FCAC CACOF FORO DEL CLIMA DE AMERICA CENTRAL

FCAC (CACOF) FORO DEL CLIMA DE AMERICA CENTRAL Meteorological drought forecasting Inter-COF collaboration with Central America Cedric J. VAN MEERBEECK 1, Simon J. Mason 2, Berny Fallas 3 and Patricia Ramirez 4 1 The Caribbean Institute for Meteorology and Hydrology (CIMH), 2 International Research Institute for Climate and Society (IRI), 3 Instituto Costarricense de Electricidad (ICE), 4 Comité Regional de Recursos Hidráulicos (CRRH) Cari. COF Seasonal Forecast Training, May 26 -27, Kingston, Jamaica

Regional Climate Outlook Forums Cari. COF - FCAC Cari. COF FCAC XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 2

Drought Early Warning (DEW) Depicting the 2009 -2010 Caribbean Drought MONITORING: Caribbean (and Central American!!) SPI and Deciles (Caribbean Drought and Precipitation Monitoring Network) PREDICTION: Caribbean and Central American seasonal precipitation outlook (FCAC and Cari. COF) 3

Drought Early Warning (DEW) Where are we now? • So we produce seasonal forecasts for the Caribbean and Central America… Cari. COF Dec 2013 -Jan-Feb 2014 & Mar-Apr-May 2014 BUT: FCAC Dec 2013 -Jan-Feb-Mar 2014 Even if reliable, how easily could we use a singular forecast of 25% … 35% … 40% of below-normal rainfall? 7

Drought Early Warning (DEW) Where do we want to go? BUT: Even if reliable, how easily could we use a singular forecast of 25% … 35% … 40% of below-normal rainfall? • What kind of drought forecasts best inform DEW? – Reliable (confiable) = forecast probabilities correspond well with observed frequencies (las probabilidades en los pronósticos corresponden bien con las frecencas observadas) – Timely (oportuno). – Understandable language (Lenguaje claro) – Salient (relevante) = forecast must relate to an outcome of direct interest to the user (los pronósticos deben relacionarse directamente con los intereses del usuario. ) – Sharp (con la minima dispersion posible) = assigned probabilities are high enough for effective sectoral resources allocation (las probabilidades asignadas son lo suficientemente buenas y efectivas para los intereses sectoriales) – Cost-effective and sustainable (efectivo y sostenible) = requiring guaranteed continuity of operations in the face of frequent understaffing and limited funding (para poder garantizar la continuidad de las operaciones, considerando las limitaciones de personal y de recursos) XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 8

Drought Early Warning (DEW) How do we get there? • Meteorological drought (and excessive rainfall) prediction using SPIs: – Reliable reliability already support usability of the forecasts (more can be done!!!) – Skillful, very high and low probabilities, and timely droughts are slow-onset climate events (most impacts 3 -12 months from onset of rainfall deficit), allowing inclusion of several months of rainfall observations into the prediction system (persistence). That brings in a lot more skill compared to the limited skill of a 0 -month lead rainfall forecast only. Moreover, if previous months were dry, the probability for drought will be highly shifted upwards. 6 -month SPI forecasting skill (from Lyon et al. , 2012, J. Appl. Met. & Clim. ) 9

Drought Early Warning How do we get there? • Meteorological drought (and excessive rainfall) prediction using SPIs: – Understandable RCOFs promote common understanding between forecasters and users, though a certain disconnect remains to date between products provided and products desired. XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 10

Drought Early Warning How do we get there? • Meteorological drought (and excessive rainfall) prediction using SPIs: – Cost-effective and sustainable minimum of 2 staff with a 5% work commitment in any (combination of) region(s); inter-COF aspect reduces costs of capacity building by increasing efficient knowledge transfer during training, and cutting costs of operations to support sustainability. XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 11

SPI OUTLOOK CASE STUDY THE 2009 -2010 CARIBBEAN DROUGHT XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 12

Expected outcomes of the SPI demo • The demo serves to (i) raise awareness of the new module in CPT and the potential of SPI outlooks; (ii) inquire on both regions’ and other regions’ technical capacity building needs; (iii) seek user feedback on the design of Cari. COF/FCAC operational SPI outlook; (iv) spearhead research on the user benefits of SPI outlooks in the water sector. XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 13

6 -month SPI demo for OND 2009 – JFM 2010 Experimental design • Input data: – Observed rainfall sums for Oct-Nov-Dec 2009 from 108 stations across the Caribbean and Central America; – NOAA-CPC’s CFSv 2 climate model hindcasts of tropical Pacific and Atlantic sea surface temperatures (SSTs) for Jan-Feb-Mar 1983 -2009. • In CPT: – Experiment consists of a persistence component (OND 2009 observed rainfall) and a predicted component (JFM 2010 rainfall forecast); – The predicted component applies a regression model (using Canonical Correlation Analysis) model considering the link between past SSTs and rainfall in the Caribbean and Central America; – Rainfall is then expressed as SPIs; – 6 -month probabilistic SPI outlook results mapped as SPI threshold exceedence probabilities and odds (relative to the historical record). XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 14

