Discussion

Model performance

Sensitive parameters determined the model performance. Their initial ranges were referred to Zhao et al. (2011) and the optimum parameter sets were obtained from the simulation results with the maximum NS value. Compared with previous case studies in Xitiaoxi watershed for mountain hydrology modeling by Chen et al. (2019), the NS value and R2 value of daily flows at the Hengtangcun (Fig. 2) implied that the simulation results by Xin’anjiang model were “very good” during the calibration and validation periods. According to Moriasi et al. (2007), streamflow simulations were regarded as satisfactory for SWAT when NS ≥0.5.
In lowland artificial watersheds, the NDP model considered the dominating mechanisms affecting polder water balance. They were artificial drainage, the water interactions among surface water and groundwater, as well as soil water in farmlands (Huang et al., 2018a). The NSvalues of water level were above 0.5 during the calibration and validation periods. This model fit was acceptable compared with the surveying watershed modeling cases using 257 models (Wellen et al., 2015). Our study area was close to the case study by Huang et al. (2018a) in space and under the unified polder-controlled strategy by Taihu Basin Authority of Ministry of Water Resources. Therefore, these similarities between the hydrological conditions in the previously modelled data and the current hydrological year being modelled can illustrate the significance and reliability of the model fits.

The different responses of hydrological processes to climate change and land use

Our study found that climate and land use change can cause higher surface runoff, which was consistent with previous study by Berihun et al. (2019). However, Berihun did not summarize the differences of hydrological responses between his three study watersheds with different biophysical characteristics, which was the key augmentations of our study. Wang et al. (2019) found that the corresponding sensitivity of streamflow to changes in climatic conditions and human activities varying from watershed to watershed. Therefore, it was important to analyze the hydrological processes and their driving factors at watershed scales.
In our study, climate change played a more critical role on the hydrological process in mountain watersheds than in lowland artificial watersheds. The variation of mean annual discharge resulting from climate change mainly due to the precipitation factor during the study period across two watersheds (R2 =0.78 in mountain, R2 =0.96 in lowland). In mountain watersheds, the annual average discharge would increase 10~200% from 2015 to 2100 under three climate scenarios (RCP2.6, RCP4.5 and RCP8.5 from CMIP6), comparing to 10~60% in lowland watersheds. In terms of seasonal discharge variation, it showed remarkable change during autumn and winter in mountain watersheds as their small base values. However, rice seasons were the “hot moments” of flood in lowland artificial watersheds.
Land use played a more critical role on the hydrological process in lowland artificial watersheds than in mountain watersheds. Converting cultivate land to residential land made a significant improvement on annual average discharge comparing to other land use scenarios, with an increment of 22.0% in lowland and 7.8% in mountain respectively, under the same conversion rate of 36.1%. In terms of the “hot moments” of flood under land use conversion scenarios, in lowland watersheds, the effect of converting cultivate land to residential land on seasonal runoff variation was more than 20.0% in the whole year expect winter. Moreover, when converting cultivate land to water area, seasonal runoff increased more than 10.0% in rice seasons especially during irrigation periods. In mountain watersheds, urban expansion from other land use types such as forestland and grassland caused seasonal runoff increasing especially in May, during the crops and plants growth seasons when consuming a lot of water. Another found in mountain watersheds was that runoff showed a positive correlation with the slope, implicating that the weak water storage capacity of sloping regions lead to more frequent rainstorm-runoff processes.
Land use can enhance climatic impact on hydrological process in lowland artificial watersheds, comparing to the non-significant tendency in mountain watersheds. For the 2050s, climate change would cause an increasing annual runoff by 312.9‬ mm and 349.7 mm in mountain and lowland watersheds, respectively. When combined with land use change, annual runoff increased another 55.1‬ mm and 269.3 mm in mountain and lowland watersheds, respectively.

Implication for water managements

As land use and climate changes are expected to be intensive in future, many regions in the world may suffer from frequent droughts and floods. The relative effects of climate change and human activity vary among different watersheds as well as different periods (Ye et al., 2013). However, the effects of climate change and land use are always underestimated or overestimated by different methods, interfering decision makers to manage water resources in a sustainable way (Wang et al., 2019).
In our study, increased precipitation and temperature under future climate conditions would cause higher streamflow especially in mountain watersheds, which seems to have slight impact on evaporation. Therefore, floods in sloping regions may become more frequent under future climate effects, leading to more severe flood disasters. The major effects of land use change on hydrological processes for the 2050s will happen when the urban area expanding rapidly at the expense of cropland. Cropland is an important land use type in lowland artificial watersheds, for the disappearance of its water retention capacity will destroy flood prevention by polders. Moreover, pumping stations setting for cropland can increase the hydraulic retention time in lowland watersheds. The “hot moments” of flood in lowland artificial watersheds would be rice reasons when converting cropland to residential area, comparing with crop growing seasons in mountain watersheds without pumps. To sum up, precisely forecasting the mountain torrents during rainstorms, controlling urban expansion and maintaining the cropland area of polders could be potential strategies for flood prevention and water resources protection in highland-lowland watersheds.