Two Measures, One Story: Multi-Scale Assessment Reveals Ecological Recovery in the Qinghai Lake Basin

Between 1990 and 2020, the Qinghai Lake Basin, a closed alpine watershed of roughly 29,661 square kilometers on the northeastern Qinghai-Tibet Plateau, underwent a measurable improvement in ecological security, according to a new multi-scale assessment that combined landscape ecological risk indexing with habitat quality modeling. The mean landscape ecological risk index across the basin fell by 31.25 percent, from 0.016 in 1990 to 0.011 in 2020, while mean habitat quality rose by 9.83 percent, from 0.58 to 0.64. Both indicators pointed to the decade between 2000 and 2010 as the pivotal transition, when the basin’s risk landscape restructured from a mosaic of extreme levels into a broadly homogeneous, low-risk pattern.

The study was designed to resolve a genuine contradiction in the literature. Earlier assessments based on the landscape ecological risk index had classified Qinghai Lake itself as a high-service, low-risk area, while evaluations using ecosystem service increments had labeled the same basin high-service and high-risk. Rather than treating the two measures as points on a single causal chain, the researchers used them as complementary lenses: the risk index captures external stressors through the interaction of landscape pattern and process, whereas habitat quality reflects the ecosystem’s intrinsic resistance and resilience. Where the two indicators converge, confidence in the assessment strengthens; where they diverge, the mismatch itself becomes diagnostic of scale-dependent mechanisms.

Methodologically, the framework distinguished two dimensions of scale. The grain dimension concerned the resolution at which landscape metrics were computed and the size of the units into which results were aggregated. Building on prior scale-optimization work identifying 90 to 100 meters as the optimal grain for the basin, the team resampled all raster data to 100 meters and tessellated the watershed into 7,686 grid cells of 2 by 2 kilometers, a unit size chosen because it falls between 2.91 and 3.50 times the mean patch area, within the recommended range of two to five times. The extent dimension spanned two levels of analysis: the natural watershed as a whole and its four constituent counties, Tianjun, Gangcha, Haiyan, and Gonghe.

The landscape ecological risk index itself was assembled from Fragstats-computed metrics of fragmentation, separation, and dominance, weighted 0.5, 0.3, and 0.2 respectively, and combined with a landscape vulnerability coefficient in which grassland ranked most vulnerable and construction land least. Habitat quality was modeled with the InVEST model, treating land cover types as proxies for habitat and incorporating threat factors including distances to roads, towns, and settlements, construction land, desert, population density, and GDP per capita. Classification thresholds for both indicators were fixed using the 1990 baseline with the Natural Breaks method and applied unchanged to 2000, 2010, and 2020, so that observed shifts reflect actual landscape dynamics rather than shifting criteria.

The temporal results showed that the basin’s risk structure changed fundamentally. In 1990 and 2000, the lowest and highest risk levels together covered more than 40 percent of the basin, a striking coexistence of extremes. By 2010 the distribution had converged on the medium level, which covered 47.90 percent of the area, and by 2020 the lower risk category alone dominated 72.88 percent of the basin, with the lowest category adding another 26.75 percent. Hotspot analysis using Getis-Ord Gi* statistics confirmed the trend: high-risk clusters collapsed from 25.31 percent of the basin in 1990 to just 0.54 percent in 2020, while the most significant hotspots shrank from 13.44 percent to 0.34 percent.

County-level trajectories diverged sharply. Tianjun County, the highest-elevation headwater area with the most fragile baseline, recorded a 29.41 percent risk reduction yet retained the highest absolute risk throughout the period, before transforming dramatically so that lower-risk land covered 95.11 percent of its area by 2020. Gangcha County posted the steepest decline, 37.50 percent, while Gonghe County fell 35.71 percent and ranked among the lowest-risk counties from 2010 onward. Haiyan County followed a non-linear path, showing little change before 2010 and then dropping to 0.009 by 2020. Gravity-center analysis echoed this heterogeneity: the basin’s risk center leapt 11.26 kilometers northwest between 2000 and 2010, driven largely by long-distance shifts in Tianjun and Gangcha, whereas Gonghe and Haiyan exhibited small, oscillatory movements of under about 1.2 kilometers.

