An adaptive spatiotemporal agricultural cropland temperature prediction system based on ground and satellite measurements

Temperature data at cropland level is important for precision farming, yield forecast and agricultural risk management. On the other hand temperature measurement tolerance rises due to topographic and structural variation when this data is generated with respect to temporal measurements at reference stations. Use of agrometeorological measurement devices at each cropland has operational sustainability and data fusion integrity problems when the average cropland size is reduced. Temporal to spatiotemporal temperature data conversion in real time appears as a solution when a common operated agroinformatic network is installed. This kind of data regeneration requires manageable and bounded average and absolute error rate in agricultural applications since each cropland has different temperature sensitivities. Spatial layout of temporal data is possible by using land references together with a model that was adapted or calibrated by spatial data of the same region. Surface heat capacity, topologic structure and tissue change increase local heterogeneity in smaller areas. Land surface temperature (LST) is a way of spatial temperature measurement via remote sensing satellites but it is mainly affected by material of the surface. In this study, we proposed a method that uses classified land surface temperature (LST) maps from satellite images and used them together with Inverse Distance Weighted Interpolation (IDW) method. We have shown that adaptive LST modified IDW yields mean absolute error (MAE) better than both IDW and IDW with elevation correction (IDW-EC).