Evolution Characteristics and Simulation Prediction of Forest and Grass Landscape 1 Fragmentation Based on the “Grain for Green” Projects on the Loess Plateau, P.R. 2

: Forest fragmentation is one of the major environmental issues that the international community 9 is generally concerned about under the background of global climate change. Studying the impact and the 10 interaction mechanism of land use change processes on landscape fragmentation is important to gaining a 11 comprehensive understanding of the ecosystem response to human activities and global climate change. 12 Based on the implementation background for the “Grain for Green” Project, we selected the Loess Plateau 13 as the research area and used the coupled future land use simulation (FLUS) model and landscape 14 fragmentation model to explore the temporal and spatial changes in forest and grass landscape fragmentation. 15 The results showed that (1) Woodland, grassland, and cropland are the main landscape types, accounting for 16 about 90% of the total area. In addition, the area of cropland initially increased and then decreased, while the 17 area of woodland and grassland exhibited the opposite trend Oover the last 35 years. In particular, the period 18 from 2000 to 2015 was a forest and grass restoration stage, and the average annual rate of forest and grass 19 restoration reached 0.56%. (2) The FLUS model was used to predict the land use on the Loess Plateau in 20 2030. The kappa coefficient was 0.85, and the figure of merit coefficient (FOM) was 0.11 for a 1% random 21 sampling, which are within a reasonable range, and the simulation results are also consistent with the 22 objective change in the current social and economic development. (3) The fragmentation of woodland and 23 grassland were dominated by edge type and core type. The core type had a concentrated distribution and an 24 absolute advantage, accounting for more planning and objective evaluation of woodland and grassland spatial allocation and quality improvement, 28 and provide an important basis for the formulation of ecological protection and land management policies.


31
Forests play a vital role in the service functions of global ecosystems by providing a series of service 32 functions such as maintaining the biodiversity, soil and water conservation, global carbon and water cycles, 35 and as the main body of the terrestrial ecosystem, it accounts for 31% of the total land area (Bonan, 2008; 36 Keenan et al., 2015). However, with the rapid development of agriculture and cities, about 40% of the 37 woodland has been converted to land cover types such as cropland, pastures, and other artificial building 38 lands around the world (Achard and Hansen, 2012), and the problem of forest loss and fragmentation is 39 increasing (Riitters et al., 2000). Forest fragmentation refers to the process in which a large continuous forest 40 is divided into smaller, independent patches of forest (Riitters et al., 2002). At present, 70% of the forests on

43
Numerous studies have shown that forest fragmentation will lead to various negative effects, such as a 44 decrease in the regional biodiversity, increased soil erosion, increased risk of invasion by invasive species,  Riitters et al., 2000). Especially in developing countries, forest fragmentation has led to huge losses 47 in economic, environmental, and cultural benefits (Zuidema et al., 1996). In particular, in China, the growth 48 of forest area has attracted worldwide attention, but the problem of forest fragmentation has become 49 increasingly prominent, which has led to low forest quality and weakened ecological service capabilities (Wu   technology, large-scale data acquisition and dynamic monitoring capabilities can provide reliable, high-56 precision data (Zhao et al., 2020). The powerful spatial information processing and analysis capabilities of 57 the geographic information system (GIS) can be used to accurately evaluate and analyze forest resources 58 (Franklin, 2001), and the combination of GIS and RS can be used to analyze forest fragmentation and to 59 reveal the dynamic spatial evolution (Carranza et al., 2015). Currently, traditional forest fragmentation 60 studies have mostly used landscape pattern indexes to describe the composition and structural characteristics 61 of landscape types. For example, using the patch size, the total patch area, the patch size variation coefficient,

73
The Loess Plateau is one of the areas in the world with serious soil erosion and forest fragmentation 74 (Feng et al., 2016). The coexistence of drought, water shortages, and soil erosion is a bottleneck restricting

98
Project on the Loess Plateau has achieved initial results, but understanding the spatial distribution and 99 landscape fragmentation is very urgent, and it needs to be improved as the ecosystem is facing huge changes 100 (Yu et al., 2018). In particular, several studies have shown that the land use change caused by the 101 implementation of this project has made the ecosystem more fragmented (Zhang and Yin, 2019), but there is 102 no precedent for the combined study of the landscape fragmentation of forest and grassland ecosystems. In 103 particular, regarding the evolution parameters of forest fragmentation, there is still a lack of clear spatial 104 significance and a quantitative description at the regional scale. This is important for establishing regional 105 ecological corridors, improving biodiversity, controlling soil erosion, and enhancing the continuity of the 106 landscape.

