Seasonal Forecasting
- Use and interpretation -

Predictability and climate model

It has been recognized that the predictability of successive instantaneous state of the atmosphere is limited and very little skills exist in forecasting weather phenomenon beyond 10-14 days. Nevertheless, there is some skill, though far from perfect, in predicting anomalies in the seasonal average of the weather. The evolution of the atmosphere is partly driven by the changes in external forcing. Much of the skill in predicting climate anomalies has its origin in the slowly changing conditions at the earth surface that can influence the climate, such as the sea surface temperature (SST) and soil wetness, etc.

The experimental regional climate model, or climate "crystal ball" as is frequently refered to in climate related literature nowadays. is only a tool in short range climate forecasting. Such a model, at its very best, can represent the large scale climate pattern and departures from normal with a modest level of accuracy (Gilman 1985; Livezey 1990; Barnston 1999). One should not infer the occurrence of a single event from seasonal forecast.

Experiments conducted by the Hong Kong Observatory using a regional climate model (RCM) showed that the model has some skills and potential to provide guidance for formulating seasonal forecast (Hui et. al 2002, and Hui and Shum 2005). The regional model was replaced by the Global-Regional Climate Model (G-RCM) suite with improved skill in 2007 (Lee et. al 2007).

Standard climate normal

A climate normal is defined, by convention, as the arithmetic mean of a climatological element computed over three consecutive decades (WMO 1989). The current period of standard normals adopted by HKO is 1971-2000.

Climate anomalies

At the seasonal scale, climate anomalies are commonly expressed in terms of different categories e.g. above normal (AN), near normal (NN) and below normal (BN). There is no standard classification scheme, though equal-probable categorization is preferred (WMO 2000). Some centres determine the boundaries of categories by ranking method. Others apply techniques involving fitting data to a statistical distribution (IRI 2003).

HKO defines the lowest 30 percentile as 'below normal' and the highest 30 percentile as 'above normal'. The middle 40% is being taken as 'near normal'. For those elements with good normality (i.e. close to normal distribution), the middle 40% roughly corresponds to the value within half a standard deviation (±0.5σ) of the climatological mean. For simplicity, 'near normal' in the HKO forecasting scheme (Chang and Yeung 2003) is defined as a value within the range of ±0.5 standard deviation of the mean.

Model climatology and anomaly forecast

It is widely recognized that one practical and important problem in dynamic climate modelling is the systematic tendency of a model to approach its own climate, giving rise to systematic error or bias. A standard practice to minimize the impact of the model systematic error is to calculate the anomaly from "model climate" rather than from the observed climate. The model climate is a function of initial calendar month, forecast period and location. Anomaly forecast is obtained by subtracting the model climate from direct model output.

It is assumed that the model climate is close to a normal distribution and is close to the climate in the real world. The value within ±0.5σ of the mean is defined as 'near normal (NN)', those greater than 0.5σ as 'above normal (AN)', and those less than -0.5σ as 'below normal (BN)'.

Common practice

Knowledge of the local climate is important in assessing model output, and indispensable in translating it into realistic statements about local and regional prospects (ECMWF 2003). In some centres, the final forecast is the result of a consolidation of many inputs, and the process is partly objective and partly subjective involving forecasters' judgment (IRI 2003).

To formulate the seasonal forecasts for Hong Kong, the forecasters analyze the output from G-RCM and dynamical forecasts from various centres. Forecast probabilities are assigned to each category i.e. "above normal", "near normal" or "below normal". For public dissemination, the most likely category is taken as the forecast.

References

Barnston, A.G., Y. He, and D.A. Unger, 2000: A Forecast Product that Maximizes Utility for State-of-the-Art Seasonal climate Prediction. Bulletin of the American Meteorological Society, Vol. 81, No. 6, pp. 1271-1279.

Chang, W L, and K H Yeung, 2003: Seasonal Forecasting for Hong Kong - A Pilot Study. Hong Kong Observatory Technical Note No. 104, Hong Kong Observatory.

ECMWF, 2003: Seasonal Forecast User Guide (http://www.ecmwf.int/products/forecasts/seasonal/documentation/index.html)

Gilman, D.L. 1985: Long-Range Forecasting: The Present and the Future. Bulletin of the American Meteorological Society, Vol. 66, No. 2, pp. 159-164.

Guttman, N.B., 1989: Statistical descriptors of climate. Bulletin of the American Meteorological Society, vol. 70, no. 6, pp. 602-607.

Hui, T.W., Chang, W.L., and Shum, K.Y.: 2002, Long-range forecasting for Hong Kong with ensembles from a regional model - some preliminary results. Second APCN Working Group Meeting, Seoul, Republic of Korea,11-13 June 2002. Hong Kong Observatory Reprint No., 480 (http://www.weather.gov.hk/publica/reprint/r480.pdf)

Hui, T.W., and Shum, K.Y., 2005: Prediction of Seasonal Rainfall in Hong Kong Using ECPC's Regional Climate Model. The Sixth International RSM Workshop, Palisades, New York, USA, 11-15 July 2005. Hong Kong Observatory Reprint No. 602 (http://www.weather.gov.hk/publica/reprint/r602.pdf)

IRI, 2003: The Science and Practice of Seasonal Climate Forecasting at the IRI (http://iri.columbia.edu/climate/forecast/tutorial2/)

Livezey, R.E. 1990: Variability of Skill of Long-Range Forecasts and Implications for their Use and Value. Bulletin of the American Meteorological Society, Vol. 71, No. 3, pp. 300-309.

Phelps, M.W., A. Kumar, and J.J. O'brien, 2004: Potential Predictability in the NCEP CPC Dynamical Seasonal Forecast System. Bulletin of the American Meteorological Society, vol. 17, no. 19, pp. 3775-3785.

WMO, 1989: Calculation of Monthly and Annual 30-Year Standard Normals, WCDP-No. 10, WMOTD/No. 341, Geneva: World Meteorological Organization.

WMO, 2000: Draft Standard Verification System (SVS) for Long-range Forecasts (LRF) (http://www.wmo.ch/pages/prog/www/DPS/SVS-for-LRF.html)

Lee, S. M., Yeung K. H. and Yeung K. K., 2007: Adaptation of Global and Regional Spectral Model for Seasonal Forecasting. The 21^st Guangdong-Hong Kong-Macao Seminar on Meteorological Science and Technology, Hong Kong, 24-26 January 2007. Hong Kong Observatory Reprint No. 689 (http://www.weather.gov.hk/publica/reprint/r689.pdf, in Chinese only).