The document discusses the development of early warning systems for floods and droughts in Africa based on the European models. It describes the European Flood Alert System (EFAS) and European Drought Observatory (EDO), which provide early warnings up to 15 days for floods and 1 month for droughts. The authors propose establishing a pan-African early warning system by applying the EFAS and EDO methodologies to African river basins, starting with pilot projects in East and Southern Africa. The system could provide lead time to authorities for flood prevention and drought response planning.
Pan-African Flood and Drought Early Warning System
1.
2.
3.
4.
5. Towards Probabilistic Forecasts: Using Ensembles: Meuse (Borgharen) 19-01-1995 / 28-01-1995 P Q ECMWF LISFLOOD 1km Low flood risk Medium flood risk High flood risk Extreme flood risk
6. EFAS thresholds compared to the real river cross section River cross section Flood plain EFAS Extreme Alert ~ 10-100+ year return period Critical Q Bankfull Q EFAS High Alert ~ 2-10 year return period ~ > bankful conditions
7. Real-time Weather Forecasts: DWD LM & GM & COSMO-LEPS ECMWF DET & EPS (2x69 runs per day ) Static European Datasets : -topography -land-use -river channel dimensions -geology Historic observed Meteo data: JRC MARS (station data from 1990 onwards) Q-Thresholds Q>Threshold yes Persistent yes Real-time processing, 2x a day Offline processing External alerts Initial conditions LISFLOOD 1-6-24 h (EPS) 1 2 3 5 Real-time processing, after decision 4 Real-time Observed Meteo Data: EU-FLOOD-GIS station data (~1300 stations across Europe) LISFLOOD 1-6-24 h LISFLOOD daily Theoretical background E uropean F lood A lert S ystem (EFAS) ! ! Early Flood Warning in Africa: The Potentials of the European Flood Alert System (EFAS) for African Basins Flood forecasts Flood forecasts Flood forecasts Flood forecasts Previous Flood forecasts
8. Number of alerts is increasing More hits than false alarms Hit-rate: 60-70%
13. Example: Vistula at Warsaw (PL) Peristent forecasts from 10 May 00:00 onwards
14.
15. European, MS, RB, … Authorities Communication on Drought and Water Scarcity Drought Management Plans EDO Map Server EDO System Setup VEGETATION STATE Meteorological Data Stations Fields Fore- casts European Data Layers LC/LU Soil DEM … RBs Hydrological Processes MONITORING & MODELLING Land Surface Processes Remote Sensing Data FAPAR NDWI Time Series Products Rainfall Anomalies Soil Moisture Anomalies Vegetation Vigour …
22. SPI forecasting using ECMWF monthly forecasts 3 – monthly SPI The precipitation forecast for the next month is added to the accumulated observed rainfall of the past two months, and then the 3-months forecasted SPI is calculated 2 - monthly cumulative precipitation map ECMWF monthly forecast average over 50 ensemble forecasts SPI calculation using historical time series forecasted 3 – monthly SPI EDO – Drought Forecast Products: SPI
23. observed 3 monthly SPI forecasted 3 monthly SPI Comparison of observed and forecasted SPI (August - 05) EDO – Drought Forecast Products: SPI
24. Next step: Probabilistic SPI forecasting (up to 1month)… 2 1 0 -1 -2 Observations Ensemble predictions EDO – Drought Forecast Products: SPI
25. EDO – Drought Forecast Products: SPI 1-month Probabilistic Forecast of SPI-3 January 2007 Probability that SPI-3 for the next month is “severe dry” or worse … still experimental! Forecasted Probability for SPI < -1.5 Observed SPI
26.
27.
28. Pilot test in East Africa: Juba & Shabelle Two pilot studies: Juba/Shabelle river basins Somalia - Ethiopia) Zambesi river basin (Southern Africa) 1977 flood 1981 flood hindcast Spring 1981 for Belet Weyne (Shabelle, Somalia)
29.
30.
Editor's Notes
As for EFAS, also AFAS should provide added value – in terms of early flood warning – to existing national systems – which can predict floods more precisely in the short run. Hence AFAS is not meant to replace any national operating system but to complement them and to help international aid organisations during crisis. u
Within the European Drought Observatory, a direct up- and downscaling will be possible. Information on droughts (e.g. drought indices) will be comparable across scales. All players produce drought information on their level of competence, with their data. EDO will facilitate the access, comparability, and harmonisation of information produced. JRC‘s contribution to EDO is the provision of novel, independent information on droughts at the European level, and the development of the EDO technical infrastructure.
