Effects of resolution, model formulation and topography representation in the South African Regional Eta model

1 INTRODUCTION

Unexpected rainfall occurred over Pretoria on August 30, 1997. The then operational Eta model failed to predict the correct positioning of the precipitation, most probably as a result of low resolution. Since it was observed that the model diagnosed well in other respects, it was suggested that a slight change to the Eta topography might improve the model's handling of the rainfall.

It was hypothesised that orographic upliftment caused the necessary trigger to produce the light rain which occurred where none was predicted by the model. There seemed to be no dynamical error in the Eta fields which could suggest why it did not predict correctly. The only suggestion at the time was that the necessary trigger was not in the right place - i.e., the topography.

This case became the focus of a study in which, firstly, the Eta model topography was manually modified to give a better resolution of the topography over the northern interior of the country. Secondly, the effects of improvements in model formulation and resolution were observed. The 80-km/38-level regional Eta model operational at the Weather Bureau until August 1998 was used in the topography experiments, and served as control in the resolution trials. In the latter the recently installed upgrade model was used, incorporating the many improvements in the assimilation and forecasting systems introduced by the developers at the National Centers for Environmental Prediction, USA, at various horizontal (80, 48 and 29 km) and vertical (38 and 50 levels) resolutions, and on various domains.

2 EXPERIMENTS WITH MODEL TOPOGRAPHY

The "step-mountain" formulation of model topography which is a feature of the Eta model, requires adjustment of a given fine-resolution terrain field to the heights of the reference coordinate surfaces of the model. In the first experiment the vertical distribution of model surfaces in the vertical was taken to be that of the upgrade model at the same resolution. This experiment was preliminary to a test where the eta levels were manually altered to be of benefit to higher lying terrain. In the third experiment the adjusted model terrain was manually modified. The effects on the predicted precipitation turned out to be rather small, although other fields were significantly affected. These effects and the connection to the rainfall will be discussed in detail in the paper.

3 EFFECT OF MODEL FORMULATION AND RESOLUTION

The various upgrade models were then run for the 97083000 case with no modifications to the standard treatment of the topography or the eta levels. It was immediately seen that the 24-hour prediction from the upgrade model correctly produced rain over Pretoria. Thus the extensive NCEP enhancements comprising the upgrade have proved very beneficial, in this one case at least. Detailed evaluation of the effects of vertical and horizontal resolution and domain size was carried out. It is to be noted that the model topographies change as the resolution changes, so that terrain influences were still playing a part in these runs. All of these evaluations will be presented.