Estimation Of Vegetation Structure Parameters In Semi-arid Mine Dump With Remote Sensing Imagery

Vegetation structure parameters are important indicators of ecological resilience and biodiversity,but their traditional measurement methods are slowinspeed,low in efficiency,high in cost and small in coverage,while remote sensing technology has advantages such as all-weather,wide range,fast speed and short cycle,which is broadly used instead of traditional methods.However,the vegetation structure reconstructed in open pit mine is the result of both artificial and natural effects,which has the characteristics of complex structure,various types and strong heterogeneity,so it brings difficulties to the quantitative estimation of vegetation structure parameters.In view of this,the thesis builds the characteristic factor database and estimation model of the vegetation structure parameters of the open pit mine dump and applies it to the four dumps of the Heidaigou based on the World View2 image and field investigation.The results and conclusions are as follows:(1)The thesis develops a method for surveying and analyzing data of vegetation structure parameters in semi-arid dumping sites.The results show that the ground digital photo ingestion and photo processing method can effectively extract the vegetation coverage information;the vegetation structure type simplification method and use the existing empirical model to calculate the biomass on the basis of simplification,which can greatly reduce the traditional harvesting method.The heavy workload is brought about,and the biomass basic data can be used to determine the spatial structure parameters and reduce the repeated investigation.The vegetation structure survey method formed in this study can provide reference and reference for other vegetation surveys.(2)Through the comparative analysis of typical vegetation community spectrum,vegetation index and texture characteristics of the dumping site,it is found that the spectral difference of typical vegetation is most obvious in the Red Edge,Nir1 and Nir2 bands,especially the Nir1 and Nir2 bands;The main difference of vegetation index is in RVI;The difference of texture features is affected by the dual factors of the band and texture calculation window size.The overall difference is that the texture difference in the near-infrared band is larger than that in the visible band.In the visible range,the difference in texture parameters increases with the increase of the calculation window.In the near-infrared range,the overall difference is that the smaller calculation window is larger than the larger calculation window,especially the difference between 3×3,7×7 and 9×9 is most significant.(3)The random forest feature importance ranking method was used to determine the factors that have important explanatory for vegetation structure parameters.The results show that the characteristic factors with high contribution to vegetation coverage were NDVI,SR and Red,mainly vegetation index information;The characteristic factors with higher interpretation of ground biomass were Red Edge,mean and Nir,mainly spectrum and mean information;Analysis of spatial structure parameter factor found: texture information has a high contribution to spatial structure parameters,according to different calculation window size,the main related texture information is correlation,information entropy and variance,etc.,and texture information Mainly calculated by the Red band.Experiments show that the feature selection method based on random forest can provide model input data for the estimation model.(4)The RF and MEA-BP neural network model can be used to estimate the vegetation structure parameters of semi-arid mine dumps.The optimal MEA-BP model displays a fine accuracy,with 2 reaching 0.863.The RMSE is 0.795.The ground biomass 2 is 0.9115,with the RMSE reaching 1.4805kg/m2;The spatial structure parameter's 2 is 0.624,with the RMSE reaching 0.0906.This indicates that the MEA-BP model has higher accuracy level and reliability in the estimation of structure parameters in semi-arid mining areas.(5)The vegetation restoration analysis of the semi-arid mine dump shows that there is a certain correlation between the vegetation coverage and the ground biomass,but the correlation between spatial structure index and vegetation coverage is weak.The overall coverage,biomass and spatial structure of the area are superior to the original site.After reconstruction,the coverage of different vegetation configuration types does not increase with the recovery time.For example,the coverage of the two modes of arbor-shrub and arbor-shrub-herb is less than the short recovery time;most of the types of ground biomass increase with recovery time;For the spatial structure parameters,except for the simple arbor and herb structure,others increase with time,indicating that the stability of vegetation structure is increasing year by year.The highest vegetation structure parameters in the dumping site are the mixed pattern of arbor and shrub,indicating that this mixed mode is conducive to improving the richness of vegetation structure,thereby improving resilience and diversity.