Spatiotemporal Evolution Dataset Development of Habitat Quality and Multi-Scenario Simulation in Yuanjiang-Jinshajiang-Minjiang Dry Valleys (2000–2040)

GUO Yalin^1*  WANG Yucai¹  YAN Weipo²  WANG Qing¹  ZHANG Yu¹

1. School of Environment and Resource, Southwest University of Science and Technology, Mianyang 621010, China;

2. Sichuan Academy of Environmental Policy and Planning, Chengdu 610041, China

Abstract: The spatiotemporal evolution of habitat quality in dry valleys serves as an important indicator of regional ecological security. With the intensification of anthropogenic activities and rapid changes in land use patterns, the risk of habitat quality degradation in dry valleys is becoming increasingly prominent, posing severe challenges to regional ecological security. This study focuses on the Yuanjiang dry-hot valley, the Jinshajiang dry-warm valley (Shangri-La-Lijiang section), and the upper Minjiang dry-temperate valley. Based on land use data from 2000, 2010, and 2020, a coupled FLUS-InVEST model was constructed to analyze the spatiotemporal evolution of historical habitat quality. Three scenarios—natural development, farmland protection, and ecological restoration— were then used to simulate land use patterns and predict habitat quality in 2040. These analyses generated a dataset describing spatiotemporal evolution and multiscenario based simulations and predictions of habitat quality in the dry valleys of the Yuanjiang, Jinshajiang, and upper Minjiang Rivers (2000–2040). Model reliability was verified using the kappa coefficient and overall accuracy; the kappa coefficients for the three regions were 0.80, 0.81, and 0.87, respectively, and all overall accuracies exceeded 0.88, indicating that the model has good applicability in each region. The dataset comprises (1) study area boundary data; (2) simulated land use type data at 1 km resolution for the study areas in 2040 under the three scenarios; (3) spatial distribution data of habitat quality at 1 km resolution for the study areas in 2000, 2010, 2020, and 2040 (under the three simulated scenarios); and (4) statistical data of habitat quality for the study areas during different periods. This dataset systematically reflects the differentiated response patterns of habitat quality to land use changes across the three types of dry valleys and can provide fundamental data support for land use optimization, ecological risk identification, and zoning management in dry valley regions. The dataset is archived in .shp, .tif, .xls and .txt formats, and consists of 179 data files with data size of 967 KB (compressed into one file with 679 KB).

Keywords: dry valley; land-use change; habitat quality evolution; FLUS-InVEST model; multi-scenario simulation

DOI: https://doi.org/10.3974/geodp.2026.05.11

Dataset Availability Statement:

The dataset supporting this paper was published and is accessible through the Digital Journal of Global Change Data Repository at: https://doi.org/10.3974/geodb.2026.06.03.V1.

1 Introduction

Dry valleys represent a unique type of ecological landscape in the Hengduan Mountains and are characterized by low precipitation, intense evaporation, low vegetation coverage, and limited ecological recovery capacity, making them extremely sensitive to external disturbances^[1]. Based on climatic, vegetation, and soil characteristics, the dry valleys in the Hengduan Mountains can be classified into 3 types: dry-hot valleys, dry-warm valleys, and dry-temperate valleys^[2,3]. In recent years, under the combined effects of anthropogenic activities and natural factors, land use patterns in dry valley areas have undergone significant changes, and habitat degradation has accelerated in some regions, posing severe challenges to regional sustainable development and the performance of ecological functions. Habitat quality is an important indicator for measuring ecosystem health and biodiversity conservation. Its spatiotemporal evolution is essential for identifying regional ecological risks and formulating conservation strategies and has become a key concern in the fields of ecological protection and sustainable development^[4].

