oresee Rainstorm over the Lower Reaches of Yangtze River Valley in Summer 40 to 50 Days in Advance 

Breakthrough Made in the Study on Extended-range Weather Forecast Method 

http://www.wokeji.com/jbsj/seb/201410/t20141009_833590.shtml

(Technology Daily, by Ma Aiping) Major breakthroughs have been made recently in the study on new extended-range forecast method, part of the project Study of Interaction between SCGT and East Asian ISO in Summer and Its Application in Extended-range Forecast of Heavy Rainfall over the Lower Reaches of Yangtze River Valley supported by the Natural Science Foundation of China (41175082) and chaired by Yang Qiuming, a senior engineer in Jiangsu Meteorological Science Institute. The study found global atmospheric 20-30-day oscillation is the overriding factor affecting weather and climate over areas extratropics of both hemispheres. Another major achievement is the lead time of 20-30-day rainfall low frequency components is extended to 40-50 days. 

    By analyzing quantity of observational data, researchers reveal that impacts of intraseasonal oscillation (ISO) on heavy rainfall over the lower reaches of Yangtze River valley (LYRV) are selective in time scale, only 20-30-day ISO intensity has the most significant positive correlation with the heavy precipitation over the LYRV. This 20-30-day ISO activity is global, changes on a quasi-two-year cycle and is independent of the interannual variation of the classic East Asian Meiyu. We have found global atmospheric 20-30-day oscillation is the overriding factor affecting weather and climate over areas extratropics of both hemispheres; and there is a strong interaction between heavy rainfall over the LYRV during May-August and the southern circumglobal teleconnection wave train (SCGT). 

     In the mean time, a multivariable lagged regressive model (MLR), a multivariable lagged regressive model/ principal component - autoregressive model (MLR/PC-CAR) with innovative significance and an extended complex number autoregressive model (ECAR) with complex number autoregressive model (CAR) built with an extended complex number matrix (ECM) composed of main low frequency sequences are set up for 20-30-day ISO with the dynamic data-driven modeling method, which are applied to the forecasting of low frequency component of rainfall over the LYRV, achieving very good results in hindcast tests made with data of many years when the 20-30-day ISO is intense and extending the lead time of 20-30-day rainfall low frequency component to 40-50 days. Partial hindcast and forecast results are released on the website of the research group (  http://www.lcjrerf30.org/english/), according to Yang Qiuming. 

     In addition, the ECAR modeling method of ECM constitution in ECAR model also provides a new description to the show of the dynamic process of internal component interaction in the climate system. 

     It is learned that the achievement of this team is mainly used in the 10-30-day extended-range forecast of summer rainstorm (heavy precipitation) over the LYRV. Global circulation data major low frequency model and forecasting methods based on data-driven modeling will be adopted to extend the lead time of 20-30-day rainfall low frequency component over the LYRV from 20 days to 40-50 days. 

     Yang also said, extended-range weather forecast is a 10-30-day forecast between weather forecast and climate prediction. It needs to combine initial meteorological conditions with influencing factors like ocean, atmosphere and climate. Observational data are complex, comprehensive and global. These scientific big data reflect and represent complex natural phenomena and relations, with a high degree of data correlation and multiple data attributes. Therefore, the forecasting process is very complicated. 

      Single classical data analysis methods haven’t been fully competent data analysis already. By use of cross-over and integrated multiple data analysis methods and technologies, also through part valid data extracted from massive data, we can obtain more comprehensive low frequency variation information than by using traditional sampling analysis. We can see the warning and insight of extreme weather messages brought by such small data as a kind of     new extended-range weather variation signal. 

     The forecast method in this project is unique because it can identify the oscillation model of a specific time scale among the observational data and can set up a simplified complex number statistics dynamic forecast model changing with time. It is an entirely data-driven forecast model needing no pre-programmed predetermined rules and nearly no theoretical support. 

      Complex ultra-high-dimensional data analysis, filtering of noise, reduction of system complexity, transformation of big data sets into small ones, and reasonable identification and extraction of main simplified dynamic process affecting regional extreme weather from massive observation data can reveal fractional variation patterns of a few major low frequency components in low-dimensional space, said Yang. Then by transforming low frequency components in real space into the complex number space through Fourier transform, we can find new complex time-lag related variation patterns between main low frequency principal components of global circulation and rainfall low frequency components; describe the nonlinear variation information of climate system’s principal components in low-dimensional space in a more comprehensive way; construct a simple model with higher forecast ability and forecast their dynamic variation process through temporal coupling of different forecast models, which reflects the interaction between scientific big data and small data. 

     Compared with the popular international extended-range forecast methods, this method based on big data is more prominent in forecasting accuracy and extended lead time. It can be said that the combination of extended-range forecast theory with practical application in real time has opened a new way in the dynamic data-driven modeling based extended-range forecast study. (2014.10.9)