Spatial structure and scaling of agricultural networks

Considering agricultural landscapes as networks can provide information about spatial connectivity relevant for a wide range of applications including pollination, pest management, and ecology. At global scales, spatial networks of agricultural land use inferred from land cover products are well-described by power law rank-size distributions. However, regional analyses of agricultural land use typically focus on subsets of the total global network. In this paper, we seek to address the following questions: Does the globally observed scale-free property of agricultural networks hold over smaller spatial domains? Can similar properties be observed at kilometer to meter scales? Does the observed scale-free structure persist as agricultural networks evolve over the growing season? We analyze 9 intensively cultivated Landsat scenes on 5 continents with a wide range of vegetation distributions. We find that networks of vegetation fraction within the domain of each of these Landsat scenes exhibit substantial variability - but generally still possess similar scaling properties to the global distribution of agriculture. We also find similar results when comparing Landsat and Sentinel-2 imagery for 3 agricultural regions in Europe, as well as in an IKONOS image of an agricultural region of China. To illustrate an application of spatial network analysis, we show an example of network disruption. We compare two networks with similar rank-size distributions that behave differently when nodes are progressively removed. We suggest that treating agricultural land cover as spatial networks can provide a straightforward way of characterizing the connectivity and evolution of complex spatial distributions of agriculture across a wide range of landscapes and at spatial scales relevant for practical agricultural applications. (C) 2016 Published by Elsevier Inc.