<p>Agricultural ecosystem models, driven by climate projections and fed with soil information and plausible management scenarios are frequently used tools to predict future developments in agricultural landscapes. On the regional scale, the required soil parameters must be derived from soil maps that are available in different spatial resolutions, ranging from grid cell sizes of 50 m up to 1 km and more. The typical spatial resolution of regional climate projections is currently around 12 km. Given the small-scale heterogeneity in soil properties, using the most accurate soil representation could be important for predictions of crop growth. However, simulations with very highly resolved soil data requires greater computing time and higher effort for data organization and storage. Moreover, the higher resolution may not necessarily lead to better simulations due to redundant information of the land surface and because the impact of climate forcing could dominate over the effect of soil variability. This leads to the question if the use of high-resolution soil data leads to significantly different predictions of future yields and grain protein trends compared to simulations in which soil data is adapted to the resolution of the climate input.</p><p>This study investigated the impact of weather and soil input on simulated crop growth in an intensively used agricultural region in Southwest Germany. For all areas classified as &#8216;arable land&#8217; (CLC10), winter wheat growth was simulated over a 44-year period (2006 to 2050) using weather projections from three regional climate models and soil information at two spatial resolutions. The simulations were performed with the model system Expert-N 5.0, where the crop model Gecros was combined with the Richards equation and the CN turnover module of the model Daisy. Soil hydraulic parameters as well as initial values of soil organic matter pools were estimated from BK50 soil map information on soil texture and soil organic matter content, using pedo-transfer functions and SOM pool fractionation following Bruun and Jensen (2002). The coarser soil map is derived from BK50 soil map (50m x 50m) by selecting only the dominant soil type in a 12km &#215; 12km grid to be representative for that grid cell. The crop model was calibrated with field data of crop phenology, leaf area, biomass, yield and crop nitrogen, which were collected at a research station within the study area between 2009 and 2018.</p><p>The predicted increase in temperatures during the growing season correlated with earlier maturity, lower yields and a higher grain protein content. The regional mean values varied by +/- 0.5 t/ha or +/-0.3 percentage points of protein content depending to the climate model used. On the regional scale, the simulated trends remained unchanged using high-resolution or coarse resolution soil data. However, there are strong differences in both the forecasted averages and the distribution of forecasts, as the coarser resolution captures neither the small-scale heterogeneity nor the average of the high-resolution results.</p>
Comparing three approaches of spatial disaggregation of legacy soil maps based on DSMART algorithm