| Taiyuan basin is located in the mid-eastern Loess Plateau of China and the center of Shanxi Province.The landform of these basins is low-lying and flat,and fits for crop plant.However,the soil in Chinese Loess Plateau is very poor,and shortage of soil nutrients is the limiting factor for crop growth and land productivity.Meanwhile,the distribution of soil nutrients are highly heterogenic and scale-dependence.Therefore,analysis of soil nutrients' spatial distributional characteristic,the multi-scale spatial correlation of soil nutrients with influencing factors,and the prediction of soil nutrients deeply and scientifically are the strong guarantees for soil processional management,dimension design of farmland,rationally fertilizing,and various projections of ecological rebuilding.In this study,three sampling transects containing 383 soil sampling points on field were collected,which were along the vertical direction of Fen River,and soil nutrients(the indexes including soil organic matter,total nitrogen,total phosphorus,total potassium,available nitrogen,available phosphorus,available potassium,soil bulk density,soil texture,soil pH,total salt content,and Vis-Nir hyperspectral)were tested.Meanwhile,27,664 points of soil nutrients from 2006 to 2013 in 12 counties were gathered.Based on the historical soil data,a sampling transect containing 128 soil samples along the direction of Fen River was extracted.To understand the spatial distributional characteristics of soil nutrients in the study area and the differences between the different sampling datasets,the descriptive statistics were done for these dataset,including three sampling transects along the vertical direction of Fen River,a sampling transect along the direction of Fen River,and the historical dataset of soil nutrients.Meanwhile,based on the three transects along the vertical direction of Fen River,the soil nutrients contents(soil organic matter and total nitrogen)were predicted by the partial least squares regression.The spatial multi-scale correlation between soil nutrients(soil organic matter and total nitrogen)and influencing factors were analyzed by the multivariate empirical mode decomposition,and the contents of soil organic matter and total nitrogen were prediction.Additionally,based on the sampling transect along the direction of Fen River,the specific scale and location variation of soil nutrients(soil organic matter,total nitrogen,available phosphorus,and available potassium)were revealed by wavelet transform.The specific scale and location correlation between soil nutrient and topographic factors were analyzed by wavelet coherency.Finally,based on the combination of three sampling transects and historical dataset of soil nutrients,the specific scale and location variation of soil nutrients in the full basin were analyzed by wavelet transformation.And the main inclusions included:(1)Compared with the contents of soil nutrients in China,soil organic matter were in the low levele,total nitrogen,total phosphorus,and available phosphorus were in the medium level,and total potassium and available potassium were in the higher level in Taiyuan basin.Compared with the variation of soil nutrients in the Loess Plateau,the variations of soil organic matter and total nigrogen were lower,total phosphorum were approximate,and total potassium were higher in the Taiyuan basin.The ascending order of soil properties' variation was pH value,bulk density,total nitrogen,silt content,total potassium,clay content,total phosphorus,soil organic matter,available potassium,sand content,available nitrogen,and available phosphorus.Compared the three sampling transects with the historical soil dataset,the averages of soil nutrients were altered littlely,which indicated the variation of soil nutrients in the time scale.(2)The results of soil nutrients prediction by hyperspectral method indicated that when the calibration dataset and validation dataset were independent in Taiyuan basin,the trends of the accuracies for calibration and validation on soil organic matter were increasing and then decreasing along the number of latent increasing,the tend of the accuracies for calibration on total nitrogen were also increasing and then decreasing while the tend of validation was unstable.Meanshile,the predicting accuracy of soil nutrients could be evaluated by the spatial structural analysis and variation in the location-scale domain,and predicting error could be obtained by the alteranation of spatial structure,or variation in the location-scale domain between measured and predicted values.(3)The spatial multi-scale correlation between soil organic matter and influencing factors by the multivariate empirical mode decomposition showed that the representive scales for upstream in the basin were 1101,and 1725 m,for midstream were 982 and 8573 m,and for downstream were 960,6753,and 11806 m.The scale of 1 km was the main representive scale of soil organic matter in the total area,and the main representive scales for the vertical directional transects of Fen River were decentralized along the Fen River's direction.Meanwhile,the correlation between soil organic matter and influencing factors at the sampling scale and spatial multi-scales indicated that the descending order was midstream,downstream,and upstream.Additionally,the correlation of spectral component 1 with soil organic matter were most significant in the three sampling transects.Soil bulk density for upstream,sand content for midstream,and topographic wetness index for downstream were very significant,while the correlation between topographic wetness index and soil organic matter in the downstream was unsignificant.Therefore,the analysis in the sampling scale could not reveal the complicated relationships in the spatial multi-scales,while the predicting accuracy of soil organic matter by multivariate empirical mode decomposition was obviously higher than the simply stepwise multiple regression.(4)The spatial multi-scale correlation between soil total nitrogen and influencing factors indicated that the main representative scale were 970,and 2637 m for upstream,were 954,8849,and 1476 m for midstream,were 1003,8438,8719,and 2810 for downstream.The scale of 1 km was the main representive scale of total nitrogen in the total area,and the main representive scales for the vertical directional transects of Fen River were decentralized along the Fen River's direction.Meanwhile,the correlation between soil total nitrogen and influencing factors at the sampling scale and spatial multi-scales indicated that the descending order was midstream or downstream,and then upstream.In the three sampling transects,clay content for upstream,sand content for midstream,and spectral component 1 for downstream were very significant with soil total nitrogen.The predicting accuracy for total nitrogen by multivariate empirical mode decomposition was higher than that of stempwised multiple regression.Therefore,the analysis in the sampling scale could not reveal the complicated relationships in the spatial multi-scales for soil total nitrogen.(5)The variation of soil nutrients in the location-scale domain by wavelet transform showed that the variance at the scale 0-4 km were high and the area of that were small;variance at scale 4-8 km were relatively low;variance at scale>8 km for soil organic matter and available phosphorus were very high,and the areas of them were very centralized.The specific scale-location correlation between soil nutrients and topographic factors by wavelet coherency along the direction of Fen River indicated that the correlation at scale>8 km were stable;The relationship between soil nutrients(total nitrogen and available potassium)and elevation or slope were fully nagetive correlated;the relationships between soil nutrients(soil organic matter,available phosphorus,and available potassium)and topographic wetness index were fully positive correlated.(6)Combining the historical data of soil nutrients from country territory and the three sampling transects along the vertical direction of Fen River,the spatial variation of soil nutrients at the location-scale domain by wavelet transform indicated that the 95%significant variance for soil organic matter and total nitrogen were relatively centralized.Meanwhile,the area of significant variance at scale 0.68-1.21 km were the biggest for soil organic matter,and the area of significant variance ta scale 1.53-10.30 km were the biggest for available phosphorus.Additionally,the area of significant variance for soil organic matter was bigger than soil total nitrogen,and the area of significant variance for available phosphorus was bigger than available potassium.Finally,the area of 95%sigenificant variance for soil organic matter,total nitrogen,available phosphorus,and available potassium at scales 7.28,10.91,4.33,and 4.33 km respectively were big,and the percentage were 21.39,17.11,26.47,and 17.71%,respectively. |
PDF Full Text Request