A long-term satellite investigation across subtropical Southwest China has revealed that the distinctive hydrogeological conditions of carbonate karst landscapes substantially amplify the sensitivity of vegetation phenology to climate variability compared to adjacent non-karst terrains. Vegetation phenology, which tracks the timing of recurring seasonal biological events including spring budburst, leaf expansion, autumn senescence, and growing season duration, functions as a direct indicator of terrestrial ecosystem responses to ongoing global environmental change. While regional warming has broadly shifted plant growth cycles across the Northern Hemisphere, researchers have increasingly sought to determine whether underlying geomorphological and lithological structures alter these relationships. By integrating multi-decadal satellite vegetation index observations with high-resolution hydroclimatic datasets and geological boundary maps between 2001 and 2022, the study provides regional empirical evidence that plants rooted in karst substrates experience heightened interannual phenological fluctuations, marked by amplified green-up advances and pronounced sensitivity during climatically unfavorable years.
The study focused on Southwest China, home to one of the largest contiguous karst expanses in the world, covering approximately 550,000 square kilometers across Guizhou, Guangxi, Hunan, Hubei, Sichuan, Chongqing, Yunnan, and Guangdong. Karst terrain in China spans roughly 3.44 million square kilometers, representing 15.6 percent of the estimated 22 million square kilometers of karst landscapes globally. These unique landscapes are underlain by soluble carbonate bedrock that exhibits exceptionally weak pedogenic capacity, leading to the development of shallow, discontinuous soil mantles interspersed with widespread bedrock outcrops. Furthermore, intensive chemical weathering and dissolution processes have established a complex dual surface-subsurface hydrological architecture consisting of sinkholes, fissures, and subterranean conduit networks. This structural arrangement induces rapid surface water percolation and drainage into deep aquifers, where surface runoff typically accounts for less than five percent of total precipitation. Consequently, shallow soils suffer from limited water retention and rapid moisture depletion, leaving vegetative communities routinely exposed to severe drought stress despite receiving substantial monsoon rainfall.
Understanding how geological foundations shape plant life cycles has emerged as a fundamental ecological question, as phenological shifts govern terrestrial carbon sequestration, evapotranspiration rates, surface energy partitioning, and overarching ecosystem productivity. Although atmospheric drivers such as temperature, precipitation, and solar radiation are universally acknowledged as primary controllers of plant development, their ecological efficacy depends on the capacity of the Earth critical zone to store and supply moisture to rooting systems. In non-karst terrains with deeper, continuous soils and greater moisture-holding capacities, the hydrological buffer moderates the physiological impact of transient meteorological dry spells. In contrast, the researchers recognized that the hydrologic vulnerability inherent to carbonate rock formations could disrupt conventional climate-phenology coupling. Specifically, while rising spring temperatures stimulate biochemical reactivation and vegetative green-up, elevated evaporative demand may rapidly exhaust sparse moisture reserves within shallow epikarst fissures, potentially generating asymmetrical biological responses that remained unquantified across regional scales.
To rigorously contrast phenological behavior between contrasting landforms, the research team assembled a continuous twenty-two-year remote sensing archive spanning from 2001 through 2022. They extracted 16-day composite Moderate Resolution Imaging Spectroradiometer Enhanced Vegetation Index data from the Terra satellite MOD13Q1 product via the Google Earth Engine cloud computing platform, resampling observations from an initial 250-meter grid to a standardized 500-meter spatial resolution. Cloud contamination and snow-covered pixels were filtered using summary quality assurance flags, followed by linear interpolation to reconstruct residual observational gaps. To isolate undisturbed natural vegetation signals from direct anthropogenic land-use conversions, the researchers cross-referenced multi-temporal land cover layers from the thirty-meter Global Land Cover product across six benchmark intervals. Croplands, built-up surfaces, bare areas, and artificial structures were systematically excluded, restricting the final analytical domain strictly to temporally stable natural woody vegetation types, including evergreen broadleaf, deciduous broadleaf, needleleaf, mixed forests, shrublands, and grasslands.
Seasonal phenological markers were derived from the smoothed vegetation index time series using the Savitzky-Golay filtering algorithm implemented within the TIMESAT software framework. The investigators determined the start of the growing season and the end of the growing season using dynamic threshold values set at twenty percent and fifty percent of the annual seasonal vegetation index amplitude, respectively, with the total length of the growing season defined as the intervening temporal duration. Applying relative amplitude thresholds effectively mitigated baseline greenness discrepancies across highly heterogeneous landscape covers, ensuring reliable seasonal cycle detection even within evergreen broadleaf canopies that exhibit muted intra-annual vegetation index amplitudes. Ground-truth validation using historical in situ phenological observations from nine regional monitoring stations in the Chinese Ecosystem Research Network confirmed strong directional and temporal consistency with the satellite-derived parameters, verifying the robustness of the remote sensing methodology across diverse topographies.
Statistical evaluation of regional temporal trajectories over the 2001 to 2022 window confirmed that both karst and non-karst terrains experienced a progressive advancing of spring green-up, a slight postponement of autumn senescence, and a net lengthening of the annual active growing period. However, the rates of change diverged substantially between the two geomorphological settings. The start of the growing season advanced at a faster rate in karst areas, shifting by negative 0.293 days per year, compared to negative 0.166 days per year across non-karst areas. Concurrently, the overall length of the growing season extended by 0.315 days per year within karst areas, reaching statistical significance, whereas non-karst areas exhibited a more moderate, non-significant elongation rate of 0.212 days per year. Autumn senescence trends remained statistically weak and muted in both environments, recording marginal postponements of 0.024 days per year in karst and 0.050 days per year in non-karst systems, establishing that growing season expansion was primarily driven by accelerated spring emergence.
