Search

Under the context of global warming, compound dry and hot events (CDHEs) will increase and bring serious losses to society and the economy. The projection of CDHEs is of great significance for policy-making and risk assessment. In this paper, two large ensemble simulations, CanESM2-LE and CESM-LE, are used to estimate the risk of extreme CDHEs under different warming scenarios in China. First, the biases of the model in the simulation of the temperature and precipitation over the China region are corrected, and the index of CDHEs is established based on a copula function. The results show that extreme CDHEs will occur more often in China with the increase in global warming and the more severe extreme CDHEs are, the greater the risk will be in the future with higher uncertainties. Events that would be attained once every 50 and 100 years in the current climate from CESM-LE (CAanESM2-LE) will be 1.2/1.6 (1.1/1.5) times and 1.3/2.3 (1.5/2.0) times more likely to occur in a 1.5°C/2.0°C warmer climate, respectively. Northwestern China will experience the greatest increase in the risk of extreme CDHEs. Extreme CDHEs expected once every 100 years in the current period over NW China are expected to occur approximately every 5 and 4 years under a 4.0°C warmer world in CanESM2-LE and CESM-LE, respectively. [ABSTRACT FROM AUTHOR]

Obtaining high-quality quantitative precipitation forecasts is a key precondition for hydrological forecast systems. Due to multisource uncertainties (e.g., initial conditions, model structures, and parameters), raw forecasts are subject to systematic biases; hence, statistical postprocessing is often required to reduce these errors before the forecasts can proceed to hydrological applications. Machine learning (ML) algorithms are canonical statistical models, and they are diverse in type and variation. It is important to verify and compare their performance in the same scenario (e.g., precipitation postprocessing). In this paper, we conduct a large-scale comparison study for the major ML models with diverse model structures and regularization strategies as postprocessors for improving the quality of precipitation forecasts. Specifically, we compare the efficiency and effectiveness of 21 ML algorithms on solving this task. Daily reforecast precipitation with lead times up to 8 days from the Global Ensemble Forecast System and corresponding observations are employed to determine the usability of different models in the Yalong River basin in China. The performance of each model is validated by a group of carefully designed experiments and statistical metrics. The results reveal that improvements in model structures are more effective than regularization strategies. Among these algorithms, the optimized extra-trees regressor exhibits the best performance, effectively reducing overestimation and achieving the best skill in forecasting precipitation. Eleven ensemble members and a 3-day forward-rolling time window can be used as predictors to obtain the best model performance. The systematic experiments and findings also offer useful guidelines for other related studies. [ABSTRACT FROM AUTHOR]

ABSTRACT The biases in the Global Circulation Models ( GCMs) are crucial for understanding future climate changes. Currently, most bias correction methodologies suffer from the assumption that model bias is stationary. This paper provides a non-stationary bias correction model, termed residual-based bagging tree ( RBT) model, to reduce simulation biases and to quantify the contributions of single models. Specifically, the proposed model estimates the residuals between individual models and observations, and takes the differences between observations and the ensemble mean into consideration during the model training process. A case study is conducted for 10 major river basins in Mainland China during different seasons. Results show that the proposed model is capable of providing accurate and stable predictions while including the non-stationarities into the modelling framework. Significant reductions in both bias and root mean squared error are achieved with the proposed RBT model, especially for the central and western parts of China. The proposed RBT model has consistently better performance in reducing biases when compared with the raw ensemble mean, the ensemble mean with simple additive bias correction, and the single best model for different seasons. Furthermore, the contribution of each single GCM in reducing the overall bias is quantified. The single model importance varies between 3.1% and 7.2%. For different future scenarios ( RCP 2.6, RCP 4.5, and RCP 8.5), the results from RBT model suggest temperature increases of 1.44, 2.59, and 4.71 °C by the end of the century, respectively, when compared with the average temperature during 1970-1999. [ABSTRACT FROM AUTHOR]

• A Bernoulli-Gamma-Gaussian model is built to calibrate S2S precipitation forecasts. • The model is effective in correcting bias and improving forecast reliability and skill. • A 15-day sliding window is suitable for pooling samples for S2S forecast calibration. Subseasonal to seasonal (S2S) precipitation forecasts can fill the gap between weather and seasonal forecasts. While raw forecasts from S2S models are informative, calibration is necessary to correct systematic biases and quantify forecast uncertainty to facilitate applications of S2S forecasts. This paper develops a seven-parameter Bernoulli-Gamma-Gaussian model to calibrate S2S precipitation forecasts. The Bernoulli distribution characterizes the occurrence of zero versus non-zero values of precipitation, the Gamma distribution accounts for the distribution of non-zero precipitation amounts and the bivariate Gaussian distribution formulates the relationship between raw forecasts and observations. A case study of the East River catchment in South China is devised for the S2S forecasts provided by the European Centre for Medium-Range Weather Forecasts (ECWMF). The Bernoulli-Gamma-Gaussian model is used to calibrate raw forecasts at the daily time step and the Schaake Shuffle is used to instill the temporal patterns to form ensemble time series forecasts. The results show that raw S2S forecasts generally suffer from considerable forecast bias and unreliable ensemble spread and can exhibit negative forecast skill. Compared to the quantile mapping, the Bernoulli-Gamma-Gaussian model with the Schaake Shuffle is effective in correcting biases, improving reliability and yielding forecast skill for daily and accumulated precipitation. Furthermore, a sliding window of 15 days tends to be suitable to pool samples to deal with sampling variability and precipitation seasonality for the purpose of calibration of S2S forecasts. Overall, the Bernoulli-Gamma-Gaussian model can serve as an effective tool to calibrate S2S precipitation forecasts for hydrological modeling and water resources management. [ABSTRACT FROM AUTHOR]

Scarcity or unavailability of precipitation observation creates difficulties in hydrologic modeling of mountainous sections of the arid region of northwest China (34°-50°N, 72°-107°E). Tropical Rainfall Measuring Mission (TRMM) precipitation products may be a potential substitute, but they should be evaluated and corrected with ground observation data before application. In this paper, two TRMM Multisatellite Precipitation Analysis (TMPA) precipitation products were evaluated by gauge observations, using indices such as frequency bias index, probability of detection, false alarm ratio, relative mean bias, Nash-Sutcliffe efficiency, and correlation coefficient. Terrain variables were extracted from a digital elevation model, and their rotated principal components were determined to establish a stepwise regression model to adjust TMPA precipitation. Additionally, a back-propagation (BP) neural network was established to correct TMPA precipitation. The results showed that TMPA had an unsatisfactory detection ability in the study area for both precipitation occurrence and amount. TMPA precipitation corrected by a stepwise regression method showed some improvement, but only the results for TRMM 3B43 on a subregion scale were acceptable. The BP neural network method showed better results than the stepwise regression method, and both TRMM 3B42 and TRMM 3B43 corrected by the former method on a subregion scale could be acceptable. Both methods were spatial-scale dependent and showed better results on a subregion scale than on a larger scale. [ABSTRACT FROM AUTHOR]