Multiple factors drive regional agricultural abandonment

Abstract

An understanding of land-use change and its drivers in agroecosystems is important when developing adaptations to future environmental and socioeconomic pressures. Agricultural abandonment occurs worldwide with multiple potentially positive and negative consequences; however, the main factors causing agricultural abandonment in a country i.e., at the macro scale, have not been identified. We hypothesized that socio-environmental factors driving agricultural abandonment could be summarized comprehensively into two, namely "natural" and "social", and the relative importance of these differs among regions. To test this postulate, we analyzed the factors currently leading to agricultural abandonment considering ten natural environment variables (e.g., temperature) and five social variables (e.g., number of farmers) using the random forest machine learning method after dividing Japan into eight regions. Our results showed that agricultural abandonment was driven by various socio-environmental factors, and the main factors leading to agricultural abandonment differed among regions, especially in Hokkaido in northern Japan. Hokkaido has a relatively large area of concentrated farmland, and abandonment might have resulted from the effectiveness of cultivation under specific climate factors, whereas the other regions have relatively small areas of farmland with many elderly part-time farmers. In such regions, abandonment might have been caused by the decreasing numbers of potential farmers. Thus, two different drivers of agricultural abandonment were found: inefficient cultivation and decreasing numbers of farmers. Therefore, agricultural abandonment cannot be prevented by adopting a single method or policy. Agricultural abandonment is a significant problem not only for food production but also for several ecosystem services. Governments and decision-makers should develop effective strategies to prevent further abandonment to ensure sustainable future management of agro-ecosystems.

Keywords: Agricultural ecosystem; Land-use change; Machine learning; Natural factors; Random forest; Social factors.