CariCOF Climate Outlook February-March-April and May-June-July 2015 Coordination – CIMH – Dr. Cédric J. Van Meerbeeck Participating territories Antigua & Barbuda, Aruba, Bahamas, Barbados, Belize, Cayman Islands, Cuba, Curaçao, Dominica, Dominican Republic, French Guiana, Grenada, Guadeloupe, Guyana, Haïti, Jamaica, Martinique, Puerto Rico, St. Barth’s, St. Kitts & Nevis, St. Lucia, St. Maarten/St. Martin, St. Vincent & the Grenadines, Suriname, Trinidad & Tobago and the US Virgin Islands

XLII FCAC – Ciudad de Guatemala, 09/ Regional Climate Outlook Forums CariCOF - FCAC 2 CariCOF FCAC

RAINFALL

FMA rainfall

CPT probabilistic FMA rainfall forecast CCA experiments: 1) Predictor is observed SST over the tropical Atlantic and Pacific over October (data source: NOAA ERSSTv3b, obtained from the IRI data library) 2) Predictor is observed SST over the tropical North Atlantic over October 3) Predictor is predicted SST over the tropical Atlantic and Pacific over DJF (data source: NOAA CPC CFSv2, October initialisation) 4) Predictor is predicted SST over the tropical North Atlantic 5) Predictor is predicted rainfall totals over the Caribbean (data source: ECHAM4.5 ensemble24, obtained from the IRI data library) Besides the major control of ENSO (here represented by Pacific tropical SST anomalies) and tropical north Atlantic SSTs on Caribbean rainfall variability, these experiments take the contrast between Pacific and Caribbean/trop. N Atlantic SSTs into account, as those factors are regarded as the most important drivers of rainfall throughout the Caribbean.

Experiment 1

CPT rainfall forecast Data: Predictor = January SST observations E-20W and 30N-20S (i.e. Tropical Pacific & Atlantic) [Source: ERSSTv3b, from IRI data library] Predictand = FMA rainfall for 474 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 7 Y modes: max = 8; optimum = 3 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 426 stations included in hindcast and forecast Goodness index (transformed Y) = !!0.084 Very Limited!!

ROC area maps

CCA modes

Experiment 1 FMA Rainfall Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 2

CPT rainfall forecast Data: Predictor = January SST observations W and 0-30N (i.e. tropical N Atl.) [Source: ERSSTv3b, from IRI data library] Predictand = FMA rainfall for 474 Caribbean stations Parameters: CCA modes: max = 5; optimum = 4 X modes: max = 8; optimum = 4 Y modes: max = 8; optimum = 7 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 426 stations included in hindcast and forecast Goodness index (transformed Y) = !!0.077 Very Limited!!

ROC area maps

CCA modes

Experiment 2 FMA Rainfall Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 3

CPT rainfall forecast Data: Predictor = FMA simulated SST E-20W and 30N-20S (i.e. tropical Pacific & Atlantic) [Source: NOAA CPC – CFSv2, November initialisation] Predictand = FMA rainfall for 474 Caribbean stations Parameters: CCA modes: max = 5; optimum = 1 X modes: max = 8; optimum = 5 Y modes: max = 8; optimum = 3 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 184 stations included in hindcast and forecast Goodness index (transformed Y) = !!0.051 Very Limited!!

ROC area maps

CCA modes

Experiment 3 FMA Rainfall Forecast Tercile maps BELOW ABOVE NORMAL

Experiment 4

CPT rainfall forecast Data: Predictor = FMa simulated SST 90-20W and 0-30N (i.e. tropical N Atl.) [Source: NOAA CPC – CFSv2, November initialisation] Predictand = FMA rainfall for 474 Caribbean stations Parameters: CCA modes: max = 5; optimum = 1 X modes: max = 8; optimum = 3 Y modes: max = 8; optimum = 2 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 426 stations included in hindcast and forecast Goodness index (transformed Y) = !!0.038 Very Limited!!