107
In this study, we selected the Loess Plateau as the research area, and comprehensively analyzed the land

Data Source and Processing
The land use data for the Loess Plateau used in this study were obtained from the Geospatial Data Cloud

145
The land use classification was conducted using the classification system of the Chinese Academy of 146 Sciences, and the land use types were divided into six categories: cropland, woodland, grassland, water bodies, 147 built-up land, and unused land ( Figure 1). After preprocessing the data using the ENVI 5.1 platform, 148 including atmospheric corrections, geometric corrections, stitching, and cropping, the method of supervised 149 classification combined with human-computer interaction visual interpretation was adopted, and Google      woodland or grassland by 20% and 20%, respectively. scenario D: increase the probability of the transition from cropland to woodland or grassland by 30% and 30%, respectively. Scenario E: increase the probability 194 of the transition from cropland to woodland or grassland by 40% and 40%, respectively.

195
The ANNs was based on the cellular automata model, which is composed of an input layer, a hidden 196 layer, and an output layer, and each neuron corresponds to a variable in the CA (Openshaw, 1998). The 197 essence of the simulation process is to establish the spatial relationships between the driving factors and the 198 initial land types (Liu et al., 2008). The specific process is described by follows:

210
This process is essentially a loop iteration process, and it makes the output result continuously approach the 211 target value . In this study, we chose to run the iterative loop in a 9×9 Moore neighborhood, Where , is the comprehensive probability that the grid p changes from the initial land use type to 216 the land use type k at time t; Ω , is the probability of land use type k appearing in grid p; is the inertia coefficient of ground type k at time t; → is the conversion cost from land use type c to land use      (Table 4). The implementation of the "Grain for Green" Project greatly affected the land use structure.

272
In contrast, the land use change was dominated by the conversion of cropland into woodland and grassland

286
Under the five different implementation scenarios for the "Grain for Green" projects, the land use 287 structure predicted by the FLUS model continued to change in 2030, but it would still be focused on cropland, 288 woodland, and grassland. Around residential areas, built-up land will continue to consume the cropland, 289 grassland, and other land, but the trend will slow down. The woodland area will increase, mainly through the 290 conversion of the surrounding cropland; and the water area will remain stable over the next ten years ( Figure   291 3, Table 5).

292
The

300
The forest and grass fragmentation on the Loess Plateau is dominated by edge and core type. The core 301 type fragmentation has a concentrated distribution and is the largest, accounting for more than 75% of the 302 entire area of woodland and grassland. The percentage of perforated and patch type is very small, less than 5%. In different years, the areas of the various spatial process types transformed into each other. From a 304 spatial perspective, except for in Inner Mongolia, the patch type fragmentation was widely distributed in the 305 other provinces, and the forest and grass landscapes in the northwest and central areas of the Loess Plateau 306 were most severely fragmented, and it was mostly concentrated in cities and forests, cropland, and in the 307 transition zone between the city and the grassland. Moreover, the gravity center of the fragmentation has 308 basically not changed, but there has been an overall westward shift. The core type fragmentation was mainly 309 distributed in Shaanxi Province and in the Luliang Mountains and Taihang Mountains in Shanxi Province.

310
The perforated and edge type fragmentation were relatively scattered.

311
The fragmentation degree generally initially increased in intensity and then slowed down (Figure 4),

312
that is, the core area continued to increase, and the proportion of patch and edge areas continued to decrease.

313
The most severe fragmentation occurred around 2000, with the patch and edge areas accounting for the 314 highest proportions (1.30% and 23.38%, respectively) ( Table 6)

321
Among them, the percentage of the large core area of the returning cropland to woodland and grassland 322 (intensity of 30%) was the highest, reaching 70.72%, and the proportion of patch area was the lowest (1.11%).

397
The main feature of this change is that the core area has increased, which shows that the impact of human fragmentation is analyzed separately, the fragmentation will be more severe.

406
Decreasing forest fragmentation is a long-term and complex process (Riitters et al., 2020). The