Nonlinear thresholds in agricultural landscape fragmentation: A GWRF-SHAP diagnostic framework for differentiated rural land management
5 affiliations
- 1College of Landscape Architecture and Art Design, Hunan Agricultural University, Changsha 410128, China
- 2School of Public Administration, China University of Geosciences, Wuhan 430074, China
- 3College of Public Management, Guizhou University of Finance and Economics, Guiyang 550025, China
- 4Hunan Institute of Agricultural Economics and Information, Changsha 410125, China
- 5College of Landscape Architecture and Art Design, Hunan Agricultural University, Changsha 410128, China; Institute for Territorial Spatial Governance and Green Development of the Yangtze River Basin, Wuhan 430074, China; Corresponding author at: Hunan Agricultural University, No. 1 Nongda Road, Furong District, Changsha City, Hunan Province, China.
Abstract
Agricultural landscape fragmentation (ALF) is a critical ecological constraint on rural sustainability, but conventional linear models fail to capture the nonlinear thresholds governing its response to rural development. This study proposes a GWRF-SHAP diagnostic framework to identify spatially explicit nonlinear driver-response relationships and operational thresholds along rurality gradients. Using Hunan Province as the study area, we classified six agricultural landscape types and stratified five rurality levels ranging from extreme rural to extreme non-rural using a three-dimensional rurality index. The GWRF model quantified local nonlinear associations while SHAP values interpreted factor contributions and threshold inflection points. Results demonstrated semi-natural production landscapes exhibit a unimodal fragmentation pattern peaking in transitional rurality zones, whereas constructed landscapes consistently maintain a relatively stable level of fragmentation. The proportion of forest land (Pfl), nighttime light index (NTL), topographic wetness index (TWI), plot distance to town center (Pdtc), and plot distance to road (Pdr) emerged as the dominant drivers of spatial variation in ALF. Together, these factors explained the observed fragmentation patterns and exhibited pronounced nonlinear threshold effects. Specifically, Pfl exhibited threshold values ranging from 0.48 to 0.68. Below this range, scattered forest patches embedded within farmland intensified edge fragmentation, whereas above the threshold, forests became the dominant landscape matrix and farmland was transformed into embedded patches, resulting in reduced fragmentation of production landscapes but increased fragmentation of natural landscapes. NTL showed threshold values between 0.45 and 1.02, reflecting distinct fragmentation processes under different stages of urbanization. Thresholds associated with Pdtc (2.49-6.28km) further distinguished human-dominated from terrain-dominated fragmentation processes in natural landscapes and separated compact urban cores from dispersed peripheral construction landscapes. These findings demonstrate that agricultural landscape fragmentation is governed by nonlinear threshold responses to rural development and provide a scientific basis for differentiated agricultural landscape governance and sustainable rural planning.
Related
- Mestizos and Mennonites Facing the Water Crisis in Northern Mexico : Differentiated Trajectories of the Social Construction of Water Scarcity and Unequal Agricultural Adaptation Strategies (Laguna de Bustillos, Cuauhtémoc, Chihuahua, Mexico)
- Editorial for the Second EJP SOIL Special Issue: “Climate‐Smart Sustainable Agricultural Soil Management for the Future II”
- Volatile organic compound emissions in croplands: The contributions of crops, soil, and agricultural management practices
- Land fragmentation, tenure insecurity, and irrigation water management in Egypt: a hybrid review and sensitivity analysis
- Soil erosion and land-use management influence bioavailable selenium in a highland agricultural landscape in Kenya.
BibTeX
@article{li2026nonlinear,
title = {{Nonlinear thresholds in agricultural landscape fragmentation: A GWRF-SHAP diagnostic framework for differentiated rural land management}},
author = {Liwen Li and Cong Xia and Qianjun Wei and Yijie Li and Penglai Ran and Jie Tan and Jingfeng Xiao and De Yu},
journal = {Ecological Indicators},
year = {2026},
month = oct,
doi = {10.1016/j.ecolind.2026.115483},
url = {https://doi.org/10.1016/j.ecolind.2026.115483},
note = {CC BY},
}
RIS
TY - JOUR
TI - Nonlinear thresholds in agricultural landscape fragmentation: A GWRF-SHAP diagnostic framework for differentiated rural land management
AU - Liwen Li
AU - Cong Xia
AU - Qianjun Wei
AU - Yijie Li
AU - Penglai Ran
AU - Jie Tan
AU - Jingfeng Xiao
AU - De Yu
PY - 2026
DA - 2026/10/23
JO - Ecological Indicators
DO - 10.1016/j.ecolind.2026.115483
UR - https://doi.org/10.1016/j.ecolind.2026.115483
AB - Agricultural landscape fragmentation (ALF) is a critical ecological constraint on rural sustainability, but conventional linear models fail to capture the nonlinear thresholds governing its response to rural development. This study proposes a GWRF-SHAP diagnostic framework to identify spatially explicit nonlinear driver-response relationships and operational thresholds along rurality gradients. Using Hunan Province as the study area, we classified six agricultural landscape types and stratified five rurality levels ranging from extreme rural to extreme non-rural using a three-dimensional rurality index. The GWRF model quantified local nonlinear associations while SHAP values interpreted factor contributions and threshold inflection points. Results demonstrated semi-natural production landscapes exhibit a unimodal fragmentation pattern peaking in transitional rurality zones, whereas constructed landscapes consistently maintain a relatively stable level of fragmentation. The proportion of forest land (Pfl), nighttime light index (NTL), topographic wetness index (TWI), plot distance to town center (Pdtc), and plot distance to road (Pdr) emerged as the dominant drivers of spatial variation in ALF. Together, these factors explained the observed fragmentation patterns and exhibited pronounced nonlinear threshold effects. Specifically, Pfl exhibited threshold values ranging from 0.48 to 0.68. Below this range, scattered forest patches embedded within farmland intensified edge fragmentation, whereas above the threshold, forests became the dominant landscape matrix and farmland was transformed into embedded patches, resulting in reduced fragmentation of production landscapes but increased fragmentation of natural landscapes. NTL showed threshold values between 0.45 and 1.02, reflecting distinct fragmentation processes under different stages of urbanization. Thresholds associated with Pdtc (2.49-6.28km) further distinguished human-dominated from terrain-dominated fragmentation processes in natural landscapes and separated compact urban cores from dispersed peripheral construction landscapes. These findings demonstrate that agricultural landscape fragmentation is governed by nonlinear threshold responses to rural development and provide a scientific basis for differentiated agricultural landscape governance and sustainable rural planning.
SN - 1470-160X
SN - 1872-7034
ER -