基于神经网络的遥感图像解译预测岩溶区地下水富集性——以广西桂林市东北部岩溶区为例

Prediction of groundwater enrichment feature in the karst area based on remote sensing information:To take the northeast area of Guilin as an study example

岩溶区地质条件复杂,地下水利用较为困难,研究岩溶区地下水的分布特征对于缓解水资源压力具有现实意义。遥感图像数据分析通过高空遥感图像分析获取地下水信息,具有可大面积观测、时效性强、耗时较短、经济节约等优点。文章以Landsat 8卫星数据信息为基础,从时空方面优选了广西桂林市东北部全州县至灌阳县一带碳酸盐岩分布区的遥感数据,并利用神经网络机器学习(Pytorch框架)和Python系统软件联合分析技术对研究区裂隙溶洞水遥感数据进行学习和预测,从而获得预测区地下水的分布特征,准确率接近96%,获得了良好的地下水预测效果。

Due to the complex geological condition in karst areas, there is facing a great difficulty for the groundwater utilization. Therefore, the study on the groundwater distribution feature in the karst areas is of practical significance to weaken and to ease water resource pressure. Analysis of remote sensing image can obtain groundwater information from remote sensing data at a high altitude. There is a large observation area, timeliness, less time consumed and a lower cost. Based on Landsat8 satellite information, remote sensing data of the carbonate area from Quanzhou County to Guanyang County in Northeast Guilin City are selected out in both temporal and spatial aspects. To use neural network machine learning program(Pytorch framework), Python software system and analytical technique, remote sensing data about fissure water at the karst cave in the research area are learned and forecasted, thus the groundwater distribution feature are obtained in the forecast area. The method shows a good forecasting performance and an accuracy of up to 96%.