fAPAR = fraction of Absorbed Photosynthetically Active Radiation; indicator of plant activity, vigour fAPAR algorithm has been produced at IES – GEM unit (Nadine Gobron), and will be applied quasi-operational in H07; production envisaged at ESA-ESRIN, Frascati fAPAR is a completely independent indicator of the response of plants to (water) stress Image Legend : light grey : low fAPAR, green : medium fAPAR, red : high fAPAR values Graph Legend : red = fAPAR, black = SPI-3; both drought indicators compare very well for major drought events in 1999 and 2005 ESA = European Space Agency ENVISAT = ESA’s environmental satellite MERIS = Medium Resolution Imaging Spectrometer, onboard ENVISAT
Latest results of a study to use ECMWF monthly ensemble forecasts for drought prediction. For the forecasting study, the 3-months SPI is computed from 2 months of observations and 1 month of forecasting data. A threshold value of SPI < -1.5 is applied (-1.5 < SPI < -2.0 corresponds to the drought level “severe dry”). As ECMWF provides 50 ensembles (realizations) of the monthly forecast with slightly changed initial conditions, the probability of exceedance of the chosen threshold (i.e. SPI values smaller/”drier” than -1.5) can be computed. All areas in the upper left image that are red have a probability > 80% that the SPI will be -1.5 (severe dry) or less. The lower right image shows the corresponding SPI-3 computed from observations for validation purposes. The correspondence of red/orange areas is very well visible, showing the value of using monthly forecasts. - Forecasts are produced by ECMWF on a 0.5 degree grid globally. - Based on these data DESERT calculates SPI globally (the shown images show a subset for Europe) - Be aware that ECMWF data are provided to us for research purposes only (not for operational processing – this would require different agreements/licenses) The selected case shows an example with good forecast results (01 2007). Of course there are better or worse results, depending on the area and season. Currently the comparison between forecasted SPI and observed SPI was made for a period of 36 month (2005 to 2007). The first preliminary results shows that SPI is better forecasted in winter and summer and low forecasts skills are observed in spring and autumn. This corresponds with long term meteorological forecasts skills, where better results are associated with the stable weather patterns (anticyclones). At the current state of the analysis we get good forecast results in around 60% of all cases . A little better in the dry areas of Europe (Turkey, Mediterranean). An in-depth analysis is on-going.
First of all, EFAS is the only probabilistic flood warning system that is especially designed for large-scale (transnational) river basins on a continental scale. Second, EFAS can cope with a limited amount of input data. The extent of the available data basis at the initial stage is in fact quite similar in Europe and Africa. Third, EFAS uses weather forecasts (deterministic and probabilistic) to create early flood forecast with increased lead times of up to 10 days (typical lead times of national systems: 2-3 days). This provides additional preparation time to national hydrological services which is important in terms of planning , coordinating and realizing effectively prevention, protection and mitigation measures. Another advantage of EFAS arises due to their clear, concise and unambiguous visualization and decision support products that enable an intuitively understanding of the results. This is surely of particular importance to Africa since they hold a large number of transnational river basins that are managed by different national institutions using different tongues, which exposes a great potential to create easily inadvertently misunderstandings between those. Lastly, the expert knowledge gained during the development of EFAS and the strong commitment of FAO SWALIM and ECMWF, supporting this enterprise with data and expert knowledge represents a great potential to the progression of AFAS.
Without Lag-Dere From the headwaters to the coastal delta it measures a length of 1,100 km (Juba) i.e. 1,700 km (Shabelle) and covers an area of 783,000 km², which is comparable to the Danube River Basin (801,463 km²; ICPDR 2008). The altitude of both rivers ranges from well over 3,000 m AMSL in the eastern Ethiopian Highlands where they originate to just above sea level in terminal areas. Climate: four different seasons which are Jilaal, Gu, Xagaa and Deyr. While the Gu and the Deyr reflect the rain seasons; the Xagaa and Jilaal mark the dry seasons. The average annual rainfall for the Juba and Shabelle basin is 550 mm and 455 mm respectively The mean annual temperature varies from 23 – 30 °C The potential rate of evapotranspiration ranges from 1,500 – 2,000 mm/year Geologic and geomorphologic realities mountains, hill lands, plains and valleys Land cover and land use The land cover of the whole Juba-Shabelle River Basin consists mainly of natural vegetation such as riparian forest, bush lands and grasslands. Other land cover types include crop fields, urban areas, dunes, bare lands and natural water bodies. Extracting the floodplains and the alluvial plains the land cover composes of 66 % wooded vegetation (mainly open shrubs), 18 % rangeland (mainly savannah), 15 % agricultural land and 1 % other types of land cover Hydrological conditions Although the catchment area of the Shabelle is about one third larger than the catchment area of the Juba, the mean annual runoff of the Juba is about three times larger than the one of the Shabelle The progressive discharge reduction is due to three factors: (1) The first one is the lack of any significant flow contribution in Somalia. (2) natural losses, i.e. evaporation and infiltration. (3) human driven withdrawals Flood forecasting and early warning methods in use S omalia F lood F orecasting M odel (SFFM), SFFM simulates the daily discharge at downstream locations by using observed discharges at upstream locations and simple regression equations
calibraion not yet satisfactory: choice of the objective function The setting of the boundary values of the calibration parameter. LISFLOOD’s modeling techniques and structure.
calibraion not yet satisfactory: choice of the objective function The setting of the boundary values of the calibration parameter. LISFLOOD’s modeling techniques and structure.