A study of the spatiotemporal dynamic of land cover types and the driving forces of grassland area change in Gannan Prefecture and Northwest Sichuan based on CCI-LC data

Abstract

Land cover change is key information when evaluating ecosystem health and such data is of great significance to planning sustainable development. Remote sensing is an effective way to achieve rapid and dynamic monitoring of land cover changes on a large scale. In this study, field measurement data and Landsat 8 OLI images were used to evaluate the classification accuracy of ESA Global land cover products (CCI-LC) in Gannan Prefecture and Northwest Sichuan, using the CCI-LC data to analyze the spatiotemporal dynamic of land cover types from 1992 to 2015, and the factors driving changes in grassland area changes was studied, considering both natural and artificial factors. The results indicated that: 1) The overall accuracy of CCI-LC products in Gannan Prefecture and Northwest Sichuan is 76.79%. Among the six land cover types (grassland, forest, bare land, water, permanent snow and ice, artificial surfaces), the user accuracy of grassland is the highest (94.88%). 2) From 1992 to 2015, the grassland area showed a generally decreasing trend, with a marked change during the period 1994-2001, with an overall decrease in area of 6760.53 km2 and a rate of loss of -0.53%•year-1. 3) From 1992 to 2000, the main changes in land cover types in the study area involved grassland, forest and artificial surfaces, and grassland area was mainly reduced in this period; from 2000 to 2010, the changes in land cover type in the study area were more diverse. The grassland cover was again reduced in most instances, but at a greatly reduced rate compared to 1992-2000; from 2010 to 2015, the transformation of land types was more complex, but a common pattern was transfer of grassland to forest and artificial surfaces. 4) Agricultural GDP and animal husbandry GDP are the main driving factors of grassland area change in Gannan Prefecture and Northwest Sichuan, and both of them are negatively correlated with grassland area. This study provides a macroscopic, rapid and low-cost remote sensing monitoring method to analyze the dynamic changes in land cover types over a long time period.