Climate and Economic Triggers of Cropland Change in the US: A Fine-scale spatial analysis

Abstract:

There is a large number of studies that have examined the extent and determinants of land-use change. However, these studies are either not causal or analyze changes at an aggregate level. Our analysis is at a fine spatial resolution and examines the extent to which a broad range of biophysical, climate, and market factors trigger a change in the use of land for crop production. We are particularly interested in analyzing the land that is at the margin of crop production and frequently transitioning between crop and non-crop use. We are interested in examining which land is more likely to change use and which factors are important drivers of this change. We compile a novel data set that includes land-use history, frequency of land conversions, climate information, and biophysical characteristics at a 500-meter resolution for the continent of the U.S. in 2000-2017. We first exploit the variations in extreme weather events during the planting seasons, and levels of market proximity in a panel data model. We then develop a new approach that incorporates machine learning and recurrent survival analysis framework to address the “displacement” or “censoring” concerns documented in related literature. Finally, we utilize the machine learning inference to explore the treatment effect heterogeneity across land pixels with various biophysical characteristics. Both approaches find the effects of extreme weather events and market proximity are less widespread and mainly confined to the low-quality land in the midwestern region. Our findings also indicate that the change in the use of land for crop production may not be inelastic and static even in the medium-term.