In view of this context, 3 typical dry valleys in the Hengduan Mountains—the Yuanjiang dry-hot valley, the upper Minjiang dry-temperate valley, and the Jinshajiang dry-warm valley (Shangri-La-Lijiang section)—were selected as the research objects (Figure 1). Based on land use data from 2000, 2010, and 2020, a coupled FLUS-InVEST model was constructed, and 3 scenarios—natural development, farmland protection, and ecological restoration—were established. A dataset of spatiotemporal evol­ution, multi-scenario simu­lation and prediction of habitat quality in the dry valleys of the Yuanjiang, Jinshajiang, and upper Minjiang Rivers (2000– 2040) was developed, elucida­ting the differentiated response patterns of habitat quality to land use changes in the three types of dry valleys. This dataset can provide fundamental data support for land use optim­ization, ecological risk identifi­cation, and zoning management in dry valley regions.

2 Metadata of the Dataset

The metadata of the Spatiotemporal evolution dataset of habitat quality and multi-scenario simulation in Yuanjiang-Jinshajiang-Minjiang dry valleys (2000–2040)^[5] is summarized in Table 1. It includes the dataset full name, short name, authors, year of the dataset, temporal resolution, spatial resolution, data format, data size, data files, data publisher, etc.

Table 1  Metadata summary of the Spatiotemporal evolution dataset of habitat quality and multi-scenario simulation in Yuanjiang-Jinshajiang-Minjiang dry valleys (2000–2040)

3 Methods

3.1 Data Sources

The basic data used in this study cover 3 major categories—land use, physical geography, and socioeconomic—with a total of 13 indicators (Table 2). The land use data were obtained from the National Cryosphere Desert Data Center[1] at a spatial resolution of 30 m and included 8 land use types: cropland, forestland, shrubland, grassland, water body, ice/snow, bare land, and impervious surface. Physical geography data included elevation, slope, aspect, mean annual temperature, annual precipitation, and the normalized difference vegetation index (NDVI). Elevation data were acquired from the Geospatial Data Cloud platform[2] at a 30 m spatial resolution; slope and aspect data were derived from the Digital Elevation Model (DEM) using GIS tools. The mean annual temperature, annual precipitation, and NDVI were sourced from the Earth Resource Data Cloud platform[3] at a spatial resolution of 1 km. Socioeconomic data includes the nighttime light index, GDP density, population density, road distribution, and distance to the nearest government seat. Specifically, the nighttime light index data were obtained from the Resource and Environment Science Data Platform[4]; GDP density and population density data were derived from the LandScan dataset[5], all at 1 km spatial resolution. Road data were freely downloaded from OpenStreetMap. The distance to the nearest government seat was characterized by the Euclidean distance, which is based on the government seat locations provided by the National Catalogue Service for Geographic Information[6]. To meet the operational requirements of the FLUS model, all data were uniformly resampled to a 1 km resolution and projected to the WGS_1984_UTM_46N coordinate system.

Table 2  Main data sources and descriptions

3.2 Algorithm

3.2.1 Land Use Simulation based on the FLUS Model

The future land use simulation (FLUS) model is a tool for multi-scenario simulation and prediction of spatiotemporal changes in land use types^[7]. The model incorporates an artificial neural network (ANN) algorithm to derive the suitability probability of each land use type from a single period of land use data and multiple driving factors. Moreover, it employs a self-adaptive inertia competition mechanism based on roulette wheel selection, which addresses the uncertainty and complexity of mutual conversions among multiple land use types by adjusting parameters such as the neighborhood factor, inertia coefficient, and conversion cost, thereby improving simulation accuracy^[8,9]. The simulation performance of the model largely depends on the rational selection of driving factors. Land use change results from the combined effects of physical, geographic, and socioeconomic factors. Therefore, by considering the regional context and referring to the relevant literature^[10], this study selected a total of 12 indicators from 2 categories—physical geography (elevation, slope, aspect, mean annual temperature, annual precipitation, NDVI, and distance to river) and socioeconomics (nighttime light index, GDP density, population density, distance to road, and distance to the nearest government seat)—as the driving factors (Table 2) for the FLUS model to simulate and predict future land use patterns.

Based on the policies and development goals of the study area and with reference to the relevant literature^[11], 3 scenarios—natural development, farmland protection, and ecological restoration—were established to simulate land use patterns for 2040.

