Characterisation of global cropland bright spots

Agriculture has major impacts on the environment: it is the first cause of biodiversity loss, freshwater withdrawals and nutrient flows disruption, as well as an important source of GHG emissions. At the same time, it is an essential human activity to sustain the life of current and future generations. Thus, a drastic increase in food systems sustainability is crucial in the coming years. To address this huge challenge, a mix of local- and global-scale studies assessing impacts and exploring possible solutions are needed. At the global scale, studies that are spatially-explicit and account for multiple impacts are particularly precious. Such studies often focus on hotspots of environmental degradation and tend to overlook the analysis of existing best practices.

In this work, we instead look at bright spots, that we define as regions where agricultural production is relatively important but does not cause the exceedance of local environmental sustainability thresholds. Making use of a circa 2020, 5 arcmin resolution dataset on global crops distribution and four associated environmental sustainability indicators (biodiversity loss, freshwater stress, excess nitrogen application and GHG emissions), we derive bright spot maps for 46 crop categories including individual cereals (wheat, maize, rice, barley, …) and other major crops (soybean, rapeseed, …).

We then train a random forest classification model to identify bright spots based on a number of land-use, biophysical and socio-economical variables. Using feature importance metrics such as SHAP values, we identify key characteristics of these regions.

Further, we simulate several prospective scenarios assuming the widespread adoption of the best practices identified, such as allocating more land to natural habitat, reducing irrigation and fertiliser use, or establishing crop rotations. We quantify the consequences of these scenarios in terms of agricultural production loss and sustainability increase, and estimate their ability to feed the human population by combining them with different human diet scenarios.