NOAA Operational Forecast System Gulf of Mexico, Demonstration since Sep 1999 Operational since Sep

1
NOAA Operational Forecast SystemGulf of Mexico,
Demonstration since Sep 1999 Operational since
Sep 2004 An Operational forecast of impact
0-3 daysand a prediction of likelihood of a
bloom for mgmt

• Bulletin to managers and
• public advisory on web

Goal is to help other U.S. HAB areas
http//www.csc.noaa.gov/crs/habf
2
Steps to HAB regional forecast

• Pre-requisites
• Product of use to managers
• Requirement identified by state/regional partners
• State data available
• Product in usable format for partners
• Continuing
• Skill/usefulness assessment
• Review with partners for product changes and
  improvement
• Assessment of new research to improve quality

3
Proposed National Operation Harmful Algal Bloom
Forecast System
Recreational risk

• System based on regional forecasts
• eastern G.Mex, western G.Mex, Lake Erie,
  G.Maine, Washington, California, Oregon, etc.

4
Forecasts and Modeling?

• Nowcast There is a HAB, where is it today?
• A location and a transport (modeling) problem
• Forecast There is a HAB, where will it be
  tomorrow?
• A location and a transport modeling problem
• Prediction (better than a crystal ball)
• When will HAB initiate or dissipate
• No HAB today, but HAB next week/month?
• This is an ecological modeling problem

Nowcast/Forecast and Prediction are different
5
Nowcasts

• Are simplest in theory Just need real-time
  observations
• But most complex in practice requires
  comprehensive real-time observing system that
  detects HABs everywhere the information is
  needed.
• Alternative Know about it yesterday or in some
  areas, forecast and interpolate the rest of today
• This means transport is required
• Insufficient data for transport, so transport
  models are needed

6
Forecasts, not much different from Nowcasts

• Still need real-time observations
• Still complex in practice requires
  comprehensive real-time observing system that
  detects HABs everywhere the information is
  needed.
• Take yesterday and today and transport to
  tomorrow
• Location and transport model

7
Prediction initiation or dissipation

• This is the role of ecological models
• They require ecological understanding
• Preferences for
• temperature
• season
• light
• nutrients
• Factors
• grazing

8
Prediction Requirements

• Ecological understanding (nutrient and
  environmental requirements, etc. etc.)
• Ecological models (growth, death in response to
  ecology)
• Physical models (to produce environmental
  conditions and move the bloom)
• Messy problem, leads to improved understanding
• However, once the HAB shows up, prediction
  matters less (except for determining the demise).

9
Ecological Models

• 3-D Hydrodynamic/Ecological
• e.g. Gulf of Maine model, has ecological
• initation from cysts
• (then used as a transport model)
• 3-D Norway moncoze model for flagellates
• Heuristic (expert fuzzy logic)
• e.g. HABES project (Alexandrium inputs)

10
For Forecast (or many ecological) models,
observations need to be interpreted for
assimilation

• Synthesis of Data through Analysis
• Combination with Models
• Assessment for Forecasts

11
HAB Forecast System is an example of an
Integrated Ocean Observing System (slide from S.
Hankin, DMAC)
Observing Subsystem
Satellites
Aircraft
Fixed Platforms
Ships
Drifters Floats
AUVs
12
Nowcasts

• Where is it now (or where was it yesterday)?
• Toxin (Cell) concentrations
• Other information
• interpreted water color
• interpreted satellite imagery
• What about ecology? How wrong is a prediction
  of 0-3 days without including ecology?
• Not a problem for growth of dinoflagellates.
  Possibility of a messy problem as dinoflagellates
  swim.

13
LOCATION NOAA Gulf of Mexico Harmful Algal
Bloom System uses cell counts (from states),
satellite and interpretation
Nov 21 chl-a with previous week samples
14
Location What besides cell counts?

• AUV gliders with sensors
• Moored sensors (e.g. Brevebuster of Kirkpatrick
  et al.)
• Handheld units low end
• (e.g. Jellet used in ORHAB)
• Rapid assessment equipment (e.g. ESP, Doucette,
  Greenfield et al. PCR, etc.)
• Remote sensing (if appropriate)

15
Forecasts

• Needs everything for a Nowcast
• Namely, Where was/is it?
• Toxin (Cell) is it now (or concentrations)
• Other information
• interpreted water color
• interpreted satellite imagery
• Analysis goes into
• Transport model
• 1-D (includes wind-based)
• 2-D, 3-D
• heuristic (e.g., toxicity model)

16
Forecasting Transport, 1-D model
Adapt wind drift transport for along-coast with
Karenia brevis Provides good estimate of
along-coast transport of major blooms where wind
forcing is local. Used currently in the NOAA
forecast. Effective for Florida. Not effective
for Texas.
17
2-D model, good for along-coast movement, blooms
often at surface near the coast. Example for
Florida being evaluated now. Lack of spatial
data is a problem. (2-D model is poor choice for
complex 3-D systems)
Bloom A
Bloom B
Day 2004
18
Example 3-D Transport, (hydrodynamic organism)
Running from initial conditions. Gulf of Maine
McGillicuddy, He, Anderson et al.
Assimilation next more obs
19
Models lead to forecast. Bulletin to managers
Analysis Interpreted imagery with cell counts,
polygons Interpreted winds and wind
forecasts Sampling recommendation Also public
condition
http//www.csc.noaa.gov/crs/habf
20
Models lead to public condition forecast
Public notice
A harmful algal bloom has been identified from
northern Pinellas to northern Lee County. Patchy
very low to high impacts possible in Pinellas and
northern Sarasota Counties every afternoon
through Thursday. Patchy very low to moderate
impacts are possible in southern Hillsborough and
Manatee Counties every afternoon through
Thursday. Dead fish have been reported in the
past few days from central Pinellas to southern
Manatee County.
Information only on verified HABs No jargon,
specific wording
http//www.csc.noaa.gov/crs/habf
21
For Forecast (or many ecological) models,
observations need to be interpreted for
assimilation

• Synthesis of Data through Analysis
• Combination with Models
• Assessment for Forecasts

22
Evaluation of forecasts

• NOAA Operational system Assessment of
  new transport models

23
Operational Nowcast/Forecast Models

• Location. If you are fortunate you have a
  known start, e.g. cyst bed.
• Ecology can be simplified for short-term
  forecasts. (Growth rates are often negligible
  over several days)
• Transport models can be heuristic, hydrodynamic,
  etc.
• Analysis Required for assimilation into
  operational transport model.
• Necessary for evaluation of results and
  uncertainties for managers.
• Poor understanding of how to evaluate model
  predictions. How to determine if a model is
  better than another? This must be done by
  facts, not by conjecture (one model is not a
  priori better than another).

24
Ecological Models

• The value
• What conditions start a bloom? When and where do
  you start looking?
• Hypothesis testing for heuristic models (e.g.
  Anderson et al., Gulf of Maine)
• However, for an operational ecological model,
  there is no technique that allows assimilation of
  HAB locations without analysis (I cannot speak to
  nutrients, etc.) Operational models require
  analysis for best input for management
  application.
• Location is nice, but not necessary (when will
  bloom start?)
• Poor understanding of how to evaluate model
  predictions. How to determine if the model is
  good AND better than another.

25
Conclusions

• Significant Improvements needed in
• Location, all possible sources
• Analysis to combine location data
• Analysis of data to assimilate blooms to all
  models
• Ability to assess accuracy of models
• Prediction needs sophisticated ecology (and maybe
  no transport)
• Nowcasts/Forecasts need transport (and minimal
  ecology)