (ETa) is converting the water into crop, and is therefore a crucial indicator of crop productivity. ETa can be estimated from satellite remote sensing [1] [2] [3]. However, on the side of satellite platforms specifications, high spatial resolution is at about size of the largest fields (~1 ha), but is available only few times a year practically, while low spatial resolution is available daily (even 8days composites are ready from Internet). A potential solution would be to match the two type of satellite images evapotranspiration by running instances of crop models at both resolutions with proper parameters. Those crop model input parameters are changing on pixel-to-pixel basis, therefore using a data assimilation method is most interesting to try and solve this problem. Because the search domain of this assimilation problem is multi-dimensional and highly non-linear, using an evolutionary search algorithm like the genetic algorithm (GA; [4]) is preferred. Similar work by [5] and [6] used some remotely sensed information combined with GAs and the Soil Water Air Plant model (SWAP) in the objective of optimization of soil hydraulic parameters. This paper describes the implementations issues of the program [GA+SWAP] whereby the ETa data from remote sensing images could be used to generate the ETa data during the periods when satellites are not available. SWAP is used by a GA to estimate pixelbased plant/water parameters controlling the pixel ETa observed by the satellite images. Two open source options have been investigated for geographical linkage, the first one being GRASS GIS [7] and the second one being a remote sensing image handling library [8].