Oceanographic Links to Alexandrium-imposed Toxicity in the Gulf of Maine

PROJECT SUMMARY:
This is a data mining project based on remote sensing, numerical modeling and statistical analyses that will identify and quantify links between coastal toxicity caused by Alexandrium in the Gulf of Maine and oceanic variability. The Maine Department of Marine Resources (DMR) monitors toxin levels in multiple species of shellfish at over 300 Maine coastal sites throughout non-winter months. The same sites and/or species are not necessarily sampled each time. Nevertheless, over 25 years of these data provide an unparalleled documentation of HAB variability along the Maine coast and our best window into the interannual variability of Alexandrium dynamics. Annual in situ surveys of potentially relevant oceanographic characteristics to compare with the DMR record are not available. Three sources of data do provide systematic and consistent metrics of oceanographic variability for extensive, overlapping time periods. 1) Satellite data. Our sea surface temperature (SST) image database provides 4-5 images/day, 1985 - today. These data define our study period (1985-2006), allowing analysis of interannual variability in SST and surface thermal patterns indicative of circulation as time/space series within each year. A shorter timeseries (1997-2004) of daily SeaWiFS multispectral data supplement the SST data. 2) Model fields. We will reconstruct major 3D hydrographic structure and circulation in each study year using our Gulf of Maine numerical model. Based on the 3D Princeton Ocean Model, our hindcasts will use meteorological forcing from the NCEP Eta reanalysis, river discharge, assimilation of daily satellite SST fields and climatological open ocean boundary conditions. 3) River discharge and meteorological records in each study year provide coincident ancillary data.

The overarching hypothesis that structures our investigation is: Interannual differences in the location, timing and magnitude of toxicity events along the Maine coast are associated with interannual variability in Gulf of Maine oceanographic patterns. We will statistically isolate and quantify dominant patterns in the multidimensional and gappy toxicity record. A suite of metrics indicative of oceanographic structure and forcing will be extracted from the image time series, numerical model output fields and ancillary data (e.g. dominant time/space SST variability, location/strength of specific frontal zones, timing/location of patterns indicative of specific circulation, pycnocline depth, timing of stratification, cross-shelf salinity structure, coherence of alongshore currents, etc.). We then use correlation functions and multivariate statistical tools to identify and quantify those characteristics of Gulf of Maine oceanography most consistently linked to toxicity events. We propose an iterative system of analysis, basing initial approaches on previous ECOHAB results and earlier analyses of subsets of the toxicity record, modifying our metrics as we learn which environmental parameters vary most closely with toxicity timeseries.

Our results 1) simplify dominant patterns of variability in the 22+ year toxicity record 2) identify those Gulf of Maine environmental characteristics most closely linked to toxic events 3) deliver a system of easily monitored HAB ocean indices to managers and 4) point to the most promising oceanic features upon which to focus future, more physiologically, based research.

Results:

The following interannual metrics were calculated based on DMR shellfish toxicity time series records for over 300 monitoring stations throughout Maine (1985-2005):

Metric Number	Description
1	Date of first toxicity event of the year.
2	Temporal extent of annual toxicity
3	Magnitude of annual toxicity (mean of highest 3 values)
4	Total toxicity over the year (annual integral)
5	Date of annual maximum in toxicity
6	Presence or absence of any toxicity in a year

CLIMATOLOGIES:

Figure 1: Climatological mean at each station of the date of first toxicity (Metric 1).

Figure 2: Climatological mean at each station of the duration of toxic events (Metric 2).

Figure 3: Climatological mean (geometric) at each station of the magnitude of toxicity (Metric 3).

Figure 4: Climatological mean (geometric) at each station of the integrated annual toxicity (Metric 4).

Figure 5: Climatological mean at each station of the date of maximum toxicity (Metric 5).

Figure 6: Percentage of years at each station that had toxicity (Metric 6).

MULTIVARIATE ANALYSIS RESULTS (Station Clusters):

Figure 7: Results of the multivariate analysis of station similarity based on interannual variability in station toxicity for Metric 1, date of the first toxicity event of the year. Data shows locations of the stations and their respective cluster membership (A) across the whole study area, (B) the same data enlarged over the Casco Bay / mid coast region of maximum station density, and (C) the interannual variability over the study period for each station cluster, as the mean within the member stations each year. Whiskers on each annual mean are standard errors.

Figure 8: The same as Figure 7, but for Metric 2, the temporal extent of annual toxicity.

Figure 9: The same as Figure 7, but for Metric 3, the maximum annual toxicity.

Figure 10: The same as Figure 7, but for Metric 4, total toxicity over the year (annual integral).

Figure 11: same as Figure 7, but for Metric 5, date of annual maximum in toxicity.

Figure 12: The same as Figure 7, but for Metric 6, presence or absence of any toxicity in a year (a binary metric). The annual value for each cluster group (C) is calculated as the percentage of member stations that are toxic in each year. 

MULTIVARIATE ANALYSIS RESULTS (Year Clusters):

Figure 13: Results of the multivariate analysis for year similarity based on station toxicity for Metric 1, the date of first toxicity. At each station, maps show the median yearday of first toxicity within the years of each of the 4 identified year-groups.  White indicates a median start day of 0, used to indicate no toxicity development.

Figure 14: The same as Figure 13, but for Metric 3, the magnitude of maximum toxicity. Maps show the geometric mean of the maximum toxicity values at each station within each of the 4 clustered year-groups.  White stations are those for which the median over the years within the group is zero.

Figure 15: The same as Figure 13, but for Metric 4, total integrated toxicity over a year. Maps show the geometric mean at each station over the years within each of the 4 identified year-groups.  White stations are those with a median of zero.

Figure 16: The same as Figure 13, but for Metric 5, date of annual maximum in toxicity.  Maps show median yearday of the toxicity maximum at each station across the years within each of the 3 groups. White stations are those where the median date was 0, indicating it did not become toxic.

Figure 17: The same as Figure 13, but for Metric 6, presence or absence of any toxicity in a year (a binary metric). Maps show the proportion of years within each group that a station has toxicity, for each of the 3 identified year-groups.