ROC area maps

CCA modes

Experiment 4 FMA Rainfall Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 5

CPT rainfall forecast Data: Predictor = JFM simulated rainfall W and 5S-35N (i.e. broader Caribbean) [Source: ECHAM4.5 ensemble 24 – from IRI data library] Predictand = JFM rainfall for 474 Caribbean stations Parameters: CCA modes: max = 5; optimum = 1 X modes: max = 5; optimum = 3 Y modes: max = 5; optimum = 2 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 430 stations included in hindcast and forecast Goodness index (transformed Y) = !!0.058 Limited!!

ROC area maps

CCA modes

Experiment 5 FMA Rainfall Forecast Tercile maps NORMAL BELOW ABOVE

MJJ rainfall

CPT probabilistic MJJ rainfall forecast CCA experiments: Dec 1) Predictor is observed SST over the tropical Atlantic and Pacific over Dec (data source: NOAA ERSSTv3b, obtained from theIRI data library) MJJ 2) Predictor is predicted SST over the tropical Atlantic and Pacific over MJJ (data source: NOAA CPCFSv2, Dec(initialisation)

Experiment 1

CPT MJJ rainfall forecast Data: Dec Predictor = DecSST observations E-20W and 30N-20S (i.e. tropical Pacific & Atlantic) [Source: ERSSTv3b, from IRI data library] Predictand = MJJ rainfall for 423 Caribbean stations Parameters: CCA modes: max = 5; optimum = 5 X modes: max = 8; optimum = 8 Y modes: max = 8; optimum = 8 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 236 stations included in hindcast and forecast Goodness index (transformed Y) = ! Limited !

ROC area maps

CCA modes

Experiment 1 MJJ Rainfall Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 2

CPT MJJ rainfall forecast Data: MJJ Predictor = MJJ simulated SST E-20W and 30N-20S (i.e. tropical Pacific & Atlantic) [Source: NOAA CPC – CFSv2, Dec initialisation] Predictand = MJJ rainfall for 423 Caribbean stations Parameters: CCA modes: max = 5; optimum = 1 X modes: max = 8; optimum = 2 Y modes: max = 8; optimum = 1 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 390 stations included in hindcast and forecast Goodness index (transformed Y) = !! Very Limited !!

ROC area maps

CCA modes

Experiment 2 MJJ Rainfall Forecast Tercile maps NORMAL BELOW ABOVE

Supporting probabilistic rainfall forecasts CARICOF national/sub-regional objective forecasts & outlook guidance: 1)Puerto Rico & USVI:FMAA = 45% ; N = 30% ; B = 25% MJJA = 55% ; N = 25% ; B = 20% Note: CPT CCA showed limited skill. 2)Guadeloupe:FMAA = 35% ; N = 35% ; B = 30% Martinique :FMAA = 32% ; N = 33% ; B = 35% French Guiana:FMAA = 30% ; N = 34% ; B = 36% French Northern Antilles:FMAA = 32% ; N = 35% ; B = 33% Note: For the Lesser Antilles there is an agreement between several dynamic models (Eurosip, JMA) ) with no visibility over the area. Statistic SST/RR (with skill) for a below forecast martinique and normal or above for Guadeloupe and Iles du Nord in FMA (predictor stay SST over tropical North Atlantic and Pacific, with positive ENSO anomaly). Over French Guiana : The statistic model has good skill and says below. The dynamic models exhibit a drier than normal forecast. 3)Barbados:FMAA = 34%; N = 30% ; B = 36% MJJA = 48%; N = 30% ; B = 22% Note: CPT CCA with ERRSTs and CFSv2 SSTs showed limited to good skill and moderate to fair discrimination 4)St. Maarten & NE Caribb.:FMAA = %; N = % ; B = % Note:

Supporting probabilistic rainfall forecasts CARICOF national/sub-regional objective forecasts & outlook guidance: 5)Trinidad and Tobago:FMAA = 29% ; N = 36% ; B = 35% MJJA = 40% ; N = 34% ; B = 26% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed moderate to good skill and no to good discrimination for FMA, and no to good discrimination for MJJ. 6)Jamaica:FMAA = 30% ; N = 28% ; B = 42% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed limited skill and no to good discrimination. 7)Suriname:FMAA = 34% ; N = 27% ; B = 37% MJJA = 36% ; N = 35% ; B = 27% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed very limited to good skill and no to moderate discrimination for FMA, and no discrimination for MJJ. 8)Belize:FMAA = 25% ; N =30 % ; B =45 % MJJA = 40% ; N =30 % ; B =30 % Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed verylimited to limited skill and no to fair discrimination. 9)St. Vincent:FMAA = 41% ; N = 23% ; B = 37% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed fair skill and poor to moderate discrimination. 10)Curaçao:FMAA = % ; N = % ; B = % Note:

