Comparison of Surface Air Temperature Products from Reanalysis over United States and Ukraine: Application to Wheat Yield Forecasting

Abstract

In this work we analyzed the impact of using surface air temperature data from different reanalysis products on the forecast made by a winter wheat yield model. The forecast model uses information of the accumulated Growing Degree Day (GDD) during the growing season to estimate the peak NDVI signal [3], [4]. Four surface air temperature datasets generated by the NCEP2, MERRA2, JRA55, and ERA-Interim reanalysis projects were compared to NCEP1 data over the United States and Ukraine. For this, the bias was analyzed both spatially on a per-pixel basis, and temporally over the whole country. In both cases, the highest agreement with NCEP1 was found for NCEP2 and MERRA2. The per-pixel spatial analysis revealed that the largest differences (BIAS of up to 7.5°C) were found in pixels on mountainous areas of complex terrain. The temporal analysis of the spatially-averaged values showed a strong seasonality of the BIAS on both countries for all datasets, with a range of differences that varied from less than 0.1°C to 1°C during the summer months, to between 1 °C and ∼4°C during the winter months. Analysis of the model's forecasts revealed differences consistent to those found in the temperature analysis, which shows the sensitivity of the forecast model to different surface air temperature input datasets.