基于信息量模型的三江源东部草地退化易发性评价——以青海省果洛州为例

草地退化的预警对维护草地生态系统稳定及保障牧民财产安全意义重大。本文探索并验证了信息量模型跨领域应用于草地退化易发性评价的可行性,可为草地退化防治管控提供决策依据。以青海省三江源东部果洛藏族自治州为研究区,将像元二分模型和信息量模型耦合,并根据地形、气候、人类活动、水文、土壤5个维度,进行草地退化易发性评价。结果表明：①2008—2018年果洛藏族自治州草地退化区与草地退化易发性评价中退化高易发区分布特点一致,主要集中于玛沁县中部、玛多县南北部等地区;②处于超过4400 m的高程,高于25人/km²的人口密度,大于30°的坡度等位置的草地容易发生退化。而位于年平均降水量大于600 mm,年累计日照时数小于2400 h,依附土壤类型为灰褐土、沼泽土等地区的草地具有更好的抗逆性;③共有68.34%的退化草地分布于极易退化区和较易退化区,15.92%分布于中易发区,15.74%分布于不易退化区和低易退化区,评价模型精度较好,AUC（Area Under Curve）值达到0.764。由此可见,像元二分模型与信息量模型结合以评定草地退化易发性的方法是可靠的,该思路对其他地区草地退化易发性评价具有借鉴意义。

Grassland degradation susceptibility assessment of the eastern area of the Three Rivers Source region based on the information quantity model:A case study of Golog Tibetan Autonomous Prefecture,Qinghai Province

Early warning of grassland degradation is of great significance for maintaining the stability of grassland ecosystem and for ensuring the property security of herding households. This study explored and validated the feasibility of applying the information quantity model to assess grassland degradation susceptibility, which can provide some guidance for grassland degradation management. Taking Golog Tibetan Autonomous Prefecture in the eastern area of the Three Rivers Source region as the study area, this research coupled the dimidiate pixel model and the information quantity model to assess the grassland degradation susceptibility by taking into consideration topography, climate, human activities, hydrology, and soil types. The results show that: (1) The identified distribution characteristics of grassland degradation are consistent with the grassland degradation high-risk areas in the prefecture, mainly concentrated in the central part of Maqin County and the south and north of Maduo County. (2) Locations such as those with elevation higher than 4400 m, population density above 25 people/km², or slope greater than 30° can easily be susceptible to grassland degradation. In contrast, grasslands with average annual precipitation more than 600 mm, annual cumulative sunshine hours less than 2400 h, or soil types such as grey cinnamon soil and boggy soil have higher resilience. (3) A total of 68.34% of the degraded grasslands are distributed in areas extremely or highly prone to degradation, 15.92% are in areas moderately prone to degradation, and 15.74% are in areas unlikely or not prone to degradation and the area under the curve (AUC) value of the evaluation is 0.764. Therefore, it is reliable to combine the dimidiate pixel model with the information quantity model to evaluate grassland degradation susceptibility, and the method has reference values for the assessment of grassland degradation susceptibility in other areas.