Supporting probabilistic rainfall forecasts CARICOF national/sub-regional objective forecasts & outlook guidance : 11)Grenada:FMAA = % ; N = % ; B = % MJJA = % ; N = % ; B = % Note: 12) Aruba:FMAA = 48% ; N = 25% ; B = 27% *** MJJA = 37% ; N = 34% ; B = 29% *** Note: CPT CCA experiment using ERSSTv3b showed very limited skill and no to fair discrimination.***After multi-model ensemble: A30 N40 B30. Model prediction shows that a warmer-than-normal condition in the tropical Pacific like El Ni ñ o Modoki will continue, but decay in the boreal summer. Therefore leaning more towards normal for rainfall and temperature for FMA fcst and more regular El Ni ñ o like for MJJ. St. Lucia:FMAA = 25% ; N = 40% ; B = 35% MJJA = 40% ; N = 30% ; B = 30% Note: Cayman:FMAA = 31% ; N = 16% ; B = 53% MJJA = 59% ; N = 27% ; B = 14% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed moderate to good skill and no to good discrimination. Dominica:FMAA = 49% ; N = 23% ; B = 28% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed limited skill and no to poor discrimination. Antigua:FMAA = 33% ; N = 34% ; B = 33% MJJA = 58% ; N = 27% ; B = 15% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed very limited to fair skill and no to fair discrimination.

Supporting probabilistic rainfall forecasts CARICOF national/sub-regional objective forecasts & outlook guidance : 18)St. Kitts:FMAA = 26% ; N = 32% ; B = 42% Note:CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed fair skill and moderate to fair discrimination 19)Bahamas:FMAA =38 % ; N =25 % ; B = 37% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed limited to fair skill and moderate to fair discrimination 20)Cuba - West:FMAA = 42% ; N = 35% ; B = 23% MJJA = 23% ; N = 16% ; B = 61% Cuba - Central:FMAA = 33% ; N = 36% ; B = 31% MJJA = 24% ; N = 11% ; B = 65% Cuba - East:FMAA = 41% ; N = 34% ; B = 25% MJJA = 30% ; N = 14% ; B = 56% Note: CPT CCA experiments using ERSSTv3b and CFSv2 SSTs showed very limited to fair (CFSv2) skill and poor to moderate discrimination.

CariCOF CPT output (background) and National/sub-regional rainfall outlooks

CariCOF precipitation outlooks

Supporting probabilistic precipitation forecasts Eight data sources (incl. GPCs): 1)IRI multi-model probability forecast (only multi-model forecast that recalibrates & weights the models based on past performance); 2)UK Met Office GCM (UKMO) probability forecast; 3)European Center for Mid-range Weather Forecast GCM (ECMWF) and EUROSIP (multi-model) probability forecasts; 4)APEC Climate Center (APCC) multi=model probability forecasts; 5)WMO Lead Centre for LRF – MME. 6)CFSv2 model 7)MétéoFrance Arpège model. 8)JMA model Rainfall outlooks are generated next to T 2m outlooks (see Appendix). Rainfall tendencies from supporting models and local expertise: 1)ENSO conditions have recently been warm (SST anom °C). Most new model runs suggest a possible progression into an El Niño by DJF , with an estimated overall confidence of 60-65% for FMA and % for MJJ. Similarly, SSTs are above average much of the Caribbean Islands, a pattern that is expected to persist into MJJ, Further east in the tropical N Atlantic, temperatures recently cooled to near-average, and might cool further into MJJ. 2)The warmer SSTs around the islands may lead to average to above-average air moisture in the Greater Antilles and Lucayan archipelago. In the south-east, a pattern of drier air could be observed in FMA. Further, the suggested weak El Niño may sustain increased vertical wind shear over the tropical North Atlantic as well as over eastern and southern Caribbean (including the Guianas). 3)The signals of added moisture in the northern portions, and drier air with possibly stronger wind shear may define a gradient in anomalous rainfall for FMA. Thus, global models are suggesting a shift to above-normal in the NW and below-normal rainfall in the SE of the Caribbean.