To evaluate sensitivity to interannual climatic fluctuations, the authors calculated annual phenological anomalies relative to baseline means and modeled their relationship using reduced major axis regression. This bivariate analytical approach, which accounts for measurement error in both variables, demonstrated substantial phenological amplification in karst vegetation. Regression slopes reached 1.280 for the start of the growing season and 1.256 for the end of the growing season, demonstrating a 28.0 percent and 25.6 percent amplification of interannual anomaly amplitudes in karst ecosystems relative to non-karst counterparts. The overall length of the growing season similarly displayed an amplified slope of 1.164, representing a 16.4 percent heightened responsiveness. Independent validation using the standard MODIS land surface phenology product confirmed an amplification slope exceeding one, verifying that carbonate substrates consistently magnify seasonal vegetation variability under fluctuating annual weather conditions.
Crucially, the research revealed a pronounced directional asymmetry in how karst vegetation registered these anomalies. When grouping observation years into positive and negative anomaly phases, the researchers found that karst amplification was heavily concentrated in climatically unfavorable years rather than favorable seasons. During years characterized by delayed spring emergence, the average absolute start-of-season anomaly in karst areas reached 4.27 days, significantly exceeding the 2.90 days observed across non-karst areas. Conversely, during years featuring advanced green-up, the anomaly magnitudes were nearly identical between landforms, measuring 2.95 days in karst and 2.89 days in non-karst. Similarly, during years marked by shortened growing seasons, karst terrains sustained an average anomaly of 4.51 days versus 3.71 days in non-karst zones, whereas differences during prolonged growing years remained negligible. This asymmetry indicates that thin karst regoliths offer minimal buffering capacity when moisture or thermal stresses arise, provoking disproportionate biological distress.
Spatial trend mappings using Theil-Sen median slopes and non-parametric Mann-Kendall tests illustrated striking geographical divergence across Southwest China. Pixels exhibiting advancing spring green-up occupied 59.09 percent of karst areas, with 10.04 percent reaching statistical significance, compared to 54.70 percent and 5.71 percent in non-karst regions. These significant spring advances formed distinct spatial clusters in priority rocky desertification management districts spanning southern Guizhou, northwestern Guangxi, and southeastern Yunnan. In non-karst landscapes, significant spring advances were largely restricted to the western Sichuan gorges and the Hengduan Mountains. Likewise, pixels with lengthened growing seasons covered 53.05 percent of karst territory, with 7.23 percent exhibiting significant prolongation, surpassing non-karst proportions of 51.69 percent and 5.74 percent. This pronounced spatial clustering highlights the combined influence of ambient climate change and two decades of targeted ecological engineering aimed at restoring degraded carbonate landscapes.
To untangle the specific climatic mechanisms governing phenological changes, the researchers deployed pixel-level Random Forest machine learning models interpreted through Shapley Additive Explanations. The analysis established that phenological responses were highly directional rather than governed by a uniform driver. For the start of the growing season, drivers promoting earlier green-up accounted for 69.8 percent of karst pixels compared to 58.0 percent in non-karst areas, with negative temperature effects dominating 28.6 percent of karst terrain, followed by negative Palmer Drought Severity Index effects across 16.0 percent. Non-karst pixels displayed higher shares of delayed-green-up classes linked to positive moisture variables. For overall season length, positive-effect classes prevailed across 58.8 percent of karst pixels, where positive temperature, precipitation, and soil moisture effects formed spatial clusters across desertification control belts. This pattern underscores that maintaining vegetative growth and extending annual productivity in karst terrains remains exceptionally dependent on adequate water availability.
Finally, the authors examined causal regulatory hierarchies by coupling Spearman rank correlations with partial least squares path modeling. Direct bivariate correlations between geological background indicators and phenological timing were negligible, yet geology exhibited strong negative correlations with mean annual soil moisture and the Palmer Drought Severity Index. In the structural equation models, which exhibited good predictive fit with goodness-of-fit metrics ranging from 0.558 to 0.575, geographical position dictated broader hydrothermal gradients, while geological background operated indirectly by suppressing moisture availability and exacerbating drought stress. Thermal conditions and solar radiation exerted direct positive influences on season lengthening and autumn timing, but soil water deficits persistently constrained vegetative vigor. The authors emphasized several research limitations, including the application of uniform dynamic extraction thresholds across distinct plant functional types, the 500-meter sensor resolution relative to fine-scale karst microtopography, unmodeled interactions between warmth and moisture, and the complex challenge of isolating climatic signals from intensive regional ecological afforestation initiatives.
Subject of Research: Hydrogeological modulation of vegetation phenological responses to climate variability
Article Title: Hydrogeological constraints amplify phenological responses of karst vegetation to climate variability relative to non-karst vegetation
Article References: Zhang, J.-X., Liu, F., Yuan, D.-M., Jia, S.-B.-W., Cai, J., Yu, L.-F., Liu, Y., Chen, Z.-F., & Yan, L.-B. (2026). Hydrogeological constraints amplify phenological responses of karst vegetation to climate variability relative to non-karst vegetation. Landscape Ecology, 41(10), Article 165. https://doi.org/10.1007/s10980-026-02453-5
Image Credits: AI Generated
DOI: 10.1007/s10980-026-02453-5
Keywords: vegetation phenology, karst hydrogeology, Southwest China, start of growing season, end of growing season, length of growing season, Enhanced Vegetation Index, MODIS, climate variability, drought stress, soil moisture, carbonate bedrock, rocky desertification, Random Forest SHAP, partial least squares path modelling, Timesat, Theil-Sen estimator

