by Wilco Terink, Futurewater
Crop growth models play a major role in sustaining the world-wide food security. These models are used to simulate crop growth during the growing season, and the final crop yield at the end of the growing season, given the farmers’ management practices. At a more strategic level, these crop growth models play an important role to decision makers to take timely decisions regarding food import and export strategies.
The simulation accuracy of crop growth models relies on the quality of the input data. Since crop yield forecasting applications are often applied over large areas that rely on a spatially distributed crop growth model, the uncertainty in the spatial variation of the input data increases.
Remote sensing images are often used in crop growth models because remote sensing images provide spatially distributed input data to these models. These images are available in numerous spatial resolutions, where coarse-resolution images are often freely available compared to the more expensive high-resolution images. Therefore, the objective of this study was to evaluate the added value of high-resolution satellite imagery above coarse-resolution satellite imagery in crop yield forecasting.
The focus of this project was on a small area in the Nile Delta in Egypt where berseem and wheat are grown during winter season. To evaluate the added value of high-resolution satellite imagery above coarse-resolution satellite imagery in crop yield forecasting, the Soil-Water-Plant-Atmosphere (SWAP) model was driven by high-resolution ASTER imagery and coarse-resolution MODIS imagery. The imagery consisted of remotely sensed Leaf-Area-Indexes (LAIs), which indicates the vegetation coverage in a certain area; a higher LAI corresponds to a higher crop coverage in the area. The yields from the high- and coarse-resolution runs were compared with a reference situation, which can be seen as the yield that would have been measured in the field. For the forecasting period (2 months before harvest), no remotely sensed LAIs are available. Therefore, the SWAP model was driven with ‘standard’ LAI values for the reference run and the high- and coarse-resolution runs.
It was concluded that a farmer can obtain a much more accurate yield forecast if high-resolution satellite imagery is used. If coarse-resolution remote sensing would have been used, then for wheat the yield is overestimated with approximately 9%, and for berseem with 26%. If high-resolution satellite imagery will be used, then the forecast accuracy is much better; wheat underestimated with 1.4% and berseem overestimated with 2.1%. The use of high-resolution remote sensing enables the farmer to optimize his local specific farm and water management; irrigation applications are more accurate under the use of high-resolution imagery. This is especially true for berseem, where the reference situation shows that 442 m3/ha is really required. Based on the best estimate (high-resolution remote sensing) 486 m3/ha is applied (10% overestimated), while under coarse-resolution satellite imagery zero irrigation is applied (100% error). For wheat the reference situation shows that 1148 m3/ha is really required. Under high-resolution remote sensing, the applied amount of irrigation is 1093 m3/ha, which is a slight underestimation (~5%), while under coarse-resolution remote sensing 1194 m3/ha is applied, which is a slight overestimation (~4%).
At a more strategic level, decision makers will have considerable advantages if high-resolution remote sensing is used in crop yield forecasting; they can better take timely decisions regarding food import and/or export strategies. It was concluded that a significant amount of money can be saved if high-resolution remote sensing is used in crop yield forecasting. If coarse-resolution remote sensing is used in crop yield forecasting for wheat, then approximately 250 million US$ is lost through less export due to overestimated wheat yields. High-resolution remote sensing for wheat would lead to underestimated wheat yields, meaning that decision makers import too much wheat for a price of approximately 100 million US$. Losses are even more significant for berseem. Both coarse- and high-resolution remote sensing results in overestimated berseem yields. If coarse-resolution remote sensing will be used for berseem, then 3.2 billion US$ is lost through less exports, whereas high-resolution leads to a loss of approximately 0.3 billion US$. It was concluded that costs for the use of high-resolution remote sensing are negligible small compared to these benefits, and therefore it is well-worth investing in high-resolution remote sensing images.
For a follow-up it would be very interesting to focus on a larger area containing many coarse- and high-resolution images; this will increase the variety of crop types and at the same time the model input uncertainty.
Read the report: The added value of high-resolution above coarse-resolution remote sensing images in crop yield forecasting: A case study in the Egyptian Nile Delta by Terink W, Droogers P, Van Dam J, Simons G, Voogt M. 2012. FutureWater.