关键词: 植被物候, 季前气候因子, 黄土高原, 时滞效应, XGBoost-SHAP

Abstract: Aims The Loess Plateau is a typical ecologically fragile region in China. To clarify the response of vegetation spring phenology to preseason climatic factors, Methods this study used MODIS13 Q1 EVI data from 2003 to 2022.Vegetation phenology was extracted using the Savitzky–Golay (S-G) filter and the dynamic threshold method. Trend analysis, partial correlation analysis, and the XGBoost-SHAP model were applied to investigate the spatiotemporal changes in spring phenology and their responses to preseason climatic factors (diurnal temperature range, temperature, precipitation, and solar radiation). Important findings The results showed that: (1) The start of the growing season (SOS) on the Loess Plateau mainly occurred between 90 and 140 days, with a gradual delay from forest to grassland zones. In addition, 68.43% of the study area exhibited an advancing trend in SOS. (2) Preseason climatic factors exerted significant lagged effects on spring phenology. SOS was significantly influenced by the lagged effects of DTR and temperature from January to March, whereas precipitation and solar radiation showed longer lagged effects, with significant impacts from April to June. The lag windows varied among vegetation types (DTR and temperature predominantly affected SOS from January to April, while the lag effects of precipitation and solar radiation extended up to four to six months in some regions). (3) Spring phenology in forest areas was primarily regulated by temperature, with mean temperature being the dominant factor. Shrub areas showed significant responses to DTR and solar radiation, reflecting the synergistic effects of temperature variation and energy input. Grassland areas were highly sensitive to solar radiation and precipitation, indicating that water availability is the key factor controlling the onset of spring growth in water-limited environments.

Key words: spring phenology, preseason climatic factors, Loess Plateau, lag effects, XGBoost-SHAP