A Hindcast Experiment Using the GCAM 3.0 Agriculture and Land-Use Module
Model validation is an important scientific activity. In a model validation method called hindcasting, a model is run over a historical time period, testing its ability to reproduce the observed past. This paper reports the results of a hindcast experiment using the Global Change Assessment Model’s (GCAM) land-use module developed at the U.S. Department of Energy’s Pacific Northwest National Laboratory (PNNL). Hindcasting is widely used by the Earth system modeling community but is novel in the integrated assessment modeling community.
PNNL researchers showed that hindcasting can be a useful tool for the integrated assessment modeling community. This exercise demonstrated that hindcasting can help researchers improve model performance and identify strengths and weaknesses in models, which in turn can be used to guide future research.
Researchers employed the GCAM land-use module to conduct a hindcasting experiment. They initialized GCAM to start a forecast in the year 1990, giving the model only information that it would have had in 1990. Researchers then ran GCAM over a historical period for which key input variables such as gross domestic product, population, and crop yields are known. GCAM used this information to estimate demands for and supplies of all land products. It estimated production, and the allocation of land for all alternative purposes such as crops, urban areas, and unmanaged ecosystems. Researchers then compared GCAM model projections with historical estimates of crop production and harvested land area reported by the United Nations Food and Agricultural Organization. Researchers showed that integrated assessment models such as GCAM could benefit from such exercises. This particular exercise showed the importance of model structure, particularly with regard to the way that land-use decisions are made. The exercise also demonstrated that model performance could be assessed using some of the same statistical measurement tools that Earth system modelers use in evaluating Earth system model performance. This paper also demonstrates the importance of model validation for the integrated assessment modeling community.