(1) Natural development scenario: This scenario follows the historical trends of land use change in dry valleys from 2000 to 2020 without imposing additional policy constraints, reflecting the trajectory of natural evolution.

(2) Farmland protection scenario: In accordance with the national food security strategy and the requirements of the farmland conservation redline policy, the probability of farmland conversion is strictly controlled to strengthen the protection of cultivated land.

(3) Ecological restoration scenario: Aligned with national ecological restoration projects and green development strategies, this scenario focuses on promoting the expansion of ecological land such as grassland and forestland, restricting the growth of nonecological land, and achieving the comprehensive management goals of soil erosion and desertification control in dry valley regions.

3.2.2 Habitat Quality Assessment based on the InVEST Model

In this study, the habitat quality module of the InVEST model was used to assess the habitat quality of 3 typical dry valleys in the Hengduan Mountains. This module is based on land use data and considers human activities that cause habitat degradation as threat factors. It accounts for the maximum influence distance and weight of each threat factor, as well as the sensitivity of different land use types to each threat factor^[12]. Cropland and impervious surfaces are associated with intense anthropogenic activities, while bare land, which lacks vegetation cover and is prone to weathering and erosion, has a high risk of land degradation^[13]. Therefore, cropland, bare land, and impervious surfaces were selected as threat sources. The model-related parameter settings are shown in Tables 3 and 4.

Table 3  Threat factor attributes

Table 4  Sensitivity of land use types to habitat threat factors

In the habitat quality assessment of the InVEST model, the degree of habitat degradation serves as the basis for calculating habitat quality, and its expression is as follows:

                                                                               (1)

where D_xj denotes the degree of habitat degradation of grid cell x in land use type j, which ranges from 0 to 1; R is the number of threat factors; r represents the r-th threat factor; Y_r denotes the total number of grid cells of threat factor r; w_r represents the weight of threat factor r, ranging from 0 to 1; r_y denotes the stress intensity of threat factor r on grid cell y; i_rxy represents the influence of threat factor r from grid cell y to grid cell x, which may follow either linear or exponential decay; β_x denotes the accessibility or protection level of grid cell x; and S_jr represents the sensitivity of land use type j to threat factor r.

The habitat quality index is calculated as follows:

                                                                                (2)

where Q_xj denotes the degree of habitat quality of grid cell x in land use type j, which ranges from 0 to 1, with higher values indicating better habitat quality and greater suitability for species survival; H_j and D_xj represent the degree of habitat suitability and degree of habitat degradation, respectively, of land use type j; z is a normalization constant, usually set to 2.5; and k is the half-saturation constant used to adjust the distribution and variation range of habitat quality values. All variables above are dimensionless; however, the spatial distance used to calculate the influence decay (i_rxy) is measured in meters (m).

4 Data Results and Validation

4.1 Dataset Composition

The dataset comprises: (1) study area boundary data (.shp); (2) simulated land use type data at 1 km resolution for the study areas in 2040 under three scenarios (.tif); (3) spatial distribution data for habitat quality at 1 km resolution for the study areas in 2000, 2010, 2020, and 2040 (under the three simulated scenarios) (.tif); and (4) statistical data on habitat quality for the study areas during different periods (.xls).

4.2 Data Validation

To evaluate the applicability of the FLUS model in the study areas, this study used the kappa coefficient to verify the simulation accuracy of the FLUS model. The kappa coefficient ranges from 0 to 1, with larger values indicating higher simulation accuracy; a value greater than 0.70 is generally considered to represent relatively high accuracy^[7]. The specific validation procedure was as follows: based on the land use data of the study areas in 2000 and 2010 and the 12 driving factors, the FLUS model was used to simulate and predict the land use pattern in 2020 for each region. The simulation results were then compared with the actual land use data for 2020, and the kappa coefficient and overall accuracy (OA) were calculated. The validation results revealed that the kappa coefficients for the Yuanjiang dry-hot valley, the upper Minjiang dry-temperate valley, and the Jinshajiang dry-warm valley (Shangri-La-Lijiang section) were 0.80, 0.87, and 0.81, respectively, and the OA values were all greater than 0.88, indicating that the FLUS model has good applicability in each region and can be used for multi-scenario simulations of future land use patterns in 2040.

