Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Acta Geophysica

1 2023

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: TIGGE

Sortuj według:

Ogranicz wyniki do:

Comparison of different ensemble precipitation forecast system evaluation, integration and hydrological applications

Tang Yehai, Wu Qiang, Soomro Shan-e-hyder, Li Xiang, Sun Yue, Hu Caihong

Acta Geophysica

2023

Vol. 71, no. 1

405--421

The examination and integration of numerical forecast products are essential for using and developing numerical forecasts and hydrological forecasts. In this paper, the control forecast products from 2010 to 2014 of four model data (China Meteorological Administration (CMA), the National Centers for Environmental Prediction (NCEP), the European Centre for Medium-Range Weather Forecasts (ECMWF), and the United Kingdom Meteorological Office (UKMO)) from The Interactive Grand Global Ensemble (TIGGE) data center were evaluated comprehensively. On this basis, a study of runoff forecasting based on multi-model (multiple regression (MR), random forest (RF), and convolutional neural network-gradient boosting decision tree (CNN-GBDT)) precipitation integration is carried out. The results show that the CMA model performs the worst, while the other models have their advantages and disadvantages in different evaluation indexes. Compared with the single-index optimal model, CMA model had a higher root-mean-square error (RMSE) of 18.4%, and a lower determination coefficient (R2 ) of 14.7%, respectively. The integration of multiple numerical forecast information is better than that of a single model, and CNN-GBDT method is superior to the multiple regression method and random forest method in improving the precision of rainfall forecast. Compared with the original model, the RMSE decreases by 13.1 ~27.9%, PO decreases to 0.538 at heavy rainfall, and the R2 increases by 4~15.2%, but the degree of improvement decreases gradually with the increase in rainfall order. The method of multi-model ensemble rainfall forecasting based on a machine learning model is feasible and can improve the accuracy of short-term rainfall forecasting. The runoff forecast based on multi-model precipitation integration has been improved, and NSE increases from 0.88 to 0.935, but there is still great uncertainty about food peaks during the food season.