SHORT TERM WEATHER FORECASTING TECHNIQUES FOR HEAVY RAINFALL
Report No 1011/1/02
November 2002

EXECUTIVE SUMMARY

Abstract

The project set out to develop a computer-based quantitative precipitation forecast system for South Africa. The aims also included research on past heavy rainfall events, evaluation of their predictability and the classification of associated systems according to their dynamical characteristics. The project succeeded in developing systems that delineate heavy rainfall from tropical and coastal zone baroclinic systems. A further product, an operational verification system to evaluate the quality of Eta rainfall prognoses, became operational at the SA WS in September 2001.

Introduction

Serious floods happen over South Africa. Recent examples are the devastating floods over the Northern Province of South Africa during February 2000, the famous Kwazulu-Natal floods of September 1987, the Free State floods of February 1988 and 1996, to name but a few. These caused loss of life and damage to infrastructure easily amounting to R 1000 million. The tools available to the weather forecaster are at present still unable to provide accurately predicted rainfall amounts. Pinpointing the exact area where the heavy rainfall will be concentrated is very difficult. This is so in spite of all the recent advances in Numerical Weather Prediction (NWP) models.

The atmospheric flow field plus the many dynamical parameters derived from NWP models are complex and it is beyond the human mind to integrate them all. The project set out to test these parameters against the occurrence of heavy rainfall. In this way those combinations of dynamical parameters consistently associated with heavy rainfall over South Africa could be identified. The mechanisms responsible for heavy rainfall could subsequently be determined and products developed to delineate areas where heavy rainfall occurs.

Considering the operational environment under which weather forecasts are prepared it was considered imperative to produce a product that is easily accessible to the forecaster. The final product must also be applicable in a flood management environment.

Aims of the Project

The aims of the FORFLOOD Project detailed in the original contract, but slightly modified during the Steering Committee meetings, can be subdivided into the main objective and secondary goals:

The main objective of this project was to research the atmospherical circulation in order to:

• Make available a Quantitative Precipitation Forecast (QPF) for input into an envisaged DWAF/Weather Service River Flow Project by improving forecasting techniques and developing a new forecasting system for the period 6 to 48-hours ahead.

In order to achieve this main goal, secondary research aims of the project were listed:

• Devise a method to classify weather systems into Baroclinic, Tropical or Tropical- Temperate using synoptic data.
• Classify all past heavy rainfall events in South Africa as Baroclinic, Tropical or Tropical- Temperate.
• Evaluate all past forecasts issued for these heavy rainfall events.

Major Results and Achievements

For the first time, in southern Africa, a system is available which reliably delineates areas where heavy rainfall can be expected with a good degree of accuracy. This system is based on meteorological dynamics but depends on the accuracy of the data fields generated by NWP models. The analysis and integration performed by the heavy rainfall identification system is easily accomplished in the short time available to the meteorologist in the operational environment and empowers him/her to make predictions, which would otherwise be beyond the capacity of any meteorologist.

The heavy rainfall identification system is made of three independent sub-systems:

The first of these is MITS (model for the identification of tropical weather systems), which determines whether a weather system is tropical. It also delineates the area where rainfall from convective weather systems will develop.

The second, THERIS (tropical heavy rainfall identification system) is a powerful, unique system for isolating heavy rainfall forced by continental tropical weather systems. THERIS runs on both digital forecast data from numerical models and archived analysis data. THERIS should also perform well north of South Africa.

BARCOZ (baroclinic rainfall system for the coastal zone) isolates areas where heavy rainfall can be expected from typical baroclinic weather systems. Most important amongst these were the cut-off low-pressure systems. The research revealed new parameters: Convergence of the gradient of the moisture flux by the southeastern (meridional) component of the wind. These new parameters are referred to as SEC (MC). When SEC was applied with upward motion under specified conditions, BARCOZ was able to isolate heavy rain from baroclinic systems.

The results of the heavy rainfall identification systems are assimilated by a Geographical Information System (GIS) and are available for publication on the World Wide Web.

During the development of MITS, THERIS and BARCOZ it became clear that synoptic scale data do not have the resolution to resolve some of the severe and important smaller scale rainfall events. The systems were very successful in isolating areas of heavy rain but were unable to explain severe events such as the Port Elizabeth flood in 1968 or the very heavy rainfall along the escarpment.

Ongoing research on the geographical and spatial characteristics of widespread rain indicated that the frequency and magnitude of these systems changed, probably during the long period with above normal rainfall of the 1970's. The research also indicated that the relationship between the frequency of occurrence of these systems and the southern oscillation index (SOI) is tenuous at best.

In the quest to evaluate past forecasts of heavy rainfall the project team failed to get a record of past rainfall forecasts from the SAWS, apart from a short period of some 4 years. The historic rainfall forecasts were unavailable to the project team, lost or irretrievable. The forecasts, which were available, formed the basis of research on the reliability of rainfall forecasts over the Vaal Dam catchment. Research to test the accuracy of SAWS rainfall forecasts over the Vaal Dam catchment indicated that forecasters tended to err on the conservative side. The project also evaluated the Eta model rainfall prognoses statistically against rainfall figures computed from radar reflectivity data as well as conventional rainfall gauge data. Verifying the Eta model predictions against gauge data became operational at the SAWS in September 2001. Over the 6-month operational period results show that the model tends to over-predict rainfall. The model fared well with identifying rainfall areas but the rainfall totals were not simulated equally well. It is . recommended that the verification method be expanded to include all available real-time and countrywide rainfall data.

The project has made a significant contribution to the understanding and recognition of situations, which give rise to heavy rainfall over South Africa and to the prediction of heavy rainfall events. In doing so the major objectives of the project have been met.

Recommendations

The systems developed are powerful and provide tools, to the meteorologist, not available before. Urgent operational implementation is recommended parallel with an evaluation programme. The operational system is very user friendly but the project team can assist with on site training of weather forecasters. The systems, especially MITS and THERIS should work equally well in countries north of South Africa. The WRC should consider making these systems available to the weather services of these countries. The SAWS has extended a formal invitation to the project team to install the systems: MITS, THERIS and BARCOZ on their computers.

Unfortunately the preparation of a proper quantitative precipitation forecast proved impossible. Uncertainty about the accuracy of the NWP rainfall prevented the creation of a first guess rainfall field. These NWP generated rainfall fields should serve as a basis in the creation of a QPF. It is recommended that the SAWS upgrade their models to small resolution regional models.

The project determined that the high South African topography and especially the escarpment modify rainfall significantly. Further research to address this factor is recommended.

A possible side benefit, of the Vaal Dam catchment rainfall forecast verification might be that the SAWS will again look at the way rainfall forecasts are worded. It is also recommended that the Eta precipitation verification method, developed during this project, be expanded to include all available rainfall data.

Conclusions

This project was very cost effective which delivered an important tool for forecasting heavy rain. It provided the opportunity for capacity building as well as the training of forecasters both local' and abroad. Several research papers were published or are in the process of publication. The University of Pretoria awarded an MSc (Meteorology) degree, to the Principle Researcher, for research during this project. Several previously disadvantaged students participated and benefited from this project.

The support of the Water Research Commission made the research possible. The SAWS provided data, guidance and allowed some of their staff to participate in the project.