Exploring global irrigation patterns: a multilevel modelling approach

Globally, areas that are equipped for irrigation have almost doubled in size over the past 50 years and further expansions are expected for the future, to meet a growing food demand. For developing countries, the Food and Agriculture Organization of the United Nations (FAO) expects these areas to be expanded by 40 million ha, by 2030. Knowledge about the constraints to irrigation and spatially explicit information about the potential for irrigation expansion, however, are lacking on a global scale. The objective of our study was to explain the global pattern of irrigated croplands and to identify cropping regions where irrigation is likely to be expanded. We accounted for biophysical determinants, such as humidity and slope, mainly at grid-cell level. Socio-economic and governance determinants, for example, Gross Domestic Product (GDP) and control of corruption, were primarily considered on a country level, given the limitations in availability of sub-national data and the role of national level governance in irrigation decisions. To identify the variability of the determinants within these two spatial levels, we conducted a multilevel analysis. This is a method employing regression models that explicitly account for hierarchically structured data. Results show significant variability in terms of irrigation. While 56% of the global variance in irrigation occurs between countries, 44% occurs within countries. Our results suggest that it is necessary to consider biophysical, socio-economic and governance information for identifying cropland areas that are likely to be under irrigation. Under current conditions, conversion from rainfed to irrigated cropland is most likely in eastern China, northern Africa, and parts of the Mediterranean region.