景观多样性指数（LDI）不仅是景观生态学研究中的一个重要指标，而且也是生物多样保护中的重要层次。基于土地利用栅格数据（30 m分辨率），在ArcMap环境中采用Neighborhood和Focal工具对宁夏沿黄绿洲LDI时空变化和尺度依赖特征进行了研究。结果表明：（1） 基本单元（正方形）边长90~6000 m 5个时期重复统计分析表明宁夏沿黄绿洲景观在空间上具有明显的尺度依赖特征，转折点在3000 m。（2） 近50 a来研究区LDI变化具有周期性，2000年是转折点。其中，1975—2000年LDI呈降低趋势，LDI分区分析显示以退化区斑块类型面积（CA）最大和好转区CA最小为主要特点，分别为6840 km²和1332 km²。2000—2020年LDI呈增加趋势，以稳定区CA最大和退化区CA最小为主要特点，分别为7848 km²和792 km²;由于此阶段初始LDI（2000年）最低，LDI后期好转程度没有达到前期水平。（3） LDI分级面积转换以前期好转区向后期稳定区流转（796 km²，占好转区60.5%）和退化区分别向稳定区（3519 km²，占退化区51.5%）、好转区（3036 km²，占退化区44.4%）流转为主要特征。（4） 景观多样性变化格局以CA与相对分裂度指数（RSI）呈负关联为特点，这种关系机制在不同时期和不同变化类型中具有普遍性。总的来看，在区域景观变化研究中确定分析指标在时空尺度上的变化转折点既是保障研究结果具有借鉴和共享的必要条件，也是区域景观多样性可视化表达与分析的基础。

The landscape diversity index (LDI) is not only an important indicator in landscape ecology research but also an important component in biodiversity conservation. Based on the land use/cover raster data (30-m resolution), the spatiotemporal variation and scale-dependence characteristics of the LDI of an oasis along the Yellow River in Ningxia, China, have been studied using Neighborhood and Focal tools in ArcMap from 1975 to 2000. The results are as follows: (1) The LDI, which is measured by a square with a side length from 90 m to 6000 m, had obvious spatial scale-dependent characteristics based on five times repeat, and its turning point was 3000 m. (2) The change trend of the LDI had been cyclical in the past decades, with a turning point of 2000. During the study period, the LDI exhibited a decreasing trend from 1975 to 2000, and the analysis of LDI zoning indicated that the main characteristics were as follows: the class area (CA) of the degraded area was the largest, and the CA of the improved area was the smallest, which were 6840 km² and 1332 km², respectively. In contrast, there was an increasing trend for the LDI from 2000 to 2020, mainly characterized by the maximum CA in the impervious area and the minimum CA in the degraded area, which were 7848 km² and 792 km², respectively. Because the initial LDI in 2000 was the lowest in the entire period, its improvement status in the later period did not reach that of the early period. (3) The conversions of the LDI-grading area were mainly characterized by the transfer from the early improved area to the late impervious area (796 km², 60.5% of the improved area) and the transfer from the degraded area to the impervious area (3519 km², 51.5% of the degraded area) and the improved area (3036 km², 44.4% of the degraded area), respectively. (4) The change in the landscape diversity pattern was characterized by a negative correlation between CA and relative splitting index, and this relationship mechanism was universal in different periods and change types. Thus, it is to best understand the landscape diversity change with credible spatiotemporal scales in a regional landscape study. It is necessary to ensure that the research results are not only used for reference and sharing but also used to visualize and analyze regional landscape diversity.