Spatio-Temporal Changes Of Rubber Plantations Based On Multi-Source Remote Sensing

As an important strategic resource, natural rubber is a kind of renewable resources and plays a crucial role in the development of countries. Rubber trees (Hevea brasiliensis) are the only source for natural rubber production in the world. Driven by industrial society and economic benefits, rubber dramatically expanded all over the world in the past decades. Global demands for natural rubber production and increasing rubber prices have promoted great rubber planting to enter an era of unprecedented prosperity in China. At present, Xishuangbanna Thai Nationality Autonomous Prefecture (Xishuangbanna), Yunnan, is the second-largest base in natural rubber plantation and production in China, which is also a typical region with density rubber plantations in northern border area of tropical regions. Rubber plantation expansions have mainly dominated the land use and land cover change (LULCC) of Xishuangbanna in recent decades. The expansion of rubber plantations in Xishuangbanna is mainly converted from tropical rain forests, which leads to a series of serious environmental and ecological problems relating to sustainable development, including regional droughts, and biodiversity decrease. Hence, understanding the spatio-temporal dynamics of rubber plantation expansions is important for regional environmental protection and sustainable development. Monitoring rubber plantation dynamics becomes an international hotspot in various research fields. The development of the remote sensing and Geographic Information System (GIS) technologies enable monitoring spatio-temporal dynamics of rubber plantations. This study presented a monitoring model for spatio-temporal expansion dynamics of rubber plantations based on unique and stable phenology features through the analysis of multi-source remote sensing data. Then this model was applied to monitor the spatio-temporal dynamics of rubber plantations in Xishuangbanna.An rubber plantation mapping algorithm, a stand age prediction algorithm, and a spatio-temporal monitoring model of rubber plantations were established based on multi-source satellite data and phenology characteristics. This study mapped accurate areas, reconstructed stand ages of rubber plantations, and discovered its expansion characteristics in Xishaungbanna. The study not only provides novel theories, methods, and technologies to identify and understand the area changes, stand age, spatio-temporal expansion pattern and its crucial driving forces of rubber plantations, but also provides scientific basis for understanding regional LULCC and its ecological and environmental effects. The main results and findings of this study includes the followings:(1) Combining growth characteristics of rubber plantations, this study developed a novel method for extracting rubber plantations area by multi-source remote sensing data. Firstly, this study found defoliation (early January to middle February) and fast growth (Late February to early March) stages are two crucial phenology phases for distinguishing rubber plantations from natural forests. Secondly, an accurate forest map was generated based on PALSAR data with a strong ability to penetrate cloud layers in tropical regions. The PALSAR forest map defined a forest distribution map and reduced the misclassification between rubber plantations and other land cover types (such as tea gardens and shrubs). Since rubber plantation growth was strictly constrained by air temperature, a suitable region map for growing rubber plantations was generated to refine the mountainous rubber plantation distribution by combining MODIS land surface temperature (LST) data. This method took advantages of PALSAR, Landsat, and MODIS LST data effectively, and the data could complement each other. It made full use of archived remote sensing data and improved the mapping results accuracy of rubber plantations. The algorithm provides a novel method for integrating multi-source remote sensing data. Finally, this algorithm was applied to Xishuangbanna and got the area and distribution extent of rubber plantations in 2010. According to the confusion matrix, the overall accuracy of the resultant map is 91%, the user’s and product’s accuracy are 91% and 94%, respectively. The study shows that combining multi-source remote sensing data and phenology features of rubber plantations is very effective for distinguishing rubber plantations from natural forests based on a PALSAR-based forest map. This study provide a simple, effective and robust method for mapping rubber plantations. Its application prospect is very broad in tropical edge where is a nontraditional place for planting rubber plantations.(2) This study presented a novel model for predicting stand ages of rubber plantations based on multi-temporal Landsat time series. Firstly, based on abundant field survey samples, this study analyzed the characteristics of rubber plantations with different stand ages on Normalized Difference Vegetation Index (NDVI), Enhanced vegetation index (EVI), and Land Surface Water Index (LSWI) in time series. LSWIdefoliation<0 is an effective threshold for predicting stand ages of deciduous rubber plantations in tropical border. Secondly, based on LSWIdefoiiation<0, a stand ages predicting model of rubber plantations was presented based on annual Landsat datasets. Thirdly, this model was applied to construct stand ages of rubber plantations in Xishuangbanna in 2010. Furthermore, this study analyzed the relationship between rubber plantation area and topography based on digital elevation model (DEM). The study result shows that:(1) It is an effective way to predict stand ages of rubber plantations by using LSWIdefoliation<0.According to confusion matrix, the overall accuracy of rubber plantation stand age map is 85%; for ≤5,6-10, and ≥11 years old rubber plantations, their user accuracy were 87%,81%, and 90%, respectively, and their producer accuracy were 87%, 81%, and 90%, respectively. (2) Stand ages of rubber plantations distributed gradually from low to high elevation, from plain to mountain regions.≥11 and 6-10 years old rubber plantations mainly distributed in plain with low elevations (600-900m) and (600-1000m), while≤5 years old rubber plantations mainly distributed in mountain regions with high elevation (900-1200m). The novel stand age prediction model provides accurate and timely basic scientific data for rubber plantation monitoring and regional LULCC research.(3) Spatio-temporal change dynamics model and their crucial drivers of rubber plantations. Firstly, this study built a spatio-temporal database for rubber plantations monitoring based on rubber plantation maps (1990,2000, and 2010), DEM, and social economy data (population, rubber price, and land use policies). The rubber plantation maps of 1990,2000, and 2010 were generated through forest base map and their phonology characteristics. Based on theories of LULCC and the database, this study constructed a model for monitoring rubber plantations. According to this model, the rubber plantation area of Xishuangbanna has been raised to 3.5×105 hm2 in 2010 (3.94 times of that in 1990). During 1990-2000, the area change of rubber plantations is small, and most occurs in the state-owned rubber plantation garden, mainly driven by regional land use policies. During 2000-2010, rubber plantations have experienced a rapid increase, areas of private enterprises and small household rubber plantations have increased substantially, driven by natural rubber prices. Rubber plantation have been expanded into unsuitable regions including high-elevation, steep, and northern regions, which caused by the increasing sharp contradictions between protection policies of farmland and demands of rubber production.In summary, the integration of GIS, multi-source remote sensing, and phenology characteristics is an effective way for monitoring rubber plantations. Based on multi-temporal Landsat datasets, LSWIdefoliation<0 is an effective metric to rapidly retrieve the stand ages of rubber plantations. This study built the monitoring model from the view of spatio-temporal process and land use theories, which is simple and effective for understanding the spatio-temporal changes of rubber plantations. This study provides reliable decision-making basis for scientific management and plans in rubber plantation planting and regional land use.