DRAFT Probabilistic FMA rainfall forecast map DRAFT

Probabilistic MJJ rainfall forecast map DRAFT

2m TEMPERATURE

FMA

CPT probabilistic FMA 2m temp. forecast CCA experiments: 1) Predictor is observed SST over the tropical Atlantic and Pacific over December (data source: NOAA ERSSTv3b, obtained from the IRI data library) 2) Predictor is observed SST over the tropical North Atlantic over December 3) Predictor is predicted SST over the tropical Atlantic and Pacific over FMA (data source: NOAA CPC CFSv2, December initialisation) 4) Predictor is predicted SST over the tropical North Atlantic over FMA 5) Predictor is predicted T2m over the Caribbean over FMA (data source: ECHAM4.5 ensemble24, obtained from the IRI data library)

Experiment 1

CPT T2m forecast Data: Predictor = December SST observations E-20W and 30N-20S (i.e. Tropical Pacific & Atlantic) [Source: ERSSTv3b, from IRI data library] Predictand = FMA T2m for 58 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 2 Y modes: max = 8; optimum = 4 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 37 stations included in hindcast and forecast Goodness index (transformed Y) = Good

ROC area maps

CCA modes

Experiment 1 FMA T2m Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 2

CPT T2m forecast Data: Predictor = December SST observations W and 0- 30N (i.e. tropical N Atl.) [Source: ERSSTv3b, from IRI data library] Predictand = FMA T2m for 58 Caribbean stations Parameters: CCA modes: max = 5; optimum = 3 X modes: max = 8; optimum = 7 Y modes: max = 8; optimum = 7 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 37 stations included in hindcast and forecast Goodness index (transformed Y) = Fair

ROC area maps

CCA modes

Experiment 2 FMA T2m Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 3

CPT T2m forecast Data: Predictor = FMA simulated SST E-20W and 30N-20S (i.e. tropical Pacific & Atlantic) [Source: NOAA CPC – CFSv2, Dec initialisation] Predictand = FMA T2m for 58 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 4 Y modes: max = 8; optimum = 6 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 38 stations included in hindcast and forecast Goodness index (transformed Y) = Good

ROC area maps

CCA modes

Experiment 3 FMA T2m Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 4

CPT T2m forecast Data: Predictor = FMA simulated SST 90-20W and 0-30N (i.e. tropical N Atl.) [Source: NOAA CPC – CFSv2, Dec initialisation] Predictand = FMA T2m for 58 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 3 Y modes: max = 8; optimum = 4 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 38 stations included in hindcast and forecast Goodness index (transformed Y) = Good

ROC area maps

CCA modes

Experiment 4 FMA T2m Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 5

CPT T2m forecast Data: Predictor = FMA simulated T2m W and 5S-35N (i.e. broader Caribbean) [Source: ECHAM4.5 ensemble 24 – from IRI data library] Predictand = FMA T2m for 58 Caribbean stations Parameters: CCA modes: max = 3; optimum = 2 X modes: max = 5; optimum = 5 Y modes: max = 5; optimum = 2 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 37 stations included in hindcast and forecast Goodness index (transformed Y) = Good

ROC area maps

CCA modes

Experiment 5 FMA T2m Forecast Tercile maps NORMAL BELOW ABOVE

MJJ T2m

CPT probabilistic MJJ 2m temp. forecast CCA experiments: DEC 1) Predictor is observed SST over the tropical Atlantic and Pacific over DEC (data source: NOAA ERSSTv3b, obtained from theIRI data library) MJJ 2) Predictor is predicted SST over the tropical Atlantic and Pacific over MJJ (data source: NOAA CPCFSv2, December (initialisation)

Experiment 1

CPT MJJ T2m forecast Data: DECEMBER Predictor = DECEMBER SST observations E-20W and 30N-20S (i.e. tropical Pacific & Atlantic) [Source: ERSSTv3b, from IRI data library] Predictand = MJJ T2m for 58 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 7 Y modes: max = 8; optimum = 2 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 38 stations included in hindcast and forecast Goodness index (transformed Y) = Moderate

