Abstract
Agricultural services often need to represent variations in expected benefits from input use within a designated area. Our aim was to produce maps that integrated data from research trials with researchers’ knowledge using weight of evidence modeling. A validation data set was used to generate maps indicating areas in which response to fertilizer was more or less likely. Maps of predicted yield were also generated for Takeo province, Cambodia using EXPECTOR software to combine, using Baysian statistics, weighted evidence from soil, a flood risk and N fertilizer response maps. The data set comprised 2336 on-farm trial sites. The two sets of maps were in broad agreement, suggesting that provincial-scale maps of yield and yield response could be generated using the weight of evidence approach when there is a low density of field trial data. Discrepancies between the two data sets were only significant where the spread of map data points exceeded the maximum likelihood distance and where increased N fertilizer rates appeared to exacerbate insect damage to crops. At the provincial scale, the maps identified areas that can be used to target advice on appropriate fertilizer rates. This could enhance the strategic planning capability for delivery of fertilizer inputs.