生态环境学报 ›› 2023, Vol. 32 ›› Issue (6): 1133-1139.DOI: 10.16258/j.cnki.1674-5906.2023.06.015

• 研究论文 • 上一篇    下一篇

下辽河平原土地沙漠化程度及预测研究

周玉祥1,*(), 赵玉1, 聂仁东1, 丁丁1, 郭立华2, 周佳峥1   

  1. 1.辽宁工程技术大学矿业学院，辽宁 阜新 123000
    2.华硕智慧科技（北京）有限公司技术部，北京 102200
  • 收稿日期:2023-02-26 出版日期:2023-06-18 发布日期:2023-09-01
  • 通讯作者: *
  • 作者简介:周玉祥（1979年生），男，讲师，博士，硕士研究生导师，主要从事矿山环境生态恢复与治理等科研工作。E-mail: 46103178@qq.com

Characterization and Prediction of Land Desertification in the Lower Liaohe River Plain

ZHOU Yuxiang1,*(), ZHAO Yu1, NIE Rendong1, DING Ding1, GUO Lihua2, ZHOU Jiazheng1   

  1. 1. College of Mining, Liaoning Technical University, Fuxin 123000, P. R. China
    2. Technology Department of ASUS Intelligent Technology (Beijing) Co., LTD, Beijing 102200, P. R. China
  • Received:2023-02-26 Online:2023-06-18 Published:2023-09-01

RichHTML

PDF

摘要：

近年来中国北方春季沙尘暴频繁发生，土地沙漠化地质灾害问题日趋严重，已对当地的经济发展和人民生命财产的安全构成威胁，研究土地沙漠化影响程度及发展趋势显得尤为重要。以下辽河平原作为研究对象，结合研究土地区沙漠化的规模大小和空间分布情况，将涉及的新民、辽中、台安、盘山和黑山等5个县采取多源信息复合方式进行实地调查，总结出该区域土地沙漠化类型有沙地、固定沙丘、半固定沙丘、流动沙丘等，确定了“土壤质地（C1）”、“植被状况（C2）”、“沙丘类型（C3）”和“裸沙占地百分比（C4）”等4个自然评价指标，将土地沙漠化程度级别划分为轻度（<50）、中度（50—70）和重度（>70）；结合本地区土地沙漠化程度的实际情况进行定性定量评价，利用传统的层次分析法（AHP）确定指标权重，通过层次分析法（AHP）分别与随机森林模型（RF）和支持向量机模型（SVM）结合运算，对下辽河平原土地沙漠化程度及发展趋势进行了分类预测。经过对模型进行参数调优和对数据集优化，最终RF模型的AUC值达到了0.89，优于SVM模型的AUC值（0.73），表明随机森林模型更适用于研究区沙漠化的分类预测。采用7:3的比例对数据集进行划分，以38组数据为基础数据，结合“土壤质地”、“植被状况”、“沙丘类型”和“裸沙占地百分比”等4个指标构建的AHP-RF模型。结果表明，植被状况（C2）、裸沙占地百分比（C4）为本地区土地沙漠化的主控指标，改善植被状况的治理措施是重中之重。研究结果可以为当地土地沙漠化防治工作提供有效借鉴意义。

关键词: 土地沙漠化, 层次分析法, 随机森林, 支持向量机, 分类预测, 防治措施

Abstract:

Northern China has recently experienced frequent spring sandstorms, which have contributed to deterioration of the natural environment and pose a severe threat to local economic development and human safety. This study aimed to characterize the scale and spatial distribution of desertification in Northern China. Liaohe Plain, representative of a typical arid and semi-arid area of Northern China, was adopted as a study area. Desertification in Xinmin, Liaozhong, Tai ‘an, Panshan, and Heishan was characterized using multiple forms of data. Following which desertification in the region was categorized as sandy, fixed, semi-fixed, and mobile dunes. This study analyzed land desertification in the lower Liaohe River Plain through the application of the “soil characteristics” (C1), “vegetation status” (C2), “sand dune type” (C3), and “percentage bare sand” (C4) indicators. Quantitative and qualitative methods were applied to characterize desertification in the region. Desertification of the study area was assessed according to categories of low (<50), medium (50-70), and high (>70) land desertification and the weights of the indicators were determined using conventional analytical hierarchy process (AHP). The levels of land desertification in the lower Liaohe River Plain were classified and predicted by combining AHP with random forest (RF) and support vector machine (SVM) models, with basin units used as the evaluation units. After optimization of model parameters and datasets, the area under the curve (AUC) of the RF model reached 0.89, exceeding that of the SVM model at 0.73. This result indicated that the RF model was suitable for classification and prediction of land desertification in the study area. The datasets were divided according to a ratio of 7:3 and an AHP-RF model was developed based on the four indicators, with 38 groups of data used as fundamental model input data. The results demonstrated that “vegetation status” C2 and “percentage bare sand” C4 were the dominant indicators of land desertification in the study area. Measures to improve vegetation conditions should be prioritized. This study can act as a reference to guide the prevention and regulation of land desertification in the study area.

Key words: land desertification, analytical hierarchy process, random forest, support vector machine, classification and prediction, prevention and control

中图分类号: