C.R. Mi - Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, University of Chinese Academy of Sciences, Beijing ۱۰۰۱۰۱, China.
Q. Zu - Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, University of Chinese Academy of Sciences, Beijing ۱۰۰۱۰۱, China.
L. He - National Meteorological Center, Beijing ۱۰۰۰۸۱, China.
F. Huettmann - EWHALE Lab, Department of Biology and Wildlife, Institute of Arctic Biology, University of Alaska Fairbanks (UAF), ۴۱۹ Irving I, P.O. Box ۷۵۷۰۰۰, Fairbanks AK ۹۹۷۷۵, USA.

China’s sugar production and consumption continues to increase. This process is alreadyongoing for over 15 years and over 90% of the sugar production comes from sugarcane(Saccharum officinarum). Most of the sugarcane is planted in the south (e.g. the Chineseprovinces of Yunnan, Guangxi, Guangdong and Hainan) and it represents there a majoreconomic crop in these landscapes. As found virtually worldwide, climate change is generallyexpected to influence such suitable planting areas. Here we started a first empirical assessmenthow climate change would influence the spatial distribution of those current and future suitableplanting areas of this strategic crop in China. We employed an ensemble machine learningalgorithm (Random Forest; bagging) and increasingly used and robust species distributionmodels (SDMs). These are based on our compiled and best publicly available crop data sampledfrom the Chinese sugarcane industry map. They were linked with bioclimate variables fromthe Worldclim database. This powerful concept allowed us to project sugarcane’s current andfuture (2070) suitable distributions based on the climate niche. Our results were extrapolated tothree Global Circulation Models (GCMs; BCC-CSM1-1, CNRM-CM5 and MIROC-ESM)under three representative concentration pathways (RCPs of 2.6, 4.5 and 8.5). The evaluationsof these models indicated that our results had a powerful performance (AUC=0.97, TSS=0.96)for robust inference. Bioclimatic variables related to temperature were the most importantpredictors for sugarcane planting. All models showed similar increasing spatial trends insuitable distribution area and just a few original suitable areas would be lost. Our finding putsemphasize on new growing areas, their soil and management. It is the first to provide thenecessary background in the future to safely cultivate sugarcane in climate-suitable areas and toobtain more sugar production for farmers and the industry; it is of large and strategic importancefor food security and national autonomy of this central commodity.

Sugarcane, Climate Change, China, Species distribution model (SDMs), Random forest (bagging) and machine learning, Food security