Implications of input data aggregation on upscaling of soil organic carbon changes

Abstract

Dynamic process models are increasingly used to predict changes in soil organic carbon (SOC) stocks of agricultural soils on the large scale. This study examines the aggregation effects of climate and soil data on regional SOC modeling for varying simulation periods based on a multi model ensemble. For a NUTS2 region in central Europe (North Rhine-Westphalia) data on soil properties and daily weather available on a spatial resolution of 1 km have been aggregated to 10, 25, 50 and 100 km resolution. Soil data aggregation (DA) showed a bigger effect on modeled SOC stock changes than climate DA, which was one order of magnitude smaller. The DA effect determine the spatial resolution of model output (scale of interest). Model errors, calculated as the difference between respective DA level and 1 km outputs, were high at low model output DA level (scale of interest: 1 km) and decreased with increasing scale of interest (10-100 km). Additionally, a large variability of simulated SOC contents amongst models was observed. Contrary to model errors induced by input DA, this variability was not leveled out by increasing the scale of output data. The regionalization of SOC stocks and changes is highly influenced by input DA. Factors like the length of the modeling period, the modeling region and the type of input DA control the resulting errors. The presented study describe a detail of these relationships.