ROC area maps

CCA modes

Experiment 1 MJJ T2m Forecast Tercile maps NORMAL BELOW ABOVE

Experiment 2

CPT MJJ T2m forecast Data: MJJ Predictor = MJJ simulated SST E-20W and 30N-20S (i.e. tropical Pacific & Atlantic) [Source: NOAA CPC – CFSv2, Dec initialisation] Predictand = MJJ T2m for 58 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 5 Y modes: max = 8; optimum = 2 Missing Y data threshold: 20% missing values, 10% missing stations, leading to 39 stations included in hindcast and forecast Goodness index (transformed Y) = Moderate

ROC area maps

CCA modes

Experiment 2 MJJ T2m Forecast Tercile maps NORMAL BELOW ABOVE

Supporting probabilistic T 2m forecasts CARICOF national/sub-regional objective forecasts & outlook guidance: 1)Cayman:FMAA = 40% ; N = 33% ; B = 27% MJJA = 32% ; N = 12% ; B = 56% Note: CPT CCA with ERSSTv3b showed limited to fair skill and poor to good discrimination. 2)St. Lucia:FMAA = 65% ; N = 20% ; B = 15% MJJA = 50% ; N = 30% ; B = 20% Note: 3)Puerto Rico & USVI:FMAA = 60% ; N = 25% ; B = 15% MJJA = 55% ; N = 25% ; B = 20% Note: CPT CCA showed virtually no skill and limited to good discrimination. 4)St. Maarten & NE Caribb:FMAA = %; N = % ; B = % Note: 5)Trinidad and Tobago:FMAA = 51% ; N = 23% ; B = 26% MJJA = 50% ; N = 24% ; B = 26% Note:CPT CCA with ERSSTv3b & CFSv2 SSTs showed moderate to good skill and fair to good discrimination. 6)Jamaica:FMAA = 69% ; N = 12% ; B = 19% Note:CPT CCA with ERSSTv3b & CFSv2 SSTs showed fair skill and poor to fair discrimination. 7)Belize:FMAA = 55% ; N =25 % ; B =20 % Note: CPT CCA with ERSSTv3b & CFSv2 SSTs showed good skill and poor to fair discrimination. 8)St. Vincent:FMAA = 45% ; N = 31% ; B = 23% Note: CPT CCA with ERSSTv3b & CFSv2 SSTs showed good skill and fair discrimination. 9)Dominica:FMAA = 28% ; N = 52% ; B = 20% Note: CPT CCA with moderate to good skill using ERSSTs & CFSv2 SSTs, resp., with poor to good discrimination.

Supporting probabilistic T 2m forecasts CARICOF national/sub-regional objective forecasts & outlook guidance: 10)Barbados:FMAA = 47% ; N = 41% ; B = 12% MJJA = 58% ; N = 16% ; B = 26% Note: CPT CCA with ERSSTs and CFSv2 SSTs showed good skill and fair to good discrimination. 11)Guyana:FMAA = % ; N = % ; B = % Note: 12)Grenada:FMAA = % ; N = % ; B = % MJJA = % ; N = % ; B = % Note: 13)Suriname: FMAA = 55% ; N = 35% ; B = 10% MJJA = 45% ; N = 35% ; B = 20% Note: 14)Aruba:FMAA = 60% ; N = 32% ; B = 8% MJJA = 46% ; N = 28% ; B =26 % Note: CPT CCA with ERSSTs and DJF CFSv2 SSTs showed very limited skill and good to very good discrimination 15)Antigua:FMAA = 26% ; N = 35% ; B = 39% MJJA = 36% ; N = 40% ; B = 24% Note: CPT CCA with ERSSTs and DJF CFSv2 SSTs showed fair to good skill and moderate to fair discrimination. 16)Bahamas:FMAA = % ; N = % ; B = % MJJA = % ; N = % ; B = % Note: 17)St. Kitts:FMAA = % ; N = % ; B = %. Note:

Supporting probabilistic T 2m forecasts CARICOF national/sub-regional objective forecasts & outlook guidance: 18)Cuba - West:FMAA = 33% ; N = 34% ; B = 33% MJJA = 55% ; N = 27% ; B = 18% Cuba - Central:FMAA = 34% ; N = 33% ; B = 33% MJJA = 60% ; N = 26% ; B = 14% Cuba - East:FMAA = 33% ; N = 34% ; B = 33% MJJA = 60% ; N = 25% ; B = 15% Note: CPT CCA with ERSSTs Oct and CFSv2 SSTs showed limited to fair skill and not to poor discrimination.

