The Fractional Vegetation Cover (FVC) and Associated Driving Factors of Modeling in Mining Areas
The Fractional Vegetation Cover (FVC) and Associated Driving Factors of Modeling in Mining Areas
复制标题
DOI：
10.14358/pers.21-00070r3
复制
发表时间：
2022-10
期刊：
影响因子：
--
通讯作者：
Jun Yu Li;Tianyu Guo;Chengye Zhang;Fei Yang;Xiao Sang
中科院分区：
文献类型：
--
作者：
Jun Yu Li;Tianyu Guo;Chengye Zhang;Fei Yang;Xiao Sang
为确定矿区植被覆盖度（FVC）及其建模驱动因子，以6类数据为驱动因子，采用多元线性回归（MLR）、地理加权回归（GWR）和地理加权人工神经网络（GWANN）3种方法进行建模。在锡林浩特市胜利矿区进行的实验表明，未考虑空间异质性和空间非平稳性的MLR模型表现最差，GWR模型表现出明显的区位差异，因为它预先定义了一个线性关系，无法描述某些位置的FVC。GWANN模型对这些缺陷进行了改进，是最适合矿区FVC掘进过程的模型，其均方根误差（RMSE）和平均绝对百分比误差（MAPE）分别达到0.16和0.20，优于其他两种模型。与MLR模型相比，它在RMSE和MAPE方面分别改进了约24%和33%，与GWR模型相比，这些值分别增长到59%和71%。
To determine the fractional vegetation cover (FVC ) and associated driving factors of modeling in mining areas, six types of data were used as driving factors and three methods—multi-linear regression (MLR ), geographically weighted regression (GWR ), and geographically weighted artificial neural network (GWANN )—were adopted in the modeling. The experiments, conducted in Shengli mining areas located in Xilinhot city, China, show that the MLR model without consideration of spatial heterogeneity and spatial non-stationarity performs the worst and that the GWR model presents obvious location differences, since it predefines a linear relationship which is unable to describe FVC for some locations. The GWANN model, improving on these defects, is the most suitable model for the FVC driving process in mining areas; it outperforms the other two models, with root-mean-square error (RMSE ) and mean absolute percentage error (MAPE ) reaching 0.16 and 0.20. It has improvements of approximately 24% in RMSE and 33% in MAPE compared to the MLR model, and those values grow to 59% and 71% when compared with the GWR model.