The China–Mongolia railway is the core foundation for the construction of traffic connection in the China–Mongolia–Russia economic corridor. Long-term desertification has brought significant ecological risks to the railway area. Owing to the large variety of vegetation cover in this region, desertification information is easily confused with other weak vegetation cover information. This article proposes a refined desertification information extraction method based on multisource feature spaces and geographical zoning modeling. First, based on the geographical zoning, land cover, and vegetation coverage data for Mongolia, the railway area is divided into the Central provinces and their northern region, the Eastern Mongolian Plateau, and the Southern Gobi region. According to the vegetation coverage characteristics and the applicability of various feature space models to different geographical regions, Albedo–normalized difference vegetation index, Albedo–modified soil adjusted vegetation index, and Albedo–topsoil grain size index feature space models were constructed for three geographical regions. Faced with new challenges presented by global warming and the impact of monsoons on the classification and extraction of desertification information, we established a desertification classification system with six levels (severe desertification, high desertification, medium desertification, low desertification, withered grassland, and nondesertification) and complete desertification information extraction. The results show that the overall classification accuracy of the method selected in this article is 85.21%. We further analyzed the mechanism of this method, compared it with previous studies, and thereby proved that this method is feasible to extract the fine information of desertification in large areas and complex geographical environments.

中文翻译：

---

多源特征空间和地理分区建模支持的中蒙铁路沿线荒漠化信息提取

中蒙铁路是构建中蒙俄经济走廊交通连接的核心基础。长期的荒漠化给铁路地区带来了重大的生态风险。由于该地区植被覆盖种类繁多，荒漠化信息很容易与其他弱植被覆盖信息混淆。本文提出了一种基于多源特征空间和地理分区建模的精细化荒漠化信息提取方法。首先，根据蒙古国的地理分区、土地覆盖和植被覆盖数据，将铁路区域划分为中部省份及其北部地区、蒙古高原东部和南部戈壁地区。根据植被覆盖特征和各种特征空间模型对不同地理区域的适用性，分别构建了三个地理区域的Albedo-归一化差异植被指数、Albedo-改良土壤调整植被指数和Albedo-表土粒度指数特征空间模型。 . 面对全球变暖带来的新挑战和季风对荒漠化信息分类提取的影响，我们建立了六级荒漠化分类体系（严重荒漠化、高度荒漠化、中荒漠化、低荒漠化、枯萎草原和非荒漠化）并完成荒漠化信息提取。结果表明，本文所选方法的总体分类准确率为85。21%。我们进一步分析了该方法的机理，并将其与前人的研究进行了比较，从而证明了该方法对于提取大面积、复杂地理环境中的荒漠化精细信息是可行的。