2015 Chinese New Year resolutions – HEPEX’s future in China

Contributed by Yi He, Qingyun Duan, Desmond Manful and Florian Pappenberger

Let’s celebrate!

Let’s get ready, because the goats are coming. According to the Chinese lunar calendar, the Year of the Wooden Goat starts on 19th of February 2015. Very much like the science of ensemble forecasting, the Chinese zodiac (Figure 1) is an ensemble of methods that can be utilized to predict people’s personality, fate and destiny.

According to the Chinese New Year, the Year of the Sheep starts Feb. 19. Also called the Chinese Spring Festival or Lunar New Year, it has been celebrated for more than 4,000 years. This year it is celebrated from Feb. 19 through 25, with Feb. 19 as the key day. It's the grandest and the most important traditional festival for Chinese people. Whole families get together, similar to Christmas Day for Westerners, and commemorate the advent of spring.
Figure 1: The Chinese zodiac and year of Goat ( by Eileen Wacker For The Telegraph ). Yellow, blue, green, red and brown covering years in the diagram represent Metal, Water, Wood, Fire and Earth respectively. They are known in Chinese culture as the Five Elements that interact with each other.

Let us, the atmospheric scientists, the HEPEX community, and people affected by ensemble forecasts in one way or the other, celebrate for the great success China has seen in the past several years in undertaking new research and development toward the application of ensemble forecasting.

Ensemble forecasting technology has become increasingly popular and been playing an important role in China since about a decade ago. For instance, the Australia Bureau of Meteorology (BOM) provided the STEPS (Short Term Ensemble Prediction System) to generate an ensemble of rainfall forecasts for the 2010 Shanghai World Expo that lasted for 6 months. These forecasts were also used of drive hydrological models to compute the probability of river levels exceeding certain critical thresholds (Dai and Tao). Chinese meteorological forecasters have also successfully provided much needed weather services using ensemble numerical weather forecasts for high impacts events such as the Sichuan and Gansu Earthquake Relief in 2013, launch and return to earth of China’s Shenzhou-10 space mission, a number of summer heavy rainfall periods in southern China and so on.

Most of the weather services provided so far have been based on ensemble forecasts that are limited to a single numerical weather model. The massive TIGGE database is yet to be well explored in China, although this will certainly change in the future. China will also see an increasing number of local meteorological centres to use these new ensemble forecasts and provide relevant weather services.

Unfortunately, lack of data sharing and cooperation between the meteorological and hydrological organisations have to some extent prevented or delayed more useful applications and services in hydrology and catchment water resources that hydrological organisations may be capable of providing (as discussed in a previous post). Nevertheless, this has never prevented scientists in both atmospheric and hydrological sciences in China from collaborating in conducting research. Let’s hope this can aspire and promote more effective collaboration in the near future amongst different Chinese authorities such as the CMA and MWR – our first New Year resolution!

HEPEX workshop and literature in China

HEPEX conducted a successful and well-organized meeting in Beijing in Oct 2012 and enjoyed a large number of scientific presentations and energetic discussions related to ensemble prediction. It has become a unique milestone marking a new era in the rapid development of hydrological ensemble forecasting in China (see also the other post in our Chinese New Year edition for more details).

Two Chinese scientists have written a report (in Chinese) analysing the research output in the field of ensemble forecasting. They searched literature from the ISI Web of Science (SCI-E) database and the China Academic Journal Network Publishing Database (CAJD) using four key words, namely stochastic dynamic forecast, stochastic dynamic predict, ensemble forecast, or ensemble predict. Figure 2 shows the number of peer reviewed articles according to this report. The first article that can be found in the SCI-E is “Stochastic dynamic prediction” by Edward S. Epstein published in the journal Tellus in 1969 (Volume 21, Issue 6, 739-759). In contrast, the first research article that can be found in the CAJD is “Forecasting technologies” published in Chinese in 1988 in the journal titled Meteorological Science and Technology. This 19-year lag indicates a significant delay in Chinese scientists embracing ensemble forecasting research.

Three development stages were reported by the two Chinese scientists. The first is between 1970s and 1980s when research focused on theoretical studies and numerical experiments. Only a handful articles were published during the first stage. Rapid improvement of computer technologies in the 1990s pushed the development of numerical weather forecasts into the second stage. Both NCEP and ECWMF established their numerical ensemble forecasting systems respectively and used them in operational forecasting. A slight increase in the number of research articles can be seen during this stage. Since the 21st century, fast acceleration of computing power and modelling capacity has shifted ensemble forecasting research to its third stage.

The annual number of research articles in the field of ensemble forecasting has increased sharply from 10 in the second stage to more than 150 since 2009. It is worth pointing out here that the graph in Figure 2 does not necessarily include articles in hydrological ensemble forecasting. Relevant articles authored by Chinese scientists included in the SCI-E database have made a breakthrough since 2008 to 10 per annum, and steadily increased since then. The number of ensemble forecasting relevant articles authored by Chinese scientists between 2012 and 2014 totals up to 85, which accounts for 15% of all relevant articles included in the SCI-E database in the same period.

Despite a late start, Chinese scientists have claimed a noticeable share in the research of ensemble forecasting and publications. Chinese scientists must continue to realize their potential to make more contribution in the future – our second New Year resolution!

Figure 2: Changes in the number of peer-reviewed articles in the field of ensemble forecasting according to the SCI-E and CAJD database . CAJD stands for China Academic Journal Network Publishing Database. SCI-E stands for ISI Web of Science

2015 Chinese New Year Resolution

It is said that if you make too many New Year resolutions, you will not be able to accomplish them. Three New Year resolutions are just about right. So, here we have our last chance.

As early as 1969, Epstein already stated in his paper, “Stochastic dynamic predictions have significantly smaller mean square errors as deterministic procedures, and also give specific information on the nature and extent of the uncertainty of the forecast. Also the range of time over which useful forecasts can be obtained is extended.” With more than two-decade rapid development worldwide and nearly 15 years of development in China, ensemble forecasts have yet to be utilised to their full potential. One of the main obstacles is their performance is often deemed to be too poor to provide ‘harmless’ operational forecasts. Additionally, forecasts issued or communicated to the end-users can often predict more of a disaster than the actual natural disaster itself, particularly as the lead times increase. False alarms do not only cost considerably in financial terms but also damage the reputation of forecasting institutions.

Emergency managers often have to make a binary decision whether or not an action should be taken. It is not so straightforward for decision makers to utilise ensemble forecasts in terms of probabilities compared to conventional deterministic forecasts wherein a binary decision can be made based on a single forecast, albeit with inevitable errors in the single forecast. “2.95 metre plus or minus 0.5 metre, how can I decide if the sluice gate should be shut or remain open? Why cannot you provide me with one single predicted water level?”, said by one of the hydrological engineers in China, is still echoing in the author’s mind. This is often the second ‘excuse’ for not using ensemble forecasts.

Readers can also refer to Demeritt et al. (2010) for more detailed discussion on challenges in communicating and using ensembles in operational flood forecasting. China is in great need of reliable forecasts to battle against its many natural disasters, especially the most frequent floods. While so much effort has been made to improve forecasts, disproportionally low effort has been made to provide services and communicate forecasts to the end users.

In the coming year, we really hope to see more people and effort dedicated to provide effective means of communication to transfer the promising technology to practice – our third New Year resolution!

Bye for now and See you in the Wooden Goat Year!


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.