CariCOF CPT output (background) and National/sub-regional T2m outlooks

Probabilistic FMA 2m temperature forecast DRAFT

Probabilistic MJJ 2m temperature forecast DRAFT

Drought outlook

NDJFMA

CPT probabilistic NDJFMA CCA experiments: 1) Predictor is observed SST over the tropical Atlantic and Pacific over December (data source: NOAA ERSSTv3b, obtained from the IRI data library) 2) Predictor is observed SST over the tropical North Atlantic over December 3) Predictor is predicted rainfall totals over the Caribbean over JFMA (data source: ECHAM4.5 ensemble24, obtained from the IRI data library)

Experiment 1

CPT SPI6 drought outlook Data: Predictor = December SST observations E-20W and 30N-20S (i.e. Tropical Pacific & Atlantic) [Source: ERSSTv3b, from IRI data library] Predictand = JFMA SPI for 423 Caribbean stations Parameters: CCA modes: max = 5; optimum = 3 X modes: max = 8; optimum = 8 Y modes: max = 8; optimum = 3 Missing Y data threshold: 20% missing values, 50% missing stations, leading to 387 stations included in hindcast and forecast Goodness index (transformed Y) = !0.112 Limited!

Experiment 2

CPT SPI6 drought outlook Data: Predictor = December SST observations W and 0-30N (i.e. tropical N Atl.) [Source: ERSSTv3b, from IRI data library] Predictand = JFMA SPI for 423 Caribbean stations Parameters: CCA modes: max = 5; optimum = 2 X modes: max = 8; optimum = 3 Y modes: max = 8; optimum = 2 Missing Y data threshold: 20% missing values, 50% missing stations, leading to 387 stations included in hindcast and forecast Goodness index (transformed Y) = !0.116 Limited!

Experiment 3

CPT SPI6 drought outlook Data: Predictor = JFMA rainfall forecast W and 5S-35N (i.e. Caribbean) [Source: ECHAM4.5, from IRI data library] Predictand = JFMA SPI for 423 Caribbean stations Parameters: CCA modes: max = 3; optimum = 2 X modes: max = 5; optimum = 4 Y modes: max = 5; optimum = 5 Missing Y data threshold: 20% missing values, 50% missing stations, leading to 388 stations included in hindcast and forecast Goodness index (transformed Y) = !!0.073 Very Limited!!

JJASONDJFMAM

CPT probabilistic NDJFMA CCA experiments: 1) Predictor is observed SST over the tropical Atlantic and Pacific over December (data source: NOAA ERSSTv3b, obtained from the IRI data library)

Experiment 1

CPT SPI6 drought outlook Data: Predictor = December SST observations E-20W and 30N-20S (i.e. Tropical Pacific & Atlantic) [Source: ERSSTv3b, from IRI data library] Predictand = JFMAM SPI for 474 Caribbean stations Parameters: CCA modes: max = 5; optimum = 3 X modes: max = 8; optimum = 8 Y modes: max = 8; optimum = 3 Missing Y data threshold: 20% missing values, 50% missing stations, leading to 377 stations included in hindcast and forecast Goodness index (transformed Y) = !0.144 Limited!

