Multivariate correlation analysis between landscape pattern and water quality

Water quality variation is generally linked to the change of landscape pattern in watershed,which represents the main impact of human activities in macroscopic view.Therefore,identifying the crucial landscape factors that affect water quality variation is valuable for understanding the mechanism that landscape may affect water quality.Multivariate analysis tools are effective methods to deal with complex correlations between landscape pattern and water quality.Besides,advances of remote sensing(RS) and geographic information systems(GIS) technologies have made regional and watershed scale studies much more feasible.This study was conducted along Danshui River watershed,a branch of Dongjiang River in Guangdong Province.The correlation between landscape pattern and water quality of Danshui River was represented by using spatial analysis and multivariate analysis methods base on ALOS satellite image and water quality monitoring data in 2007.Landscape metrics,including information of landscape composition and spatial configuration,were used to represent landscape pattern.In order to cover overall landscape information,landscape metrics on both watershed scale and riparian scale were used.Spearman′s rank correlation analysis,multiple linear regression models with step-wise and canonical correlation analysis(CCA) were used to reveal the linkage between landscape metrics and water quality.The results show that forest,urban and agriculture land use are accounted for more than 90% of the total area in Danshui River watershed,while the area proportion of urban land exceeds 20%.Results of multiple linear regression models with step-wise and CCA showed that water quality indicators were affected by more than one landscape metric.The variation of water quality was influenced by landscape pattern significantly.The landscape metrics in watershed scale revealed more information of water quality variation than landscape metrics in riparian scale.The proportion of urban land use proportion had the greatest impact on water quality.Spearman′s rank correlation analysis and multiple linear regression models showed the proportion of urban land use was the most important contributing factor to cause variation of oxygen consuming pollutants and nutrients.However,forest and agriculture land use had less influence on water quality.On the other hand,landscape metrics about landscape fragmentation were crucial factors to affect indicators of water quality,such as pH,DO and heavy metals.The result of CCA indicated that the first ordination axis could explain 54.0% of the correlations between landscape metrics and water quality indicators,and the first two ordination axes could cumulatively explain 87.6% of the correlations between landscape metrics and water quality.The result of CCA revealed that water quality had an obvious trend with the varying landscape gradient.The first two ordination axes mainly represented urbanization gradient and landscape fragmentation gradient respectively.Landscape characteristics in the study area showed a gradient of urban,urban-rural fringe,rural from upstream to downstream of Danshui River watershed.The distribution of pollutants concentration was corresponded with the gradient of landscape pattern in the watershed.Land use and cover change is an integrated result due to human activities,and change the state of eco-system of river and watershed significantly.It′s highly reasonable that the water quality must correspond to the change of watershed landscape.