Water management in a semi-arid region: an analogue algorithm approach for rainfall seasonal forecasting Maracchi G., M. Pasqui and F. Piani IBIMET-CNR Consiglio Nazionale delle Ricerche INTERNATIONAL WORKSHOP ON AGROMETEOROLOGICAL RISK MANAGEMENT: CHALLENGES AND OPPORTUNITIES New Delhi, India, 25-27 October 2006
Seasonal Forecasting in Sahel Motivations: Why a new seasonal forecasting method is needed? New insights on African Monsoon physical mechanism and SST role on precipitation (Vizy&Cook 2001, Giannini et al. 2003): Southern Sahel Precipitation is positive correlated with Guinea Gulf SST. Northern Sahel Precipitation is negative correlated with Indian and Central Pacific SST. Interannual variability due to Niño. A monthly anomaly data is needed, at least, for any agrometeorological application: seeding time and early warning systems. Related Activity on Seasonal Forecasting: Setting up a map server based data dissemination tool for end-users: qualitatively browsing of available maps; simple extraction of data for end-users applications: agrometeorological, risk management, hydrology; Spatial Downscaling techniques
Analogues method: an overview SST (Reynolds) as Predictors over : 1.Niño-3 (5S-5N;150W-90W) 2.Guinea Gulf (10S-5N;20W-10E) 3.Indian Ocean (5S-15N;60E-90E) OUTPUT: Precip. (GPCP) Anomaly vs. 1979-2003 Clim. ISSUED: every month VALIDITY: Quarterly and Monthly Sahel Area: 10 W 10 E, 10 N 20 N Target Sahel area Nino3 Most variability during ENSO Water Vapour for African Monsoon Feed Asian Monsoon
Method Standardized* Anomalies (SSTA) obtained by: Subtraction of the 1979-2003 SST average Division by 1979-2003 SST standard deviation Standardized Change Rates to consider the trend of the predictors defined as: difference between current and previous standardized SSTA for the considered month *Standardization is used to have the same order of magnitude of all the predictors
Searching for the analogue year Each month in [1979-2003] is defined by a vector in a 6 dimentional space: Predictors P i : 1. SST Nino-3 std anomalies 2. SST Guinea std anomalies 3. SST Indian std anomalies 4. SST Nino-3 Change rate 5. SST Guinea Change rate 6. SST Indian Change rate Using a simple projection, each original predictors is weighted according to the relative correlation with the Sahel precipitation. P' New Predictors P i i = ω j i ω j P i Where ω ι represents correlation coeff. between P i and Precipitation over Sahel area (10 W 10 E, 10 N 20 N) for a specific month, furthermore we assume that: ω 1 =ω 4 ω 2 =ω 5 ω 3 =ω 6
Searching for the analogue year #2 Analog criterion: Minimization of the Euclidean distance in the 6-dimensional space of predictors P i : min 6 1 ' ' 2 ( Pi curr Pi past ) After the distance minimization the analogue year is found and used for the seasonal forecasts. Best Analog year
Seasonal Forecast: Step by Step CURRENT MONTH e.g.: April 2006 ANALOGUE YEAR e.g.: April 1994 MONTH+1 e.g.: May 2006 May 1994 MONTH+2 e.g.: June 2006 June 1994 MONTH+3 e.g.: July 2006 July 1994 CLIMATOLOGICAL AVERAGE e.g.: May, June, July 1979-2003 ANOMALIES
IBIMET Seasonal Products http://www.ibimet.cnr.it/case/sahel/
Seasonal Rainfall Forecasts http://www.ibimet.cnr.it/case/sahel/ AMJ - Anomaly May Percent Anomaly
Qualitative Comparison: 1998 Good Accordance JAS issued on June 1999
Qualitative Comparison: 2001 Good Accordance JAS issued on June 2003
Qualitative Comparison: Good Accordance 2004 JAS issued on June
Qualitative Comparison: 2000 Bad Accordance JAS issued on June 2002
Qualitative Comparison: AMJ issued on March Good Accordance 2005
Quantitative Comparison: An example of forecast maps of percentage anomaly for the year 2003 with the correspondent observed values from CMAP (Janowiak and Xie 1999): for each month, June, July and August, forecasted maps were shown with different time lags.
Quantitative Comparison: 1 Month Ahead 2 Months Ahead 3 Month Ahead Below Above Norm Below Above Norm Below Above Norm Obs.: Below Normal 25% 13% 38% 25% 49% 38% 62% 37% 13% Below Above Norm Below Below Above Norm Below Below Above Norm Below Obs.: Above Normal 38% 25% 37% Above 49% 13% 38% Above 13% 62% 25% Above Below Above Norm Below Above Norm Below Above Norm Obs.: Normal 13% 62% 25% 13% 49% 38% 49% 13% 38% Normal Normal Normal Chocolate wheels for Sahel area evaluation. Numbers represent the percentage of historical cases in the lower (light grey), middle (white) and upper (dark grey) terciles for August. Table columns represent time lags and table rows represent different observed classes (Below Above Normal from top to bottom).
Ongoing Activities Quantitative Validation Definition of POD, FAR skill indices. Further method development: Sensitivity study to the adopted metric. Possible displacements of active SST areas. Use of Forecast SST anomalies as Predictors. Possible inclusion of the SST of the Extra-Tropics North Atlantic (Mauritania) as predictor. Evaluation of a possible extension of the method to the temperature anomalies over Mediterranean Basin.
Conclusions The improving of seasonal forecasts on Sahel region, especially for agrometeorological applications, should include a full comprehension of physical mechanism including Hadley Cell dynamics. The present spatial resolution should be improved in order to obtain an useful geographical information input for agrometeorological models ( ~ 10km ). Dissemination of seasonal forecast information should take into account the new web-based tools such as Map Server. The method, due to its direct link with some physical mechanism, could be use as an independent tool for further analysis with other statistic approaches.