A method for extracting water from barrier lake in high mountain areas based on decision tree classification:A case study of Attabad barrier lake on the Karakoram Highway
A method for extracting water from barrier lake in high mountain areas based on decision tree classification:A case study of Attabad barrier lake on the Karakoram Highway
堰塞湖水体动态监测对于堰塞湖的险情评估、灾害推演、安全管理以及降险处置决策等均具有重要意义.为了高效提取高山区堰塞湖真实水体范围,以中巴公路Attabad堰塞湖为研究区,利用决策树分类结合归一化差值水体指数(normalized difference water index,简称 NDWI)、综合水体指数(comprehensive water index,简称CWI)等6种常规水体提取方法来提取堰塞湖水体范围,并对比了 6种方法用于堰塞湖水体提取的效果,筛选出适用于高山区堰塞湖的最佳水体提取方法,最后使用混淆矩阵法进行了精度评价,并做了分类后处理,准确提取了堰塞湖水体边界.研究结果表明:(1)6种水体提取模型中CWI模型水体提取效果最好;(2)基于坡度的决策树分类方法总分类精度为89.31％,Kappa系数为0.84,较为完整地提取了高海拔堰塞湖真实水体范围,有效剔除了湖岸斜坡山体阴影,湖泊边界较为清晰完整.基于决策树的高山区堰塞湖水体提取方法在高海拔山区能较为有效地提取真实水体范围,尤其是针对地形切割强烈、山体阴影较多的堰塞湖区域,能快速准确识别水体.该方法的优点是:水体提取过程较为简单,容易实现,提取效率较高,便于推广.
Abstract
The water dynamics of barrier lakes in high mountain areas are crucial for risk assessment,disaster pre-diction,safety management,and decision-making.[Objective and Methods]To accurately and efficiently extract the water boundaries of mountainous barrier lakes,this paper focuses on the Attabad barrier lake along the Karako-ram Highway,proposing a water extraction method based on decision tree classification.This method incorporates slope information into six conventional water extraction methods for decision tree classification.The effectiveness of these six methods was compared for extracting water from barrier lakes in the experimental area.The best-perform-ing methods suitable for barrier lakes in high-altitude areas were applied to extract the water body range of the Atta-bad barrier lake.Accuracy was assessed using a confusion matrix,and classification post-processing was performed to refine the water boundary extraction.[Results]The research results indicate that(1)among the six models,the CWI model demonstrates the best performance,effectively distinguishing between slope,water,and shadow wa-ter,leading to a highly accurate outline of the barrier lake.However,a limitation of this model is the presence of a few mountain shadows in the middle of the slope.(2)The decision tree classification method based on slope a-chieved an overall accuracy of 89.31％and a kappa coefficient of 0.84.It effectively extracts the actual water range,excluding slope shoreline and mountain shadows,and provides a clearer lake boundary.Nevertheless,black fragments observed in the lower area of the barrier lake,likely due to landslides and mountain shadows,remained challenging to classify.Overall,the decision tree classification-based method proved effective in identifying water bodies,particularly in areas with rugged terrain and numerous shadows.[Conclusion]This paper proposes a method for extracting water bodies from barrier lakes in high mountain areas using decision tree classification.By incorporating slope information into conventional water body extraction methods,this approach accurately extracts the water boundary,effectively eliminates shadows from steep slopes,retain shadowed water on gentler slopes,and improves extraction efficiently.The simplicity and high extraction efficiency of this method make it a practical solu-tion for widespread application.
关键词
决策树分类模型/Attabad堰塞湖/水体指数/中巴公路/水体提取方法
Key words
decision tree classification model/Attabad barrier lake/water body index/Karakoram Highway/wa-ter extraction mothod
万方数据