Landscape pattern has a great impact on the surface water quality. Many studies have focused on the relationship between the two aspects since 1970s, especially from the appearance of the landscape ecology method which remarkably promoted the studies after 1990s. This paper firstly reviewed the research theory and method in the related studies in the recent 30 years in abroad and home. It is shown that the landscape pattern metrics in the studies has developed from qualitative analysis, land use configuration, land use spatial pattern, and currently to an integration of landscape pattern and hydro-geomorghy structure, while the focus of surface water quality indices has expanded from hydro-chemical ions, normal pollutants, nutrients, heavy metals, organic pollutants, and currently to aquatic organisms and stream health as well. In the field of hydro-chemical ions, it is found that only qualitative not quantitive method were used just to compare the ion concentrations in different land use area, which may be caused by the lack of landscape theory and GIS technique; In the field of the pollutants, there is an interesting finding that physical indicators such as water temperature, total suspended solid (TSS), conductivity and pH may be more sensitive to the change of landscape pattern than the chemical indicators; In the field of nutrients, it is revealed that there is often an notable difference between N and P in their response to landscape variation, more usually, N is more sensitive than P because of their different source form human activities; and in the field of aquatic organisms and stream health, it is emphasized that the impervious area index should be used as a key environmental indicator, in which the threshold effect are of great necessity to be discussed, which often shows that biological indicators are more intended to have a lower threshold than the chemical and hydrological in their response to the change or variation of landscape pattern. Great attention has been paid to the problem of scale effect, however, the results are still confused in the problem of what is an effective scale would represent the most variation of water quality data. Critical techniques were summarized, such as step-wise multi-regression, Pearson relationship, gradient analysis, and model estimation. Some issues are put forward such as (1) watershed embedding and spatial auto-correlation in the delineation of watershed boundary, (2) the spatial configuration and hydro-morphology was often ignored in the selection of landscape metrics, and (3) the qualitative and semi-quantitative analysis method was mostly choused in the data analysis, and corresponding strategies were given. Finally, the prospect of domestic studies is tentatively suggested, which should be based on two theoretical premises, namely whether the watershed boundary could be clearly defined and whether the point pollution in the study area was completely controlled, and the most important is how to define a nice landscape classification system according to the cumulation and transformation activity rules of non-point pollutions in different landscape arrangement. Some key scientific points were proposed for the application of landscape pattern and process theory to urban non-point pollution control, which including how far is the non-point source to the target water, the spatial pattern and threshold of impervious area, and what is the effective scale to setting the pervious area like wetland and riparian buffer zone.