Ensemble seasonal forecasts for the Horn of Africa: evaluation of the skill of high-resolution and convection-permitting simulations.

Paolo Mori

Institute of Physics and Meteorology

Challenges

Seasonal forecasts are of huge importance for economical and social reasons in the Horn of Africa: the predominance of rainfall-fed crop within this region requires long time strategies in order to mitigate droughts or other extreme events. Hence it is of high relevance for policy-makers to have accurate predictions concerning the development of the rainy periods.
June, July and August are the main rainy season in the Horn of Africa and consequently the most important season for agriculture. The inter-annual variation is quite large and dependent on many factors, numerical models prediction are not always reliable and accurate.
Currently available global seasonal forecasts e.g. from the European Center for Medium Range Weather Forecasts (ECMWF) allow the national weather forecast services (e.g. the National Meteorology Agency of Ethiopia (NMA)) to issue regional warnings. Seasonal forecasts by NMA however tend to miss both rainfall events and dry spells.

Objectives

• The first objective is to set up a high-resolution ensemble-based seasonal forecast system based, downscaling global seasonal ensemble forecasts.
• The model simulation data will be analyzed considering the evolution of water and energy cycles on the seasonal scale in the Horn of Africa. Here the focus is on the study of the land-atmosphere feedback and the related development of the atmospheric boundary layer in this tropical to subtropical region.
• Seasonal forecasts based on ensemble simulations add value by the provision of uncertainty measures. The third objective of this project is to gain insights on the fluctuations of the model using the statistical nature of an ensemble-based forecast.

Expected results

The goal of the project is to produce a one-year forecast over a domain including the Horn of Africa with a resolution of 3 km, allowing a detailed representation of orography, land surface and local atmospheric phenomena, such as convection.
This is supposed to improve the understanding of the phenomena governing the rainy periods in the different regions of the Horn of Africa, giving useful insights for development of early warning and operational seasonal forecasts.
The analysis of the variance of the ensemble members will give useful information on the uncertainty of seasonal forecasts, and as a consequence on the predictability of extreme events in tropical areas with a lead time of a few months.

Aims

The results of the new ensemble-based forecast with WRF-NOAHMP will be compared to already existing coarser resolution prediction in order to estimate whether the forecast skill improves, why, and what can be done to improve on the actual limitations.
The added value of the high-resolution grid will be studied. The impact of different factors involved will be taken into account: detailed topography, dynamically resolved convection, land-atmosphere feedback, planetary boundary layer evolution.
The advantages of the high-resolution set up will be discussed, considering, among possible applications, agricultural, food security and safety purposes.

Methods

Numerical atmospheric WRF and land surface NOAHMP models are used in order to downscale the ECMWF global seasonal forecast.
A 15-km resolution domain is forced at the boundaries using the global forecast. A 3-km domain is nested within the coarser domain. This resolution is called “convection permitting” because deep convection can be resolved explicitly, therefore no cumulus scheme is necessary.
A multi-parametrization ensemble consisting of 20 members + control is implemented.
The simulation starts the 1st of April and ends in September. The first two months allow the model to spin-up and the analysis will focus on the period from June to September.
The model will be compared against ECMWF analysis and observational datasets (i.e. TRMM, GPCP).
The most important variables considered in the study are surface temperature and rainfall. However, the interplay among many other variables (winds, surface fluxes, solar radiation) will be studied in order to understand the model’s behavior.
To evaluate different attributes of the forecast some standard deterministic and probabilistic skill score will be used: anomaly correlation coefficient, root mean square error, Brier skill score and ranked probabilistic skill score to name a few.

Publications

1. Mori Paolo, Budosa-Ware Markos, Schwitalla Thomas, Warrach-Sagi Kirsten, Wulfmeyer Volker, 2020.  Downscaling of seasonal ensemble forecasts to the convection-permitting scale over the Horn of Africa using the WRF model. International Journal of Climatology.