Streamflow Forecasts for the Yangtze River at the Three Gorges Dam, China
Upmanu Lall (Professor, EEE)
Hyun-Han Kwon (Postdoctoral Research Scientist, EEE)
(Joe) Kaiqin Xu (Research Scientist, National Institute for Environmental Studies, Japan)
Casey Brown (Research Scientist, International Research Institue for Climate and Society)
1. Background and Research Objective
Climate-based forecasts of hydrologic variables benefit water resources decision making, such as reservoir release decisions. The Yangtze River has experienced severe floods in the course of history, and in the 20th century notable events occurred in 1924, 1926, 1931, 1935, 1948, 1949, 1954, 1983, 1991, 1996, 1998, and 1999. The great flood of the summer of 1998 was an extremely serious basin-wide disaster and was one of the worst natural disasters on record (3,700 deaths, 223 million people displaced, and property damages up to $30 billion) in the Yangtze River basin. There appears to be a link between major floods and warm ENSO events (El Nino). In the following years, each of which has been described as an El Nino year, major floods were noted: 1924,1926, 1931, 1954, 1991, and 1998. There are many studies focused on climate scale dynamics that characterize rainfall in the Yangtze basin. Given significant water constraints in China, a strategy to more efficiently manage large water projects using seasonal to interannual water supply forecasts is critical. In this study we attempt to identify major modes of variability in global climate variables that may be used to produce seasonal forecasts of Yangtze River streamflow. Such forecasts may aid management of Yangtze River flows in the current era of new infrastructure development and the fast growing demand for water.
2. Methodology
2.1 Study Site
Yichang hydrological station (YHS, 111.28°E, 30.70°N, figure 1) was selected as the study site. YHS has streamflow records dating back to 1882. The famous Three Gorges Dam (TGD) is only about 40 km upstream. With the completion of the TGD, future measurements at YHS will represent a significantly altered hydrologic system.
Fig. 1 Location of Yichang Hydrological Station and TGD in the Yangtze River
2.2 Approach Used for Predicting Summer Monsoon Streamflow with Climate Indices
The approach can be summarized in figure 2. Simple methods of statistical analysis were used to describe the relationship between summer (JJA) streamflow and preceding climate variables and to evaluate predictive ability. Linear correlation maps were constructed using March-April-May (MAM) values (3-month ahead) to identify regions of that exhibit teleconnections with streamflow at YHS. The indices were formed in two ways, using a spatial average over the areas of interest and by taking the first EOF of the time series for the area of interest.
Fig.2 Flowchart of Method used for Predicting Summer Monsoon Streamflow withClimate Indices
3. Results and Discussion
3.1 Characteristics of Streamflow at YHS and Influence of Climate Variables
The annual hydrograph shows the seasonality of streamflow, with monsoon runoff (June-July-August) dominating (figure 3). Linear correlation maps between selected climate variables averaged over MAM and YHS streamflow for JJA for the years 1979 to 2003 are depicted in figure 4. Table 1 lists the correlation coefficients of the spatially averaged MAM indices with JJA runoff.
Fig.3 Annual Cycle of Streamflow at YHS
Fig.4 Correlation Analysis between JJA Streamflow and March-April-May (MAM)a) SST; b) OLR; c) PREC and d) SLP
The area of negative correlation of JJA streamflow with SSTs located near the equator, which we designate SST1, likely corresponds to an ENSO signal, specifically the horseshoe area of negative SST anomalies that lies to the west of the warm anomalies in the eastern Pacific. During ENSO warm events, East Asian monsoon circulation is weakened, causing less rainfall in northern China and leading to increased rainfall, streamflow and flooding in south central China as the subtropical high remains to the south.The area of positive SST correlation in the east Indian-western Pacific warm pool area, which we designate SST2, corresponds to increased rainfall (and thus streamflow) during non-ENSO years due to warm SSTs in that area.
3.2 Prediction of JJA Runoff with Climate Indices
The results of the regression analysis are shown in table 3. The time series of observed JJA runoff and the values predicted with the quadratic model with two SST- EOF predictors are shown in figure 5.
Fig.5 Timeseries of Model Prediction Results and Observed JJA Streamflow using Quadratic Regression (open circle) with MAM Predictors
4. Conclusions
Analysis of summer monsoon streamflow at Yichang hydrological station (YHS) for the period 1882–2003 revealed
- teleconnections with streamflow at YHS were identified with linear correlation maps of climate fields
- regions in the Eastern Indian and Western Pacific oceans influence YHS streamflow, which is consistent with previous studies of regional rainfall
- a regression model for predicting streamflow using SST indices results in an R2greater than 0.6 (robust under leave one out cross validation)
Skillful prediction can improve water management in the Yangtze River Basin, possibly the Three Gorges Dam and future South to North water transfer project. Our current work evaluates this potential.
