Improving the forecasts of extreme streamflow by support vector regression with the data extracted by self‐organizing map

During typhoons or storms, accurate forecasts of hourly streamflow are necessary for flood warning and mitigation. However, hourly streamflow is difficult to forecast because of the complex physical process and the high variability in time. Furthermore, under the global warming scenario, events with extreme streamflow may occur that leads to more difficulties in forecasting streamflows. Hence, to obtain more accurate hourly streamflow forecasts, an improved streamflow forecasting model is proposed in this paper. The computational kernel of the proposed model is developed on the basis of support vector machine (SVM). Additionally, self‐organizing map (SOM) is used to analyse observed data to extract data with specific properties, which are capable of providing valuable information for streamflow forecasting. After reprocessing, these extracted data and the observed data are used to construct the SVM‐based model. An application is conducted to clearly demonstrate the advantage of the proposed model. The comparison between the proposed model and the conventional SVM model, which is constructed without SOM, is performed. The results indicate that the proposed model is better performed than the conventional SVM model. Moreover, as regards the extreme events, the result shows that the proposed model reduces the forecasting error, especially the error of peak streamflow. It is confirmed that because of the use of data extracted by SOM, the improved forecasting performance is obtained. The proposed model, which can produce accurate forecasts, is expected to be useful to support flood warning systems. Copyright © 2012 John Wiley & Sons, Ltd.     

Variability patterns of the annual frequency and timing of low streamflow days across the United States and their linkage to regional and large‐scale climate

Low‐flow events can cause significant impacts to river ecosystems and water‐use sectors; as such, it is important to understand their variability and drivers. In this study, we characterise the variability and timing of annual total frequency of low‐streamflow days across a range of headwater streams within the continental United States. To quantify this, we use a metric that counts the annual number of low‐flow days below a given threshold, defined as the cumulative dry days occurrence (CDO). First, we identify three large clusters of stream gauge locations using a Partitioning Around Medoids (PAM) clustering algorithm. In terms of timing, results reveal that for most clusters, the majority of low‐streamflow days occur from the middle of summer until early fall, although several locations in Central and Western United States also experience low‐flow days in cold seasons. Further, we aim to identify the regional climate and larger scale drivers for these low‐streamflow days. Regionally, we find that precipitation deficits largely associate with low‐streamflow days in the Western United States, whereas within the Central and Eastern U.S. clusters, high temperature indicators are also linked to low‐streamflow days. In terms of larger scale, we examine sea surface temperature (SST) anomalies, finding that extreme dry years exhibit a high degree of co‐occurrence with different patterns of warmer SST anomalies across the Pacific and Northern Atlantic Oceans. The linkages identified with regional climate and SSTs offer promise towards regional prediction of changing conditions of low‐streamflow events.     

Sensitivity of streamflow simulation in the Delaware River Basin to forecasted land‐cover change for 2030 and 2060

In order to simulate the potential effect of forecasted land‐cover change on streamflow and water availability, there has to be confidence that the hydrologic model used is sensitive to small changes in land cover (<10%) and that this land‐cover change exceeds the inherent uncertainty in forecasted conditions. To investigate this, a 26‐year streamflow record was simulated for 33 basins (54–928 km²) in the Delaware River Basin using three dates of land cover: the 2011 National Land‐Cover Dataset (Homer, Fry, & Barnes, 2012 ), 2030 land‐cover conditions representing median values from 101 equally‐likely forecasts, and 2060 land‐cover conditions corresponding to the same iterations used to represent 2030. Streamflow was simulated using a process‐based hydrologic model that includes both pervious and impervious methods as parameterized by three land‐cover‐based hydrologic response units (HRUs)—forested, agricultural, and developed land. Small, but significant differences in streamflow magnitude, variability, and seasonality were seen among the three time periods—2011, 2030, and 2060. Temporal differences were discernible from the range of conditions simulated with 101 equally likely forecasts for 2030. Development was co‐located with the most frequent landscape components, as characterized by topographic wetness index, resulting in a change in hydrology for each HRU, highlighting that knowing the location of disturbance is key to understanding potential streamflow changes. These results show that streamflow simulation using regional calibration that incorporates land‐cover‐based HRUs can be sensitive to relatively small changes in land‐cover and that temporal trends resulting from land‐cover change can be isolated in order to evaluate other changes that might affect water resources.     

Separating the effects of changes in land management and climatic conditions on long‐term streamflow trends analyzed for a small catchment in the Loess Plateau region,NW China

As an integrated result of many driving factors, significant declines in streamflow were observed in many rivers of the Loess Plateau (NW China). This can aggravate the inherent severe water shortages and threatens the regional development. Therefore, it is urgent to develop adaptive measures to regulate the water yield to ensure water security. A key step for successful implementation of such measures is to separate the response of water yield to the main driving factors of land management and climate change. In this study, the variation of annual streamflow, precipitation, potential evapotranspiration, and climatic water balance in a small catchment in the Loess Plateau (near Pingliang, Gansu province) was examined for over five decades, although the relative contribution of changes in land management and climate on the streamflow reduction were estimated. A statistically significant decreasing trend of ‐1.14 mm y^‐1 in annual streamflow was detected. Furthermore, an abrupt streamflow reduction because of construction of terraces and check‐dams was identified around 1980. Remarkably, 74% of the total reduction in mean annual streamflow can be attributed to the soil conservation measures. Based on a literature review across the Loess Plateau, we found that the impact of changes in land management and climate on annual streamflow diminished with increasing catchment size. This means that there is a dependency on catchment size for the hydrological response to environmental change. This indicates that at least at the local scale well‐considered land management may help ensure the water security at the Loess Plateau. Copyright © 2012 John Wiley & Sons, Ltd.