Supporting probabilistic SPI forecasts CARICOF national/sub-regional expert views: 1)Barbados:ONDJFM (SPI<-0.8)P = 28%Rel. Odds =0.8 (no concern) Note: CPT CCA with fair discrimination ROC ~0.7. 2)St. Maarten & NE Caribb: NDJFMA? (SPI<-0.8)P = 6%Rel. Odds <1 (no concern) Note: CPT CCA with fair to good discrimination ROC The SPI probability for below normal is quite low. 3)Trinidad & Tobago: hydrol. year (SPI<-0.8)P = 35%Rel. Odds =2 (drought watch) Note: CPT CCA with fair skill and moderate to good discrimination ROC )Jamaica: ONDJFM (SPI<-0.8)P = 30%Rel. Odds =1.67 (drought watch, with drought warning in SE most portions of the country) Note: CPT CCA with limited skill andno to good discrimination ROC )Belize: hydrol. year (SPI<-0.8)P < 20%Rel. Odds <1 (no concern) Note: CPT CCA with negative skill and no to good discrimination ROC )St. Vincent:ONDJFM (SPI<-0.8)P = 57%Rel. Odds = 6 (drought warning) Note: CPT CCA with limited skill and good discrimination ROC )Grenada:DJFMAM? (SPI<-0.8)P = 24%Rel. Odds =1.5 (no concern) Note: CPT CCA with negative skill and no discrimination. (to be dismissed; probably no persistence) 8)Aruba: hydrol. year (SPI<-0.8)P = 30%Rel. Odds = 0.43 (no concern) Note: CPT CCA with fair skill and good discrimination. 9)St. Lucia:ONDJFM (SPI < -0.8)P = %Rel. Odds = 1 (no concern) Note:

Supporting probabilistic SPI forecasts CARICOF national/sub-regional expert views: 10)Cayman: hydrol. year (SPI<-0.8)P = 69%Rel. Odds =11 (drought warning) Note: CPT CCA with very good discrimination. 11) Dominica: ONDJFM (SPI<-0.8)P = 38%Rel. Odds =2 (drought watch) Note: CPT CCA with fair discrimination. 12) Antigua: hydrol. year (SPI<-0.8)P = 29%Rel. Odds =1.6 (drought watch) Note: CPT CCA with limited skill but good discrimination. Jun – May: SPI/Drought outlook: B 29% prob of at least a moderate drought (rainfall deficit by end of May). Aug-Apr: SPI/Drought outlook: B 25% prob (rel odds of about 1.3) of at least a moderate drought (rainfall deficit by end of Apr). Oct-Mar: SPI/Drought outlook: B 27% prob (rel odds of about 1.5) of at least a moderate drought (rainfall deficit by end of Mar). 13) Bahamas: hydrol. year (SPI<-0.8)P = 18%Rel. Odds =0.8 (no concern) Note: CPT CCA with moderate skill and very good discrimination. 14) Cuba: hydrol. year (SPI<-0.8) P(West) = 31% ; P(Central) = 33% ; P(East) = 49 Rel. Odds(West) =1.9 (no concern) ; Rel. Odds(Central) =2 (drought watch) ; Rel. Odds(West) =3.9 (drought watch) 15) St. Kitts: NDJFMA (SPI<-0.8)P = 69%Rel. Odds = 7.1 (drought warning) Note: CPT CCA with moderate skill and very good discrimination.

SPI outlook Nov to Apr – areas under immediate drought concern? 124 January 2015’s update: Drought concerns have subsided throughout the Caribbean, except SE Belize as well as in portions of the Guianas. Some concern exists in the ABC Islands, Belize, Trinidad & Tobago and in portions of French Guiana and the Windward Islands. December 2014’s update

Drought outlook till May 2015 – shorter- / longer-term concern? 125 CONSERVE WATER!! Especially in Cayman, Haïti and Jamaica. Current drought situation: – Cayman, Eastern Jamaica, Haïti and to a lesser extent part of the Windward Islands are in drought and have suffered water shortages. Shorter-term: – We expect that the drought situation to possibly appear in some of the other islands and improve over others. Longer-term: – Because of below normal rainfall during the wet season, water shortages may occur in portions of the Antilles during the dry season, which is our tourist season. – Chances are that an El Niño may evolve. El Niño often results in a drier dry season (except for the NW Caribbean), and particularly in the SE Caribbean later in the dry season. However, above-average temperatures in the Atlantic suggest a wetter dry season. Regardless, areas with existing water shortages may not see any improvement until late in the dry season.

