A Study On The Relationship Between Multi-scale Landscapes And River Water Quality Response In The Miyun Reservoir Upper Watershed

It has been commonly accepted that multi-scale landscapes within a watershed have intensive influences on physical, chemical and biological features of a river. At present, such research has shifted from simply qualitative studies to advanced quantitative ones on regional or even national scale using statistical and hydrological models as tools. However, such research in China is just in the starting stage. More specifically, most studies focus on the relationship between non-point source pollutant migration and simple land use pattern on single spatial scale rather than concerning continuous processes on a series of spatial scales. Therefore, this doctoral dissertation studied the scale dependent relationship between landscape pattern and non-point pollution processes on a series of spatial scales (sub-catchment, riparian zone, and local reach) and the scale adaptability of control action of landscape pattern variables on environmental responses. This dissertation took the upper watershed of Miyun Reservoir as the study area and adopted Geographic Information System (GIS), Redundancy Analysis (RDA), Multiple Linear Regression (MLR), Factor Analysis (FA), and Spatial Regression (SP) to achieve the objectives which are:1) to explore the interactional mechanism of pollutant migration,2) to identify the key landscape types and patterns that impact surface water quality on a series of spatial scales,3) to discriminate the impacts of various factors on the relationship between landscape pattern and water quality. This doctoral dissertation is expected to provide a set of theoretical basis and technical guidance for effectively controlling non-point source pollution. Also, this dissertation will be helpful to further controlling eutrophication of surface water and promoting drinking water source protection. The main results of this doctoral dissertation are listed as follows: (1) Adopting Kruskal-Wallis Test and analysis of variation coefficient, we analyzed the water quality parameters of 1376 samples collected in about 50 monitoring sites in the Miyun Reservoir upper watershed. The results show that changes in the concentration of pollutant have significant spatial and temporal difference and the maximum pollutant concentration is mostly in the rainy season. By comparing the concentration of pollutants within nested watershed on a series of hydrological scales, we find that the concentration of Total Nitrogen (TN) and nitrate (NO3-) shows a descending trend with the increase in watershed area. Conversely, the concentration of NO2- has a converging trend with the spatial scale shifting from small to large. While the concentration of NH4+ has no noticeable trend among scale transformations, this is probably because NH4+ is affected by point source pollution. The values of Chemical Oxygen Demand (COD) and Total Organic Carbon (TOC) increase with spatial scale shifting from small to large, and the increasing trend of TOC value is more significant. The migration of Total Phosphorous (TP) is nearly not affected by hydrological scale transformations.By factor analysis of chemical and physical indexes of water quality, we find that the indexes that have leading impacts on water quality variation change with seasonal evolution.Before the rainy season, basic physical and chemical parameters (pH, DO, Temp, COND) have the most significant impacts on water quality variation, and then come the eutrophication indexes (TN, NO3-, NH4+, TP, Chl), aerobic pollutants (TOC, COD) have the minimum impacts. During and after the rainy season, the eutrophication indexes have the maximum impacts on water quality variation, and then come basic physical and chemical parameters, aerobic pollutants have the minimum impacts. TN and NO3- are the most important indexes that have greatest impacts on water quality. Inferior to the impacts of TN and NO3-, TP is also valuable for measuring eutrophication of water body.(2) Landscape composition variables (farmland proportion, forestland proportion, grassland proportion and urban proportion) and landscape structure variables (patch density index, edge density index and contagion index) are selected to represent artificial landscape features. Also, watershed area, average slope and drainage density are selected to represent physical landscape features. We carry out the study on the relationship between landscape and water quality on three spatial scales which are sub watershed, riparian zone and local reach.By factor analysis we find that landscape variations within the study area are mainly caused by anthropogenic landscape characteristics that are represented by landscape composition and configuration.Natural landscape features represented by hydrological variables have limited impacts on landscape variations, thus have high spatial homogeneity. Landscapes within 100 m on the riparian scale can best explain water quality variations. Therefore, response relationship between landscape and riparian environment within 100 m can represent the impacts of landscape features on the riparian scale on water quality. Local reach can not explicitly explain water quality variations, thus this scale is not suitable for predicting physical and chemical characteristics of the river in the study area.(3) We explore the importance of landscape factors by MLR and RDA. The results show that Drainage Area (DA) has substantial impacts on water quality. Other landscape variables that have sufficient impacts on water quality are Residential area, Forest and CONTAGE (CONT) in sequence. Well-developed fruit planting industry (mainly chestnut planting) resulted in the positive correlation between forest and TN and NO3-. Seasonal hydrological processes and the intensity of human activities have substantial impacts on the relationship between landscape factors and riparian environment. Before the rainy season, landscapes have little impacts on water quality. During the rainy season, landscapes can best explain water quality variation because of maximum river flow and extensive agricultural activities. After the rainy season, river water level keeps dropping and crop is harvested, the impacts of landscapes on water quality weaken, but still slightly stronger than those before the rainy season.The results of partial RDA show that the impacts of natural factors and human factors on water quality variation change with seasonal evolution. Before and during the rainy season, both have the similar impacts on water quality. However, after the rainy season human factors have stronger impacts on water quality than they do in the other two seasons. Land use variables can better explain water quality variations in each season than landscape structure variables. Landscapes on the riparian scale can better explain water quality variations than those on the watershed scale, because the most representative riparian width was identified and landscape patterns on the riparian scale are quite different from those on the watershed scale.(4) Landscape indexes generated from highly precise land use data are not necessarily more suitable for explaining water quality variations than the indexes with low precision. When selecting land use maps or remote sensing images with applicable precision as data source, researchers should take into account characteristics of the study area and research objects. Although the impacts of variation of land use classification systems on water quality are weaker than those of data precision, these impacts are not stable on different spatial scales, having an increasing trend with the decreases in spatial scale.Among numerous influencing factors, the increase in samples has the most significant impacts on the relationship between landscape and water quality. The analysis results by RDA show that in the dense sampling scheme the significance of natural factors noticeably is enhanced. Also, the significance of farmland and forestland that belong to landscape components is reflected. Although the seasonal variations of impacts of landscapes on water quality become more evident, the ability of landscapes to explain water quality variations gets declined.Adopting spatial regression method, we find that spatial weighting factors can really improve the ability of landscapes to predict water quality, however the effects are limited. The maximum increase in coefficient of determination is 0.18, the minimum value is 0, and the average is less than 0.05. The improvement of predictive ability does not vary with spatial scale transformation, which means no scale effects.The percentages of Agriculture or Residential area have no apparent threshold effects on water quality in the study area.