Regional Forecasting of California HABs

Transcription

1 Regional Forecasting of California HABs Chlorophyll Domoic Acid Clarissa Anderson UC Santa Cruz West Coast Regional HAB Summit February 12, 2009

2 Trainer et al 2000, Hickey and Banas 2003, in Kudela 2005 WEST COAST DA EVENT REALLY PUT ISSUE for SBC ON THE MAP! ,000 ng/l

3 Santa Barbara Channel SBC &Monterey Bay Ches. Bay Mark Brzezinski, UCSB Dave Siegel Libe Washburn Raphe Kudela, UCSC Raghu Murtugudde, UMD Christopher Brown, NESDIS Matthew Sapiano, NOAA Anne Thessen, MBL Wen Long, UMD-HPL Bala Mathakumali, UMD John Strack, UMD Raphe Kudela Jenny Lane, UCSC Peter Raimondi, UCSC Peter Miller Yi Chao, NASA JPL Dave Siegel Gregg Langlois, CDHS

4 Integrated Diatom vs. Dinoflagellate Diagnostic Pigments Plumes and Blooms Data: Integrated Peridinin ( g L -1 ) Temporal shift in bloom magnitudes? Dinos Diatoms Integrated Fucoxathin ( g L -1 ) 1 Toxic Pseudo-nitzschia spp. Blooms Anderson et al, Journal of Geophysical Research (2008)

5 PLUMES AND BLOOMS TRANSECT Latitude ( N) San Miguel MONTHLY SAMPLING: Nov 2004 June 2006 Hydrographic and Optical Properties Pseudo-nitzschia spp. Abundance Domoic Acid Concentration (surface) Santa Rosa Longitude ( W) Anacapa Santa Cruz

6 Monthly Mean Across Stations Cellular Domoic Acid (pg cell -1 ) Pseudo-nitzschia spp. Abundance (x 10 3 cell L -1 ) Cellular Domoic Acid (pg cell -1 ) Nov Dec Mar Apr May Jun Aug Sep Oct Nov Nov Jan Mar Mar Apr May Jun Pseudo-nitzschia spp. Abundance (cells L -1 ) r 2 = 0.12 p < 0.001

7 Stepwise Multiple Regression Predictor Variables: 37 physical, chemical, optical parameters Response Variables : Pseudo-nitzschia Abund., pda, cda True Skill (S T ) = Hindcast Skill Artificial Skill S H r 2 S T r 2 Davis, 1976, 1977 Siegel and Dickey, 1986 Two Statistical Models for Each Response Variable: 1) Full Model employs all predictor variables possible 2) Remote-Sensing Model employs only variables that can be remotely estimated (SeaWiFS, MODIS, AVHRR)

8 P-n spp. Abundance Significant Predictors For Full Models (412/555) 75% skill Silicate:Nitrate ratio n = 75 (555) a p (490) (510/555) Significant Predictors For Remote-Sensing Models (412/555) 63% skill (555) n = 89 (510) Chlorophyll (level 2) (510/555) Particulate Domoic Acid Cellular Domoic Acid (510/555) 58% skill Silicate:Phosphate n = 80 Temperature Salinity Temperature 58% skill a g (412) n = 75 Salinity (510/555) a p (510) (510/555) 47% skill Temperature n = 86 (490/555) Day of Year Chlorophyll Temperature 46% skill (412) n = 72 Chlorophyll (510/555) (510)

9 PREDICTING CELLULAR DOMOIC ACID Best Full Model # incorrect nonbloom # nonbloom obs = 10/34 = 29% False Pos. # correct bloom # bloom obs. = 39/40 = 98% Predicted Log(cDA +1) Log(cDA+1) Measured 58% Skill # obs = 75 Presence Threshold for Cellular Domoic Acid 1.1 pg cell -1 Log(cDA+1) = (SST) [a g (412)] (Sal) 1.27[ (510/555)] 5.06[a p (510)] Upwelling CDOM Absorp. Upwelling Chlorophyll Reflectance/Absorption

10 ADVECTION OF TOXIC BLOOM FROM PT. CONCEPTION INTO SANTA BARBARA CHANNEL 9 Sep Sep mg m -3 Anderson et al., Harmful Algae (2009)

11 Santa Barbara Channel Monterey Bay Ches. Bay Mark Brzezinski, UCSB Dave Siegel Libe Washburn Raphe Kudela, UCSC Peter Miller, UCSC Yi Chao, NASA JPL Gregg Langlois, CDHS Raghu Murtugudde, UMD Christopher Brown, NESDIS Matthew Sapiano, NOAA Anne Thessen, MBL Wen Long, UMD-HPL Bala Mathakumali, UMD John Strack, UMD Jenny Lane, UCSC Raphe Kudela Peter Raimondi, UCSC Peter Miller

12 Hindcasts using Generalized Linear Logistic Model for Pseudo-nitzschia spp. the Chesapeake Bay Hindcasted Predictors: - Phosphate + Salinity - Temperature + DOC - Silicic Acid -Month -N:P Probability of Bloom Anderson et al, in prep

13 OPTIMIZATION of the PROBABILITY THRESHOLD for DETERMINING BLOOM" FORECAST False Alarm Ratio 0.57 Proportion Probability 0.40 of Detection Heidke Skill Score 0.39 Prob. = 11-16% Probability of False Detection 0.08 Prob. Threshold Value

14 California Chesapeake Bay Currently working on linking regional ROMS model output and satellite retrievals to Pseudo-nitzschia & DA models for the SBC and later for the Monterey Bay. And validating with multiple sources of state monitoring data.

15 Predictive Modeling of Pseudo-nitzschia Blooms in the Monterey Bay (Jenny Lane, Raphe Kudela, Pete Raimondi) Lane et al., in review

16 Model Comparison (2009) Lane et al., in review

17 SETTING PROBABILITY THRESHOLDS OPTIMALLY Spring Fall-Winter Annual Bloom Non-Bloom Lane et al., in review

18 Collaboration with: Raphe Kudela, Peter Miller, Yi Chao, Dave Siegel, Gregg Langlois SST Toxic Bloom Nowcasts/Hindcasts Salinity HAB Model MODIS-Aqua Chl, ( ) % Prob. Of Bloom Validate! SBC ROMS nowcasts from Yi Chao

19 Predicted Pseudo-nitzschia spp. Abund. May 2008-August 2008 cells/l 1.0 x x (animation of time series!)

20 Predicted Particulate Domoic Acid Conc. May 2007 Aug 2007 ng/l 2.0 x 10 3 (animation of time series!) 0

21 Chlorophyll Particulate DA Hindcast for May 27, 2008 ng L -1 CODAR Surf Currents May 27 May 29

22 Predicted Cellular Domoic Acid Conc. May 2007 Aug 2007 (animation of time series!) pg/cell VALIDATION!!! 1 ) State monitoring data from Gregg Langlois 0 2) Newly funded NSF project to look at export of domoic acid to depth (PI: Claudia Benitez-Nelson, USC) Monthly surface DA and Pseudo-nitzschia Abund. data across SBC that will be used to validate nowcasts