The Effects Of Sloping Landscape Features On Water Quality In Typical Karst Watershed

In this paper,the middle and upper reaches of the Chishui River in typical karst basin were chosen as the study area,field water sampling data,remote sensing and GIS technology were used to extract the landscape features of different slope grades in the study area.The relationship between landscape features and river water quality parameters was discussed and the“best”sloping landscape characteristics-water quality response model was established for different water quality parameters based on the classical linear statistical model and the artificial neural network model.The relevant results are shown as follows:?1?The landscape composition of the study area was mainly composed of forest land,shrub-grass,cultivated land and construction land.Either in sub-basin scale or in buffer scale,the slope landscape?6°-25°?compositions were mainly slope land,slope shrub-grass,slope cultivated and slope construction land,and the steep slope landscape?>25°?compositions were mainly steep slope forest land and steep slope shrub-grass.From the perspective of the comprehensive land use index,human activities had the highest development intensity for the flat landscape?235,246?,followed by the slope landscape?218,223?and the smallest to the steep slope landscape?90,207?.Among them,the development intensity of human activities on the slope landscape?216?was almost equivalent to the development intensity of the overall landscape?219.84?.From the perspective of landscape pattern,at the landscape level,from slope landscape to steep slope landscape,the fragmentation degree,diversity and separation degree of plaque were reduced,the shape of plaque was more regular,and the continuity between landscape types was enhanced.At the level of landscape type,the patch shapes of slope woodland and steep slope farmland were more complicated,and the patch shape of steep slope-shrub was more irregular than that of slope shrub.?2?In terms of landscape composition,the correlation among total phosphorus?TP?and slope landscape features were obvious both in sub-basin scale and in 500 m buffer scale,and is significantly negatively correlated with slope forest land and total forest land at sub-basin scale?-0.556,-0.540?.this phenomenon occurred because the slope forest land that accounts for more than 50%of the slope landscape area is retained by infiltration,bioabsorption of plant roots,and adsorption of phosphorus,thus hindering the transmission of pollutants;TP is also associated with slope irrigation and The slope construction land is significantly positively correlated?0.656,0.566?,indicating that the“source”effect of slope irrigation on river water quality will be masked if the slope grade is not divided.Total nitrogen?TN?was significantly positively correlated with sloping land and total arable land at the sub-basin scale?0.616,0.492?,indicating that the application of nitrogen fertilizer combined with severe soil erosion in slope farmland increased the TN content in the rivers in the study area.Ammonia nitrogen(NH₄₊-N)was significantly negatively correlated with slope woodland at the buffer scale?-0.479?,but was significantly positively correlated with steep slope forestland?0.402?.The slope was steeper,the soil's stabilizing ability was worse,and ability of the roots to absorb nitrogen was weaker.In addition,the steep slope forest contained the fruit-bearing forest with nitrogen fertilizer,which made the NH₄₊-N positively correlated with the steep slope forest.However,the correlation between NH₄₊-N and total forest land was not obvious.It also indicated that the“source”effect of NH₄₊-N on the steep slope forest land without slope grade would be masked by the“sink”effect of slope land on NH₄₊-N.NH₄₊-N was significantly negatively correlated with flat cultivated land,sloping land and total cultivated land?-0.727,-0.628,-0.637?,and the correlation with steep slope farmland was not significant,because soil particles,soil colloids and crops were NH₄₊-N.Strong adsorption,blocking the transmission of pollutants to river water.In addition,NH₄₊-N was significantly positively correlated with slope irrigation grass,slope construction land and steep slope construction land?0.429,0.405,0.494?,indicating that the slope construction land and slope irrigation grass were“source”landscape,so it was reasonable to control the construction area of slope land.Improving river water quality had a positive effect.?3?At the landscape level,with the increase of landscape plaque fragmentation value?PD?(PD _steep-slope=8.65,PD _total=12.23,PD _slope=20.22),the correlation between TP and PD was enhanced?0.507,0.593,0.623?.At the same time,as the average patch shape index value?SHAPE-MN?increased(SHAPE-MN _steep-slope=1.19,SHAPE-MN _slope=1.25,SHAPE-MN _total=1.37),the correlation between NH₄₊-N and the average patch shape index was also enhanced?0.516,0.654,0.761?.It showed that the high degree of landscape fragmentation and the more irregular plaque shape had negative impacts on water quality.?4?At the level of landscape type,TP was significantly positively correlated with the maximum plaque index?LPI?and the SHAPE-MN of slope construction land?0.498,0.557?,and showed significant positive correlations with the SHAPE-MN of slope shrub and the SHAPE-MN of slope construction land?0.529,0.567?at sub-basin scale.That is to say,TP was sensitive to the plaque complexity and aggregation degree of the slope construction land,and was the plaque shape complexity of the slope irrigation grass.NH₄₊-N was significantly negatively correlated with slope woodland shape index?LSI?and steep slope forest LSI?-0.721,-0.691?,and negatively correlated with LSI and steep slope LSI?-0.705,-0.644?.That is,with the increase of slope,the more complex the plaque shape of forest land and cultivated land were,it had the lower the correlation with NH₄₊-N,and the adsorption capacity to NH₄₊-N was weaker.In addition,NH₄₊-N was significantly positively correlated with the LPI of the steep slope construction land?0.749?,which was consistent with the conclusion that NH₄₊-N was significantly positively correlated with the steep slope construction land.?5?The comprehensive impact of sloping landscape features on river water quality was so complex that the classical linear statistical model was difficult to express the impact of sloping landscape features on river water quality.By introducing the artificial neural network model,which is with adaptive,self-learning,self-organizing and nonlinear mapping functions.Good results in simulating the influence of sloping landscape features on water quality were achieved.The BP neural network model is the“best”sloping landscape feature-water quality response model of TN and TP.The sloping landscape features participating in TN modeling are slope arable land and sloping land IJI.Compared with classical linear statistical model,the decision coefficient of BP neural network model was increased to 0.923,the average relative error value was reduced to 14.30%,and the average absolute error was reduced to 0.332.The landscape features involved in TP modeling including slope farmland IJI and total cultivated land.The coefficient of the model was increased to 0.960,the average relative error was reduced to16.06%,and the average absolute error was reduced to 0.006.