Abstract:

Based on a Bayesian ensemble change detection algorithm, this study integrates passive microwave data, MODIS products, and meteorological and hydrological variables to comprehensively analyze lake ice phenology and its driving factors for seven large lakes in the arid region of northwest China over the period 1978—2022. The results are summarized as follows: (1) Cross-validation between passive microwave remote sensing data and Landsat observations yields a mean coefficient of determination of 0.86, with a mean absolute error of 1.56 days and a root mean square error of 2.52 days. These results indicate that passive microwave data provide a feasible and reliable approach for extracting lake ice phenology, despite minor local discrepancies. (2) Across the seven lakes in the arid region of northwest China, the average onset of ice formation occurred between November 17 and January 19 of the following year, while complete ice-off dates ranged from July 11 to October 25. The mean ice-covered duration was 143 days. All lakes exhibited a shortening trend in ice duration, with Bosten Lake showing the smallest reduction rate (0.24 days per decade) and Sayram Lake the largest (0.55 days per decade). (3) Lake freeze-onset dates displayed both advancing and delaying trends, whereas ice-off dates predominantly advanced in the arid region of northwest China from 1978 to 2022. For most lakes, the onset and complete freeze dates were delayed at rates of 0.09-0.38 days per decade and 0.24-0.29 days per decade, respectively. In contrast, a few lakes, including Jili Lake and Bosten Lake, exhibited advancing freeze-onset trends. In high-latitude regions (excluding Ebinur Lake), ice-off dates advanced significantly at rates of 0.27-0.48 days per decade, whereas high-altitude lakes exhibited weaker trends ranging from 0.12 to 0.27 days per decade. (4) Lake ice freeze-thaw cycles in the arid region of northwest China are jointly controlled by meteorological factors (wind speed, snow cover, precipitation, and near-surface air temperature) and hydrological characteristics (lake area and water transparency). Near-surface air temperature directly governs the timing of lake ice freezing and melting, whereas lake area and transparency indirectly influence ice phenology by regulating water heat capacity and solar radiation absorption.

Key words: lake ice phenology, spatiotemporal differences, Bayesian ensemble change detection, influencing factors, arid region of northwest China