An analysis of methodological and spatial differences in global cropping systems models and maps

AimAgricultural practices have dramatically altered the land cover of the earth, but the spatial extent and intensity of these practices is often difficult to catalogue. Information on the distribution and performance of specific crops is often only available through national or subnational statistics. Recently, however, there have been multiple independent efforts to incorporate the detailed information available from statistical surveys with supplemental spatial information to produce a spatially explicit global dataset specific to individual crops. While these datasets provide decision makers with improved information on global cropping systems, the final global cropping maps differ substantially from one another. This study aims to explore and quantify systematic similarities and differences between four major global cropping systems products and the subsequent implications for analyses dependent on those models. LocationThis study was conducted at a global scale. MethodsEach global cropping systems model was assessed by latitude as a measure of biophysical plausibility and each pair of models was compared using a Gaussian filter to remove trivial spatial discrepancies. Model disagreement was explored in relation to the interdependent input data of each model pair with a particular focus on cropland extent. The influence of the observed model differences on subsequent analyses was demonstrated using model-dependent estimates of the yield gap as an example. ResultsThe results of our analysis indicate that the choice of cropping systems model is non-trivial: considerable differences exist between model-specific estimates of the yield gap across nearly all climate zones and the average model difference exceeds the average estimated yield gap in certain regions. The differences in crop-specific harvested area and yield products of each model are significant, and mostly result from differences in the input datasets and downscaling methodologies. In particular, the choice of dataset on cropland extent proved to be influential regardless of the downscaling process employed. Main conclusionsDiscrepancies in the final products of cropping systems models are currently poorly understood, but have implications for basic policy decisions relating to agricultural production and food security. The considerable disagreement between models serves as a reminder of the ongoing challenges to the creation of spatially explicit estimates of harvested area and yield based on crop statistics. Our analysis helps shed light on the importance of model choice by demonstrating the implications for further analyses that depend on cropping systems models, and works to overcome these challenges by characterizing model-dependent differences in harvested area and yield.