First unofficial forecast

Since there have been delays in getting out our first Cape Cod Bay forecasts (i.e. still waiting for satellite data and flow fields), I decided to attempt a crude forecast using the limited information that we do have.  You'll see that the forecast isn't too bad, but there is quite a bit of room for improvement.  As we incorporate more data (see below), and more advanced methods (e.g. ensemble Kalman filter), we should expect to see marked improvement in the accuracy of our forecasts.

Disclaimer: As you read on, bear in mind that these are not finalized results.  This description is to provide insights into the forecasting process.

First, the data that we have:
- flow fields from previous years
- satellite data from previous years
- zooplankton samples for 2009 from the Provincetown Center for Coastal Studies (PCCS)

What's missing:
- flow fields from 2009 (coming soon)
- satellite data from 2009 (coming soon)

As the weeks roll by, we'll also be getting zooplankton updates from the PCCS, as well as updated flow and satellite data.  The missing data, however, is critical to a good forecast, so this exercise should be taken with a grain of salt.

Now for the forecast:

There's not a whole lot we can do about the missing 2009 data.  Our computation requires sea surface temperature and chlorophyll values from satellites and flow fields, so I've used representative values from previous years.  I then tuned the output to a collection of zooplankton sample data, and scaled the values to correspond to what the PCCS has been seeing in the water this year to give us our forecast (Fig. 1).


Figure 1 Forecast zooplankton abundance (ind. m-3) at selected dates.

An important note: for simplicity's sake, I'm just using the Pseudocalanus parameterization to forecast all zooplankton, omitting for now the Calanus and Centropages groups.  The forecast will lose validity, therefore, as the assemblage changes later in the winter.

If we compare the forecast to the maps from the PCCS through the end of January (Fig. 2), there are a couple of points to make.  The first part of the temporal signal appears to come through, with populations rising to high levels by late January.  Whether or not the decline toward Feb. 21 is observed in the net samples remains to be seen.  The second point is that the spatial pattern is incorrect.  That is, the high concentration that appears in the forecast should be located more toward the southeastern part of the bay.  This is likely due to the fact that the 2009 flow field is missing from the calculation.

This is a good jumping off point, but more than that, it illustrates the importance of the missing data layers.  Stay tuned for a more refined forecast in the next week or so.