Analysis Of NPP Driving Forces In Typical Arid Areas Of Northwest China

The arid region of Northwest China was located in the central and eastern part of Central Asia.The surface environment and climate in this region were caused by the high-frequency and large-scale complex climate evolution since the Quaternary of Cenozoic.Since 2.0 Ma BP,the continuous uplift caused by the moving northwards of Indian Ocean Plate had created the world ridge of the Himalayas,which affected the distribution of monsoon,trade wind and ground temperature in Central Asia and the world,and had a great and important impact on global climate change.Net primary productivity(NPP),as a quantitative index describing the cumulative efficiency of organic matter in plant bio-environment,could effectively express the cumulative benefits of material in ecosystem and the quality of system development,which was of great significance to the study of ecology and sustainable development of environment.Under the fragile ecological environment background in arid areas,the study of ecological security maintenance and system quality monitoring was particularly important for regional development and environmental protection.Completeness of this study of NPP in arid areas could effectively reveal the development trend of environmental quality.This paper aimed to explore the driving forces of important environmental factors of NPP in arid areas by taking typical sample areas in arid areas as data analysis objects.With the rapid development of modern data platform,scholars and researchers could easily access a large number of scientific datasets.In this paper,a large number of scientific data sets of environmental factors related to NPP were obtained by means of scientific data websites such as the Resource and Environment Science Data Center of the Chinese Academy of Sciences and the Science Data Center of Cold and Arid Regions.By introducing C.V(coefficient of variability,C.V = SD/Mean)spatial computation and wavelet signal analog processing methods,the original data were optimized and the spatial proximity variation relationship was highlighted.The data matrix was processed individually by using Matlab software,and then the NPP model and factor analysis were completed to obtain the data analysis results.The extraction and construction of data sets depended on the original data format and quality.The primary data obtained in this paper was environmental attribute product data of raster,which was organized by attribute layers.In the process of attribute value extraction,ArcMap software was used to adjust the projection and coordinates of data space,constructed the attribute layer of grid points in the sample area,and then used the extraction tool to extract multi-layer attribute information in batches.The main idea of the wavelet processing was to use the wavelet denoising technology of signal processing to realize the data de-redundancy and continuous interval acquisition after linear processing of scattered data,and optimize the quality of data and analysis results.In the process of data extraction and database construction,9 data sets were obtained for statistical analysis and model construction,which were the basic data sets of North Xinjiang,C.V.and wavelet data sets,and the data sets of South Xinjiang and Inner Mongolia research sample areas.Through scientific statistical analysis,the following research results were obtained:The results of factor correlation analysis suggested that both C.V data set and wavelet data set had higher significant expression and greater partial factor correlation coefficient compared with the basic data set.The correlation analysis results of C.V data sets showed that the coefficients are large,and there were great differences between C.V data sets and basic data sets and wavelet data sets.In addition,the factor linear relationship of C.V data sets was weak in numerical distribution.The correlation analysis results of C.V data sets showed that its coefficients are larger,and results had great differences between C.V data sets and basic data sets and wavelet data sets.In addition,the factor linear relationship of C.V data sets was weak in numerical distribution.The correlation between the factors of the basic data set and the two-dimensional distribution were similar to the results of the wavelet data set.In the process of wavelet processing,the outlier was eliminated,and the analog signal curve data was interpolated.The correlation coefficient was relatively larger and the significance was stronger.After factor dimension reduction analysis,it was found that vegetation coverage,soil erosion,precipitation and topographic factors were dominant factors in various environmental types.Among the results,the lowest contribution rate of composite factors was 65.35%,and the average contribution rate was 83.55%.The two-dimensional distribution of factor ranking in northern Xinjiang and Inner Mongolia was relatively random,while that in southern Xinjiang was mainly uniform in the direction of dry and heat factors.A small number of environmental attribute groups had more prominent environmental advantages in the water vapor conditions and terrain composite factors.By comparing the results of data set model building,we could get that both C.V data set and wavelet data set had more models than the basic data set,and the model fitting degree was higher.Most noticeable was the Inner Mongolia data set,because the basic data set only obtained one linear model under the restriction of 0.05 entry and 0.10 exclusion,and other factors except NDVI were not selected.The wavelet data set had 7 models,the number of factors increases by 6,and the fitting degree of the model reaches 0.767,which was significantly improved compared with the basic data set.The number of data set models constructed in other research sample areas is more than that in Inner Mongolia,and the fitting degree was also higher.However,the construction of the optimal fitting model of the wavelet data sets in the southern Xinjiang and Inner Mongolia research sample areas was very similar,which breaked the state that the optimal fitting degree was the highest in southern Xinjiang,the second in northern Xinjiang and the lowest in Inner Mongolia.Among them,because the data attribute meaning of C.V data set was different from other data sets,the original data was polarized,so,a few models had abnormally high fitting degree.With a large number of subsequent model analysis and factor analysis,the ranking results of NPP driving factors in arid areas showed that NDVI was the highest factor of average contribution rate,and the contribution rate of each model was close to 50%.Afterwards,the second factor in the study area of northern Xinjiang was Aat10(>10 ? cumulative temperature)factor,while that in southern Xinjiang was Tadem(annual mean temperature)factor.The contribution rate was close to 10%.The terrain factor in the study area of Xinjiang was higher than that in Inner Mongolia,but its contribution rate was relatively low.It was worth emphasizing that the results obtained from the wavelet data set in the model construction were the best,the number of factors in the model was relatively stable,and more factors could be expressed in the model under the premise of less change in fitting degree.After wavelet processing,the fluctuation between two-dimensional distribution and factor of datasets was relatively weak,and the quality of data analysis was effectively improved.Its application value in factor analysis and modeling analysis was enormous,and the experimental analysis work under different scales will had important scientific significance.