Page 244 - AJWEP-22-6

Qin, et al.

temperature (a biphasic hump-shaped pattern with spatially explicit modeling approaches, to capture
a threshold at 0°C) and precipitation (diminishing potential heterogeneity and provide more targeted
marginal effects with thresholds at 300 mm and 450 mm). insights for ecological management and policy-making.
These patterns align with our observed 1–2-month This study relies on MODIS NDVI (250 m resolution)
precipitation lag and topographic constraints, resulting and 1 km climate data, which may overlook micro-scale
in local anti-persistence (Hurst index of 0.44) in desert fluctuations. Future research should integrate Landsat
and mountainous areas. In the Yellow River Basin of imagery and field measurements to better assess the
Henan Province, Shi et al. reported a substantial impacts of high-temperature anomalies and combine
51
FVC enhancement (67.7%) from 1990 to 2020, with socioeconomic data to explore policy implications.
elevation as the dominant driver (q = 0.3) and strong Moreover, clustering approaches based on vegetation
coupling with precipitation, consistent with our results types or climatic zones could be applied to identify
on elevation-induced spatial heterogeneity and climate- regions with similar FVC dynamics, 54,55 estimate the
topography interactions. Liu et al. observed FVC relative contributions of climatic and anthropogenic
52
increases across 90.64% of the Jinghe River Basin drivers, and improve forecasting accuracy through
area from 1998 to 2019, with significant correlations regression or hybrid models. Such methods have
to the precipitation concentration index (18.47%), been successfully applied in other scientific domains,
supporting our identification of precipitation’s notable demonstrating their potential to deepen the analysis of
role in vegetation recovery. Furthermore, Tian et al. influencing factors in ecological studies.
53
in the Yangtze River Delta described post-2016 FVC Shache and Bachu counties need to protect their water
fluctuations (initial decline followed by growth), resources to sustain grassland recovery, while Zepu and
predominantly driven by climate variables with spatial Tumushuke counties should promote the cultivation of
modulation by elevation and slope, which validates drought-tolerant species. Yecheng and Shache counties
our post-2000 increase, post-2010 stabilization, and should optimize irrigation management, and Zepu and
regional persistence based on the Hurst index (0.81). Bachu counties should implement afforestation on
These congruences affirm that in arid regions, FVC bare land. Climate adaptation strategies, such as oasis
improvements are primarily climate-mediated, with expansion, are critical, and future monitoring should focus
topographic factors exacerbating spatial variations and on temperature thresholds and precipitation fluctuation.
thresholds governing nonlinear responses. In view of the potential uncertainties associated
The short-term forecasting results based on with the application of large-scale remote sensing
regression models indicate that FVC in the study area data, future research will strengthen the deployment
will remain relatively stable, with a slight upward trend, of ground-based sampling sites and integrate them
during 2024–2028, resulting in an overall increase of with multidimensional remote sensing information to
approximately 0.01 over the next 5 years. A closer look establish a more refined data calibration framework.
at the beginning of the forecast period reveals that in Moreover, the potential errors of the methods used in
2024, the linear and quadratic models predict FVC this study should be acknowledged. For example, the
values of 0.232 and 0.227, respectively, which increase accuracy of the pixel dichotomy approach depends
slightly in 2025 to 0.235 and 0.229, respectively. These on the selection of NDVI thresholds, which may
narrow and consistent confidence intervals highlight lead to underestimation or overestimation of FVC in
the robustness of the forecasts, suggesting that no heterogeneous landscapes. In contrast, Hurst analysis
abrupt changes in vegetation are expected in the near assumes time-series stationarity and may be sensitive
term. Such stability aligns with the relatively steady to noise and short-term fluctuations, potentially biasing
patterns observed over the past two decades. It should persistence estimates. Similarly, the Sen+Mann–
be emphasized that these forecasts are derived from a Kendall tests, though widely applied, have been
univariate annual time series model and thus primarily critically assessed by Serinaldi et al., who highlighted
56
represent an extrapolation of historical trends rather their mathematical inconsistencies and limitations
than a mechanistic explanation of vegetation dynamics. under certain conditions, such as sample size and serial
Consequently, the results are best interpreted as short- correlation. Future research may consider improved
term, forward-looking references. Future studies procedures, such as the Hamed correction, to better
57
could expand on this work by incorporating additional account for long-term persistence. These limitations
environmental or ecological drivers, such as climate should be taken into account when interpreting the
variability or land-use changes, or by adopting more results. Simultaneously, machine learning methods can

Volume 22 Issue 6 (2025) 238 doi: 10.36922/AJWEP025350269