Changes of Chinese forest-grassland ecotone in geographical scope and landscape structure from 1990 to 2020

Forest-grassland ecotone (FGE) has essential ecological and economic value. Unfortunately, it is impacted greatly by environmental changes and anthropogenic disturbance, and is considered one of the most severely threatened biomes in China. To protect Chinese FGE, identifying its exact boundary and exploring its landscape structure dynamic are badly needed, especially on nationwide scale at one-year temporal resolution. Here, we mapped the annual FGE distribution of China from 1990 to 2020, investigated its changing trends of area, location and landscape patterns, and revealed the underlying driving factors. Our results showed that FGE area over the 31 years totaled 1,011,870 km2, covering about 10.54% of China’s land. The FGE area first increased from 1990 and peaked in 1999, and then kept decreasing until 2020. The FGE gravity center has moved accumulatively 590.15 km over the 31 years, with the net moving distance of 228.76 km southwestward. From 1990 to 2020, forest area increased continuously while grassland and cropland area decreased, but these three landscape types had been dominating the FGE. The increase in forest area was largely converted from grassland. The decline in grassland mainly resulted from its conversion into cropland and forest. Meanwhile, the conversion of cropland to grassland supplemented grassland loss to a certain extent. At landscape level, the total area with decreased fragmentation is larger than that with increased fragmentation. Returning Farmland to Grassland Project and land reclamation were primary drivers for changes of fragmentation in the northern and middle part of the FGE, while temperature and precipitation were primary drivers in southern part. Our results will improve the understanding into the dynamic trends of distribution and pattern of FGE at nationwide scale, and thus help to optimize the designing of ecological projects and protective schemes for FGE as a unique and integral biome.

Methods
We combined spatial autocorrelation and spatial clustering methods to obtain Chinese forest-grassland ecotone(FGE) distribution. We mapped the FGE boundary from 1990 to 2020 in the following six steps: i) We generated a grid map with a cell size of 1.5km×1.5km in the study region and calculated the percentage of forest and grassland in each grid cell based on land cover map using ArcGIS10.7; ii) Using local spatial autocorrelation in the GeoDa V1.6.7, we produced two maps of spatial clustering, i.e., the clustering results of forest and grassland that separately include five circumstances including High-High (HH), Low-Low (LL), High-Low (HL), Low-High (LH), and Not Significant (NS), where “High” and “Low” before and after the short dash represents the forest/grassland coverage in the examined grid and its surrounding ones respectively, and “Not Significant” represents other circumstances than above-mentioned; iii) According to the paired forest and grassland clustering results, we obtained four types of landscapes including forest-grassland ecotone (FGE), forest zone (FZ), grassland zone (GZ), and the other zones (OZ); iv) We extracted FGE distribution of every year from 1990 to 2020 (hereafter referred as the original FGE), then merged the 31 FGE results to obtain an integrated FGE map, where the FGE covers an largest concentrated contiguous patch and many small dissociative patches; v) We obtain final FGE distribution map, neglecting the small patches that together occupied less than 2% of the total FGE area; vi) We obtained final annual FGE distribution by extracting the part of the original annual FGE within the final FGE distribution.