The effects of landscape pattern change on hydrological processes have become one of the major ecological concerns in the world. Some researches have been working on the yearly and monthly runoff response to the landscape pattern change based on distributed hydrological model. How the daily runoff, such as flood and low flow, responds to landscape pattern change has not been well studied. The study area of Jinjiang watershed, which is situated at southeastern China, is one of the regions where the economy grow fast and the landscape pattern change dramatically. It is worth to reveal how the change in landscape pattern influences the flood and low flow for the regional ecological security and economic development. In this paper, an integrated approach of distributed hydrological modeling and Pearson correlation analysis was applied to quantify the contributions of the changes of landscape indices on the variation in flood and low flow. Daily stream flow under the meteorological condition from 2002 to 2010 was simulated with the landscape pattern maps in two periods (1985 and 2006) using the Soil and Water Assessment Tool (SWAT), respectively. The changes of annual maximum 1-day, 5-day, and minimum 1-day, 7-day stream-flow between two simulations were calculated. Finally, the changes of stream-flow were related to the changes of landscape indices in Pearson correlation to quantify the impacts of changes in landscape on that of flood and low flow.
The results show as following: (1) All Ens and R² are above 0.75, and Ers are in the range of ±10% in the calibration and validation period for three gauge stations, suggesting SWAT model is well performance. (2) Compared with the landscape in 1985, the annual mean maximum 1-day and 5-day stream flow from 2002 to 2010, which was simulated by SWAT model with the landscape in 2006, increases by 5.46% and 4.97% in the Jinjiang watershed, while the annual mean minimum 1-day and 7-day stream flow decreases by 3.79% and 2.55%, respectively. (3) There are significant correlations between the variation of flood flow and landscape indices. The forestland area is negatively related with annual maximum 1-day and 5-day stream flow with the correlation coefficients of -0.764 and -0.721. The glass area is positively related with annual minimum 1-day and 7-day stream flow with the correlation coefficients of 0.461 and 0.478, respectively. It demonstrates that the increase of forest area can reduce the discharge of flood, while the increase of glass area can increase the discharge of low flow. Furthermore, the relationships between landscape metrics and flood are also revealed. SHDI and SHEI are significantly positively related with maximum 1-day flow with the correlation coefficients of 0.721 and 0.736, respectively. On the other hand, LPI and CONTAG show negative relationships with maximum 1-day flow with the coefficients of -0.61 and -0.596, respectively. The relationships between landscape metrics and flood reveal that equilibrium distribution of landscape classes and low connectivity reduces hydrologic cycle of the basin and eventually leads to the increase of flood flow. (4) Landscape metrics are not significantly related with low flow indicate that the impact of the change in landscape on that of the low flow is not so important.