In addition, the simulated habitat quality results were highly consistent with the regional land use transition patterns: regions with more dramatic land use change exhibited corres­pondingly greater changes in habitat quality. Under the three scenarios, the variation ranges of habitat quality index relative to the 2020 baseline were 0.006–0.012 for the Yuanjiang dry-hot valley, 0.003–0.013 for the upper Minjiang dry-temperate valley, and 0.001 for the Jinshajiang dry-warm valley (Shangri-La-Lijiang section) (the index was 0.912 under all three scenarios), reflecting the differentiated response characteristics of different regions to land use policy interventions and supporting the rationality of the assessment results.

4.3 Data Results Analysis

4.3.1 Evolution of Land Use Types in Each Region

Forests were the dominant land use type in all three study areas: the Yuanjiang dry-hot valley, the upper Minjiang dry-temperate valley, and the Jinshajiang dry-warm valley (Shangri-La-Lijiang section). From 2000 to 2020, the common trends of land use transformation in the three regions were a decrease in grassland and an expansion of impervious surfaces, but the evolutionary pathways differed significantly (Table 5). In the Yuanjiang dry-hot valley, cropland, shrubland, and impervious surface areas increased by 945.20 km², 661.42 km², and 109.94 km², respectively, whereas forestland and grassland areas decreased by 1,020.96 km² and 697.42 km², respectively. The conversion was dominated by the transformation from forestland and grassland to cropland, with some cropland converted to impervious surfaces, reflecting the encroachment of agricultural production and urban development on ecological space. In the upper Minjiang dry-temperate valley, grassland and cropland decreased by 332.56 km² and 168.79 km², respectively, while forestland increased by 334.23 km². The conversion was dominated by the transformation from grassland and cropland to forestland, reflecting an internal structural adjustment of ecological land. In the Jinshajiang dry-warm valley (Shangri-La-Lijiang section), grassland decreased by 328.76 km², whereas cropland and forestland increased by 169.30 km² and 102.56 km², respectively, and impervious surfaces expanded notably (+4.25 km², +551.95%). The conversion was dominated by the transformation from grassland to cropland and forestland, while cropland continued to be converted to impervious surfaces, resulting in spatial expansion characteristics driven jointly by urbanization and agricultural development.

Table 5   Changes in land use types by region (2000–2020)

Note: /indicates no transition.

4.3.2 Multi-Scenario Simulation of Land Use Change

Under the three development scenarios of natural development, farmland protection, and ecological restoration, the responses of land use to policy interventions varied considerably across regions (Figure 2). The Yuanjiang dry-hot valley exhibited the most pronounced responses: under the natural development scenario, cropland (+770.32 km²) and forestland (+269.89 km²) expanded simultaneously; under the farmland protection scenario, cropland expansion (+1,264.67 km²) and grassland contraction (–666.50 km²) were particularly notable; and under the ecological restoration scenario, the increase in forestland area (+2,016.62 km²) was the greatest among the three regions, indicating a high sensitivity of

Figure 2  Spatial distribution maps of land use types under different simulation scenarios (2040)

land use to policy interventions in this area. The upper Minjiang dry-temperate valley responded relatively positively to ecological restoration policies, with notable effectiveness in bare land management (–347.08 km², –70.39%); however, under the natural development scenario, the trends of bare land expansion (+158.47 km²) and ice/snow shrinkage (–28.78 km²) warrant attention. In the Jinshajiang dry-warm valley (Shangri-La-Lijiang section), land use change was the most moderate under all three scenarios, but the expansion of forestland was limited (+0.98% under the natural development scenario, +2.19% under the farmland protection scenario, and only –0.02% under the ecological restoration scenario), indicating that because forest coverage is already at a high level (>78%), the potential for further ecological restoration gains is significantly constrained by baseline conditions. The differing response characteristics of land use change in the three types of dry valleys reflect the profound influence of regional natural background conditions and differences in the intensity of human activities on the effectiveness of policy interventions.

