The global consumption of natural rubber is expected to grow by 3.7% per year to 13.8 Mio tonnes in 2018. Soaring rubber prices have led to rapid land use conversion to rubber in mainland SE Asia and although prices have fallen since 2010 rubber farming continues to be lucrative. An ability to predict where rubber is likely to be planted next is key to targeted policy. We show that the environmental space where rubber plantations have traditionally been established matches the environmental conditions of the natural South American range of Hevea brasiliensis. As much of this ‘traditionally suitable’ environmental space for plantations in SE Asia is now occupied, we have investigated the more recent spread of plantations into new environments and whether this expansion follows a spatially predictable pattern (using MODIS NDVI 250m derived maps for 2004/05 and 2009/10*). The spread of rubber 2004/05 – 2009/10 was strongly related to proximity to the nearest neighbour plantation (70% explanatory power). Of these new plantations 96% were established in non-traditionally suitable environments. Novel environments may not per se be suboptimal for specifically bred rubber clones, but 23% of the plantations were found in areas where temperature seasonality differed by >50% from the native rubber range, and almost a quarter of these replaced areas that prior to 2000 were under forest. Sudden weather extremes in areas with greater climate variability may push even adapted clones to their tolerance margins. Our results suggest that there is a rapid and partly predicable spread of plantations into marginal and potentially risky environments, often at the expense of high-biodiversity value land. We highlight areas that may be next in line for conversion to facilitate targeted policy interventions that prevent “loss-loss” scenarios, e.g. the conversion of high-biodiversity land in areas where environmental stresses may render rubber farming ultimately difficult.

* Li Z & Fox JM (2012) Mapping rubber tree growth in mainland Southeast Asia using time series MODIS 250 m NDVI and statistical data. Applied Geography 32(2):420-432.