Drought forecasting – how will it work? HOW MUCH RAINFALL WE NORMALLY EXPECT 15

Drought forecasting – how will it work? HOW MUCH RAINFALL WE EXPECTED IN JFM 16

Why add in persistence in seasonal forecasts TRADITIONAL SEASONAL RAINFALL OUTLOOK XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 17

Why add in persistence in seasonal forecasts HOW MUCH RAIN FELL SO FAR? XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 18

Why add in persistence in seasonal forecasts SAME SEASONAL RAINFALL FORECAST but adding in observed rainfall surplus/deficit 19

Why add in persistence in seasonal forecasts THIS IS WHAT REALLY HAPPENED!! XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 20

SPI thresholds – drought and excessive precip. • SPI = rainfall totals over a period expressed in units of standard deviations around the historical mean as determined using a gamma distribution. • Caribbean Drought and Precipitation Monitoring Network adopted the following thresholds: Table SPI classification used from January 2011 SPI Value Category SPI Value Impact -0. 50 to -0. 01 -0. 80 to -0. 51 -1. 30 to -0. 81 -1. 60 to -1. 31 -2. 00 to -1. 61 <= -2. 01 Normal Abnormally dry Moderately dry Severely dry Extremely dry Exceptionally dry 0. 50 to 0. 01 0. 80 to 0. 51 1. 30 to 0. 81 1. 60 to 1. 31 2. 00 to 1. 61 >= 2. 01 Normal Abnormally wet Moderately wet very wet Extremely wet Exceptionally wet 21

6 -month SPI demo for OND 2009 – JFM 2010 CDPMN Monitor – old drought categories SPI -2. 0 and less -1. 99 to 1. 5 -1. 49 to -1. 0 -0. 99 to 0. 99 1. 0 to 1. 49 1. 5 to 1. 99 2. 0 and more Climate condition Extremely dry Severely dry Moderately dry Near normal Moderately dry Severely dry Extremely dry Climatological Probability (%) 2. 3 4. 4 9. 2 68. 2 9. 2 4. 4 2. 3 22

6 -month SPI demo for OND 2009 – JFM 2010 Initial results – Drought probabilities 23

6 -month SPI demo for OND 2009 – JFM 2010 Initial results – Drought probabilities 24

6 -month SPI demo for OND 2009 – JFM 2010 Initial results – Drought relative odds Possible use of relative odds – making colour-coded drought alert maps blue (or green): odds low – no concern orange: high odds – drought warning yellow: odds raised – drought watch red: very high odds – drought alarm) 25

Drought risk forecasting – how could it look like? W H ERE IS DRO UGHT RISK IM MINENT? HO W BAD WILL I T GET 3 M ONTHS FROM NOW? 26

6 -month SPI demo for OND 2009 – JFM 2010 Region-wide verification BARBADOS GRENADA 27

6 -month SPI demo for OND 2009 – JFM 2010 Verification for Barbados • Persistence component allows a truthful estimate of rainfall deficit. • EXCEPTIONAL DROUGHT normally occurs once every 40 years, or with less than 2. 3% frequency. • For Barbados, the probability for exceptional drought by March 2010 was 45%. (For Grenada, the worst hit country, this was 60% !!) BARBADOS Abnormally dry Moderately dry Severely dry or Extremely dry Exceptionally or worse or worse dry SPI < -0. 5 SPI < -0. 8 SPI < -1. 3 SPI < -1. 6 SPI < -2. 0 climatological probabilities 29% 11% 6% 2. 30% climatological odds 1. 00 forecast SPI probabilities (OND 2009 -JFM 2010) 88% 77% 65% 53% 49% forecast relative SPI odds (OND 2009 -JFM 2010) 7. 9 8. 9 11 14. 1 15. 8 observed SPI (OND 2009 -JFM 2010) 28

SPI outlooks – what lies ahead? • Development, Tailoring, Operationalisation, Improvement and Streamlining of the SPI outlook methodology by Cari. COF and FCAC. This includes: – technical capacity building through training workshops: Jamaica, 26 -27 May + Costa Rica, 2 -6 June – enhanced stakeholder participation; – targeted climate products and services development. • Questions we will need to ask the user: 1. What is the most usable output visualisation of the SPI outlook to you? Maps? Tables? … 2. Which time scale and time lead is most usable to you? 3. Are you more interested in probabilities or odds/relative odds? XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 29

Thank you XLII FCAC – Ciudad de Guatemala, 09/10 -04 -2014 30