Monitoring annual dynamics of mature rubber plantations in Xishuangbanna during 1987-2018 using Landsat time series data: A multiple normalization approach

• Published in: International journal of applied earth observation and geoinformation 2019-05, Vol.77, p.30-41
• Main Authors: Xiao, Chiwei, Li, Peng, Feng, Zhiming
• Format: Article
• Language: English
• Subjects: Landsat time series (LTS); Normalized burn; Normalized difference moisture index (NDMI); Ratio (NBR); Rubber plantations; Xishuangbanna
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2018.12.006

•A Re-Normalized Vegetation Index was firstly developed to map rubber plantations.•32-year maps of rubber plantations were generated in Xishuangbanna during 1987–2018.•Annual area of rubber plantations reached its peak in 2014 and then decreased.•Rubber plantations expanded extensively in Sino-Lao and Sino-Myanmar border regions. Accurately monitoring rubber plantations dynamics is essential for assessing eco-environmental effects in soil, hydrology and biodiversity especially in the northern edge of the Asian tropics (e.g. Xishuangbanna, China). In this study, a novel phenology-based multiple normalization approach was firstly proposed to annually map rubber plantations between two critical phenological phase (defoliation and foliation) in Xishuangbanna during 1987–2018. It included three key steps, namely (1) Landsat non-visible bands normalization for calculating the Normalized Difference Moisture Index (NDMI) and Normalized Burn Ratio (NBR), (2) normalization of NDMI and NBR for the Normalized Vegetation Index (NVI), and (3) re-normalization of NVIs for the Re-Normalized Vegetation Index (RNVI). The NVI highlighted the temporal differences of NDMI in land surface moisture content and NBR in soil moisture as well as vegetation survival of rubber plantations during the shifting period of defoliation and foliation. The RNVI fully considered the inverse patterns of the NVIs between defoliation and foliation phase. Rubber plantations were featured by negative NVI and RNVI values within the two temporal windows, while the positive NVI and RNVI values or zero stood for other non-rubber land cover types. The average overall accuracy of five-year mature rubber plantations maps (2010, 2013, 2015, 2016, and 2018) was up to 94.7% with the average kappa coefficient of 0.88, showing the great potential of multiple normalization approach. The total area of rubber plantations increased about 5.9 times from 1987 to 2018, showing clear expansion trends from centralization to scattering in Xishuangbanna as well as continuous spread in Sino-Lao (near Luang Namtha) and Sino-Myanmar (near Mongphak of Shan) border regions on the Chinese side. In addition, annual average analysis showed that about 91.4% of rubber plantations were invariably distributed around Jinghong City and Mengla County in the past decades.