SPI outlook Hydrological Year – drought concern by the end of the Caribbean dry season (May 30 th, 2015)? 126 This 12-month SPI-based drought outlook uses observations until December 2014, with potential impacts on large surface water reserves and groundwater. Impactful hydrological drought is a concern in the Greater Antilles (except Cuba and N Hispaniola), as well as in portions of the Bahamas, Belize, Dominica southward to St. Vincent. Drought is possible in the Leeward Islands. Previous update

Key to interpreting drought alert levels – operational (to be upgraded) 127 ColourAlert levelMeaningSuggested action level GREEN No ConcernNo drought concern Business as usual YELLOW Drought WatchDrought possible Keep updated, conserve water ORANGE Drought Warning Drought evolving Be prepared, conserve water. protect RED Exceptional Drought Drought of immediate concern Take action, ration water, protect

128 ALERT LEVELMEANINGACTION LEVEL NO CONCERNNO CONCERN No drought concern monitor resources update and ratify management plans public awareness campaigns upgrade infrastructure DROUGHT WATCH Drought possible keep updated protect resources and conserve water implement management plans response training monitor and repair infrastructure DROUGHT WARNING Drought evolving protect resources conserve and recycle water implement management plans release public service announcements last minute infrastructural repairs and upgrades report impacts DROUGHT EMERGENCY Drought of immediate concern release public service announcements implement management and response plans enforce water restrictions and recycling enforce resource protection repair infrastructure report impacts

Appendix

US Climate Prediction center – Climate Diagnostics bulletin

US Climate Prediction center – Climate Diagnostics bulletin

NOAA CPC NAO index monitoring/forecasting

Explanatory variables – IRI Predicted SSTs El Niño region Caribbean & tropical Atlantic

Explanatory variables – IRI Predicted SSTs El Niño region Caribbean & tropical Atlantic

Explanatory variables – ECMWF Predicted Tropical SSTs st!sea%20surface%20temperature!1%20month!Tropics!201401!ensemble%20mean!/

Explanatory variables – CPC/IRI ENSO Forecast

IRI – multi-model probabilistic rainfall forecast

EUROSIP – multi-model probabilistic rainfall forecast ummary!1%20month!Tropics!201310!/

ECMWF – ensemble probabilistic rainfall forecast

UK Met Office – probabilistic rainfall forecast ABOVE NORMAL BELOW

NOAA CPC – probabilistic rainfall forecast

APCC – multi-model probabilistic rainfall forecast

WMO LC for LRF-MME – Multi-model probabilistic rainfall forecast =4#

WMO LC for LRF-MME – Individual deterministic rainfall forecast

courtesy of Christophe Montout Météo France Arpège – probabilistic rainfall forecast ***

Japan Met Agency – probabilistic rainfall forecast

IRI – multi-model probabilistic T 2m forecast

EUROSIP – multi-model probabilistic T 2m forecast ummary!1%20month!Tropics!201309!/

ECMWF – ensemble probabilistic T 2m forecast

UK Met Office – probabilistic T 2m forecast ABOVE NORMAL BELOW

APCC – multi-model probabilistic T 2m forecast

WMO LC for LRF-MME – Multi-model probabilistic T 2m forecast =4#

WMO LC for LRF-MME – Individual deterministic T 2m forecast

WMO Lead Centre for Long-Range-Forecast – individual GPC models inter-model consistency

Japan Met Agency – probabilistic T2m forecast

Station three-monthly Rainfall Climatology January to March sum

Station three-monthly Rainfall Climatology February to April sum

Station three-monthly Rainfall Climatology March to May sum

Station three-monthly Rainfall Climatology April to June sum

Station three-monthly Rainfall Climatology May to July sum

Station three-monthly Rainfall Climatology June to August sum

Station three-monthly Rainfall Climatology July to September sum

Station three-monthly Rainfall Climatology August to October sum

Station three-monthly Rainfall Climatology September to November sum

Station three-monthly Rainfall Climatology October to December sum

Station three-monthly Rainfall Climatology November to January sum

Station three-monthly Rainfall Climatology December to February sum

TRMM Monthly Rainfall Climatology January, February & March

TRMM Monthly Rainfall Climatology April, May & June

TRMM Monthly Rainfall Climatology July, August & September

TRMM Monthly Rainfall Climatology October, November & December