4.3.3 Spatiotemporal Variation Characteristics of Habitat Quality

An equal-interval classification method was adopted to divide the habitat quality index into 5 grades: low quality (0–0.2), relatively low quality (0.2–0.4), moderate quality (0.4–0.6), relatively high quality (0.6–0.8), and high quality (0.8–1). Overall, the habitat quality in the Yuanjiang dry-hot valley, the upper Minjiang dry-temperate valley, and the Jinshajiang dry-warm valley (Shangri-La-Lijiang section) exhibited pronounced temporal evolution and spatial differentiation characteristics (Figure 3).

From 2000 to 2020, the habitat quality of the Yuanjiang dry-hot valley tended to decrease, with relatively low-quality habitat increasing by 942.57 km² and moderate-quality habitat emerging (210.93 km²), whereas high-quality habitat decreased by 1,054.63 km². The habitat quality index decreased from 0.828±0.304 to 0.819 ± 0.310. In the upper Minjiang dry-temperate valley, the habitat quality index increased from 0.888 ± 0.176 to 0.891±0.177. Specifically, high-quality habitat remained largely stable (+17.18 km²), and relatively

Figure 3  Distribution maps of habitat quality classes under three simulation scenarios (2040)

low-quality habitat decreased by 169.33 km²; however, low-quality and moderate-quality habitat increased by 125.16 km² and 26.99 km², respectively, indicating an emerging trend of low-quality habitat expansion. In the Jinshajiang dry-warm valley (Shangri-La-Lijiang section), the area of high-quality habitat decreased by 167.82 km², while that of relatively low-quality habitat increased by 167.77 km², and the habitat quality index decreased from 0.919±0.196 to 0.913±0.210.

From 2020 to 2040, under the natural development scenario, habitat quality in all three regions tended to decrease. The farmland protection scenario had negative effects on habitat quality in the Yuanjiang dry-hot valley (the habitat quality index decreased to 0.807±0.319) and the Jinshajiang dry-warm valley (Shangri-La-Lijiang section) (habitat quality index decreased to 0.912±0.217) but had a slight positive effect in the upper Minjiang dry-temperate valley (habitat quality index increased to 0.894±0.171), reflecting regional variability in policy effects. Under the ecological restoration scenario, the upper Minjiang dry-temperate valley showed the most significant improvement in habitat quality (habitat quality index +0.013), followed by the Yuanjiang dry-hot valley (+0.012), whereas the Jinshajiang dry-warm valley (Shangri-La-Lijiang section) exhibited a slight decline (–0.001), indicating that the marginal benefits of ecological restoration are constrained by regional baseline conditions and that the space for ecological recovery in high-altitude dry-warm valleys is limited.

5 Discussion and Conclusion

The evolution of habitat quality in dry valleys has important indicative significance for regional ecological security. However, under the influence of different hydrothermal conditions, the responses of habitat quality to land use change vary considerably among different types of dry valleys, which introduces uncertainty into comparative studies across multiple types of dry valleys. The Spatiotemporal evolution dataset of habitat quality and multi-scenario simulation in Yuanjiang-Jinshajiang-Minjiang dry valleys (2000–2040) is based on a coupled FLUS-InVEST model driven by multisource data, and the model reliability was verified using the kappa coefficient and overall accuracy as dual indicators. The kappa coefficients of the three typical regions are 0.80, 0.81, and 0.87, respectively, and the overall accuracy values are all higher than 0.88, indicating credible data quality. The development of this dataset effectively fills the gap in comparative studies on the evolution patterns of habitat quality in different types of dry valleys and provides a reliable foundation for revealing the driving mechanisms underlying differences in habitat quality degradation under different hydrothermal combinations.