Habitat quality told a parallel but not identical story. Basin-wide mean habitat quality rose steadily, with the combined higher and high levels holding stable at 54 to 56 percent of the area and the lowest-quality category shrinking from 23.01 percent to 12.46 percent by 2010. Tianjun again drove the change, converting large areas from low to medium quality, with its lower-level share falling from 39.23 percent to 17.65 percent and its own mean quality rising 24.44 percent, the largest gain among the counties, though it still ended 2020 with the lowest absolute value at 0.56. Gonghe consistently held the highest quality, exceeding 0.74, while Gangcha showed high structural stability with more than 86 percent of its area in the upper three quality levels. Across the basin, 84.59 percent of land remained stable in quality class over the full period, and the improving area, 12.99 percent, was more than five times the degrading area of 2.42 percent.

The authors attribute the basin-wide improvement to the combined effects of policy interventions and a warming, humidifying climate, while stressing that policy likely amplified a climate-driven recovery trend rather than acting alone. The Qinghai Lake Basin Ecological Protection and Comprehensive Management Project, launched in 2008, covered all four counties with a total investment of 1.57 billion CNY, 92.8 percent of it allocated to wetland protection and environmental restoration. By 2016, grassland coverage in the basin had risen from 55 percent to 68 percent. Meteorological records show annual precipitation increasing by 9.26 millimeters per decade, with Tianjun recording the largest gains, up to 15.3 millimeters per decade in summer rainfall, and phenology data indicate an advanced and lengthened growing season. The temporal correspondence between the 2000 to 2010 transition and these interventions is consistent with a substantial policy contribution, though the study does not claim formal causal proof.

From these patterns the researchers propose, explicitly as working hypotheses rather than validated causal pathways, two mechanisms. The first is a cascading response from landscape pattern to ecological function to risk outcome, in which structural changes detected by the risk index precede functional changes in habitat quality; in Tianjun, habitat quality improved during 2000 to 2010 before the risk index declined and high-risk areas disappeared during 2010 to 2020. The second is a sequence of quality preceding risk reduction in counties with fragile baselines undergoing dramatic recovery. The authors caution that with only four time points formal causality tests are statistically unreliable, that both indicators derive from the same land use data, and that the risk index is a composite measure of accumulated structural risk, not an external precursor, so the observed temporal sequence should be read as correlation in time.

The practical payoff is a two-dimensional decision matrix that translates indicator combinations into management rules. Where low risk coincides with high quality, as in Gonghe, the prescription is maintenance-oriented protection; where high risk coincides with low quality, as in parts of Gangcha’s northern headwaters, intensive restoration is indicated. Divergent combinations call for diagnostic investigation or proactive stress reduction, with conservation goals prioritizing habitat quality and prevention goals prioritizing the risk index. The framework also identifies a proposed preventive window, the period when habitat quality is improving while the risk index is not rising significantly, during which early intervention could shift management from post-event remediation toward proactive risk detection. The authors note that reliable numerical thresholds and monitoring intervals require long-term local calibration, and that future work should employ higher-frequency monitoring, change-point detection, and formal causal inference to test the proposed mechanisms. As a methodological case for alpine inland watersheds, the study illustrates how pairing two imperfect indicators across multiple scales can reconcile contradictory assessments and support place-specific governance.

Subject of Research: Multi-scale assessment of landscape ecological risk and habitat quality in the Qinghai Lake Basin from 1990 to 2020

Article Title: Management implications from multi-scale linkages between landscape ecological risk and habitat quality in the Qinghai Lake Basin

Article References: Han, Y., Zhang, C., Chen, K., Yang, L., Ma, J., Yang, X., Yang, H., Ma, Y., & Cao, M. (2026). Management implications from multi-scale linkages between landscape ecological risk and habitat quality in the Qinghai Lake Basin. Environmental and Sustainability Indicators, 32, Article 101583. https://doi.org/10.1016/j.indic.2026.101583

Image Credits: AI Generated

DOI: 10.1016/j.indic.2026.101583

Keywords: Qinghai Lake Basin, landscape ecological risk, habitat quality, InVEST model, Qinghai-Tibet Plateau, ecological security, scale effects, grassland restoration, Getis-Ord Gi*, multi-scale analysis, alpine watershed, ecological management