Abstract:The influence of landscape pattern on water quality will change depending upon scale, so it is meaningful to probe how changes in landscape pattern influence water quality at differing spatial-temporal scales. For water-environment protection of reservoir watersheds in Shenzhen city, this paper took Xili reservoir watershed and its sub-watersheds as research areas. This paper studied the influence of landscape spatial differences on water quality at differing temporal and spatial scales. At the temporal scale, this paper chose water quality monitoring data at the Xili reservoir outlet in 2000 and 2001, which was more heavily influenced by non-point pollution than point pollution. Spatially, due to the limits of data, this paper divided data into two spatial scales: the Xili reservoir watershed and its sub-watersheds. The Location-weighted Landscape Contrast Index (LCI) was developed and applied by Prof. Chen Li-Ding. It is used to analyze the relative influence of landscape spatial elements upon water quality. This paper first suggested some improvements to the LCI and then used those improvements to calculate elevation, slope and distance LCI with divisional or cumulative characteristics at the Xili reservoir watershed and its sub-watersheds. Then we used this index to analyze the influence of ‘Source’ and ‘ink’ landscape spatial difference changes upon water quality at differing temporal-spatial scales. The results show that LCI can clarify the differing roles of ‘Source’ and ‘Sink’ landscape in the non-point pollution process. We can use this method to analyze how landscape spatial differences influence water quality. Moreover, LCI reveals that three main spatial elements——elevation, slope and distance——limit human activities upon landforms. So we should exploit the ability of natural landforms to intercept and decrease non-point pollution. However, LCI also has some shortcomings in practical usage. It is influenced by the complexity of the landscape spatial difference and dramatically influenced by differences in quantity and type of ‘Source’ and ‘Sink’ landscapes. It would therefore not be suitable for analysis at differing spatial scales for contrasting scale effects. So we should use the LCI method at appropriate scales, which will help the realization of research aims.