At present, this dataset has certain limitations. For example, the driving factors used in land use change simulations do not fully incorporate dynamic variables such as policy implementation intensity; the scenario settings are based on existing policy frameworks and involve idealized assumptions; and the parameter settings in the InVEST model, such as threat factor weights, maximum influence distances, and habitat suitability, are somewhat empirical. Future research can further improve simulation and assessment accuracy by introducing multisource remote sensing data, optimizing localized parameter calibration, and adopting multi-model integration methods, thereby providing more precise data support for ecological risk identification and zoning management in dry valley regions.

Author Contributions

Guo, Y. L. was in charge of overall design of the dataset development; Wang, Y. C. and Yan, W. P. were responsible for data collection and processing, and for designing the models and algorithms; Wang, Y. C., Guo, Y. L., Wang, Q. performed data validation; Zhang, Y. conducted data verification and validation; Guo, Y. L. and Wang, Y. C. drafted the data paper.

Conflicts of Interest

The authors declare no conflicts of interest.

References

[1]        Wang, Z. L., Hu, B., Bao, W. K., et al. Latitudinal patterns and underlying factors of component biomass in plant communities in the arid valley of southwest China [J]. Chinese Journal of Plant Ecology, 2022, 46(5): 539–551.

[2]        Zhang, R. Z. The Dry Valleys of the Hengduan Mountains Region [M]. Beijing: Science Press, 1992.

[3]        Fan, J. R., Yang, C., Bao, W. K., et al. Distribution scope and district statistical analysis of dry valleys in Southwest China [J]. Mountain Research, 2020, 38(2): 303–313.

[4]        Zhang, Y. Z., Hu, Y. F., Han, Y. Q., et al. Spatial distributions and evolutions of global major ecological degradation regions and research hotspot regions [J]. Acta Ecologica Sinica, 2021, 41(19): 7599–7613.

[5]        Guo, Y. L., Wang, Y. C., Yan, W. P., et al. Spatiotemporal evolution dataset of habitat quality and multi-scenario simulation in Yuanjiang-Jinshajiang-Minjiang dry valleys (2000–2040) [J/DB/OL]. Digital Journal of Global Change Data Repository, 2026. https://doi.org/10.3974/geodb.2026.06.03.V1.

[6]        GCdataPR Editorial Office. GCdataPR data sharing policy [OL]. https://doi.org/10.3974/dp.policy.2014.05 (Updated 2017).

[7]        Li, Y. X., Liu, Z. S., Li, S. J., et al. Multi-scenario simulation analysis of land use and carbon storage changes in Changchun City based on FLUS and InVEST model [J]. Land, 2022, 11(5): 647.

[8]        Lin, W. B., Sun, Y. M., Nijhuis, S., et al. Scenario-based flood risk assessment for urbanizing deltas using future land-use simulation (FLUS): Guangzhou Metropolitan Area as a case study [J]. Science of the Total Environment, 2020, 739: 139899.

[9]        Su, J. C., Tang, B., Liu, Y., et al. Multi-scenario land use change simulation and ecosystem service function evaluation in Bazhong City based on FLUS model [J]. Environmental Ecology, 2023, 5(11): 17–25.

[10]     Wang, L. Y., Zhang, Y. Evolution of spatiotemporal patterns and multiscenario simulation of land use transition in Gansu Province [J]. Arid Zone Research, 2025, 42(4): 695–707.

[11]     Zuo, Q., Zhou, Y., Wang, L., et al. Impacts of future land use changes on land use conflicts based on multiple scenarios in the central mountain region, China [J]. Ecological Indicators, 2022, 137: 108743.

[12]     Zhang, X., Li, Y. X., Lv, C. J., et al. Research progress on application of ecosystem service functions based on InVEST model [J]. Ecological Science, 2022, 41(1): 237–242.

[13]     Wang, B. X., Cheng, W. M. Effects of land use/cover on regional habitat quality under different geomorphic types based on InVEST model [J]. Remote Sensing, 2022, 14(5): 1279.