Abstract:

Desertification has become a major threat to the global ecological environment, and the desertification monitoring is crucial for desertification prevention and control. Based on the Suomi/NPP (National Polar-orbiting Partnership) remote sensing data and the observation data of 8 meteorological stations during the vegetation growing season (from May to September) from 2014 to 2021 in the Qaidam Basin, the desertification difference index (DDI) was calculated by using NDVI-Albedo (Normalized Difference Vegetation Index-Albedo) feature space. Moreover, the natural discontinuity method, Sen+M-K trend analysis method, correlation analysis method, accuracy error matrix and transfer matrix analysis were also used to explore the spatial and temporal dynamic evolution of land desertification and the influence of meteorological factors to desertification in the Qaidam Basin from 2014 to 2020 during the vegetation growing season. The results are as follows: (1) The NDVI-Albedo feature space performs a high applicability in the Qaidam Basin (R² greater than or equal to 0.65), with an overall classification accuracy of 79.38% and a Kappa coefficient of 0.62. (2) From 2014 to 2021, the degree of land desertification in the eastern and southern Qaidam Basin is lower than that in the western and central Qaidam Basin. Furthermore, DDI shows a significantly increase in some areas, especially in southern and eastern region with the increase rate of DDI over 0.01a^-1. The total area of desertification land in the Qaidam Basin shows a decreasing trend with a rate of -1 173 km²·a^-1. Additionally, a transforming characteristic occurs between different degrees desertification land that severe desertification lands transferred to mild desertification land. (3) Correlation analysis shows that precipitation and average relative humidity are significantly positively correlated with DDI (P<0.01), and correlation coefficients are 0.91 and 0.86, respectively, indicating that the water is the dominant factor affecting desertification in the Qaidam Basin.

Key words: desertification, NDVI-Albedo feature space, desertification difference index （DDI）, spatio-temporal dynamic monitoring, meteorological affecting factor, the Qaidam Basin

摘要：

荒漠化目前已成为威胁全球生态环境的主要问题，开展荒漠化监测对于荒漠化防治至关重要。基于2014—2021年植被生长季（5—9月）Suomi/NPP（National Polar-orbiting Partnership）遥感数据和柴达木盆地8个气象站点观测数据，利用NDVI-Albedo（Normalized Difference Vegetation Index-Albedo）特征空间计算荒漠化差值指数（Desertification Difference Index，DDI），运用自然间断法、Sen+M-K趋势分析法、相关性分析法、精度误差矩阵计算和转移矩阵计算等方法，探讨柴达木盆地植被生长季荒漠化土地时空动态演变及气象影响因素。结果表明：（1）基于NDVI-Albedo特征空间构建的DDI在柴达木盆地荒漠化监测中适用性较高，特征方程R²≥0.65，整体分类精度79.38%，Kappa系数0.62。（2）2014—2021年，柴达木盆地荒漠化程度东部、南部较低而西部、中部较高，且东部、南部部分地区DDI值以每年超过0.01的速率增大，部分地区增大显著；荒漠化土地总面积呈减小趋势，速率为-1 173 km²·a^-1，不同程度荒漠化土地之间存在转移特征，具体表现为荒漠化程度较重的土地向较轻的土地转移。（3）相关性分析表明，降水量、平均相对湿度均与DDI呈极显著正相关（P<0.01），相关系数分别为0.91、0.86，水分是影响柴达木盆地荒漠化的主要气象因子。

关键词: 荒漠化, NDVI-Albedo特征空间, 荒漠化差值指数, 时空动态监测, 气象影响因子, 柴达木盆地