Recently in Ecosystem Forecasting Category

2014 Lobster Forecast--Update 6

| No Comments
I think this will be my last update for awhile.  While the skill of our forecast continues to increase through May, the value of the forecast decreases as well.  I might check back on things at the end of the month, but we'll see.

Our final forecast has the start of the season exactly at the long-term average of 6/29:

Forecast2014_05.02.jpg

There is a slightly greater chance that the start date will be earlier than 6/29, but it's pretty small. 

While I'm pretty excited about this forecast, there is one thing that I find unsatisfying.  Because we're defining the start date as the percent of the total landings for the year, we can't check the accuracy of the forecast until the year is complete.  I have some ideas for how we might work around this problem (defining the start as the rate of change in landings, for example).  Here are some other features we hope to add (assuming we can find some funding, of course):

    • move beyond landings to forecasts for hard/soft shell mix
    • forecasts for different lobster zones
    • improved lead time--we think we might be able to start issuing forecasts in November
Stay tuned!

2014 Lobster Forecast--Update 5

| No Comments
The temperatures at 20m were slightly below average for a few days.  This has caused our forecasted start of the high-catch period to shift slightly later, but we are still projecting a more-or-less normal year:

Forecast2014_04.27.jpg

Note that we have changed the labeling on the left-hand axis slightly.  The forecasts are based on the mean temperature over an eight day period.  Originally, we were using the middle of this period as the "date when the forecast was made."  This is obviously not true.  The labels now indicate the last day of data used to make the forecast, as implied by the axis label.  

Although the average temperatures over the last 8 days have been slightly cool, the last two days have been a bit warmer.  When we update the forecast again (likely on Thursday), I would expect the projected start date to shift to the left by a day or so.  

2014 Lobster Forecast--Update 4

| No Comments
I'm on vacation this week, so this post will be very brief. Here's the latest update of our lobster forecast:

Forecast2014_04.23.jpg

The 20m temperatures at Buoy E have stuck pretty close to the average over the last few days, and our forecasted start-date is pretty much where it was last week: a day or so later than the long-term mean date.  I will update again this weekend and once more next week.

2014 Lobster Forecast--Update 3

| No Comments
School vacation starts tomorrow in Maine, and I'll be taking some much needed vacation time.  I will try to update the forecast at least once while I'm gone, but I wanted to make sure to get a fresh forecast up before I leave:

Forecast2014_04.17.jpg
The forecast again has shifted to the left, and we're now essentially predicting a normal season.  The best analog is 2008.  

I'm intrigued by how the season is progressing in the water.  My sense is that the ocean is warming up faster than on shore, especially below the surface.  I think this might be a result of the series of clear, bright days we've had recently.  The most important component to heating in the ocean is the penetration of light into the water.  On a clear day, especially if the winds are low, there are a lot of photons going into the ocean.  The temperature of the air above the water is a lesser factor, and can really just pull heat from the surface of the ocean.  The forecast for the next few days is for sunny and calm conditions.  My bet is that by the middle of next week, our forecast will be for an slightly early start.

2014 Lobster Forecast--Update 2

| No Comments
The warming at Buoy E has continued and the temperatures there are now pretty much at the long-term average.  This means that our forecast is for a season that is only delayed two days (now projected for July 1):

Forecast2014_04.12.jpg
As before, the start dates from previous years are plotted at the top, with blue indicating cold years and red indicating warm years.  The diamonds represent a forecast.  The center of the diamond is the forecasted value and the width is the 95% confidence interval.  Each diamond represents a forecast made using data centered on the date on the left.  

Note that the width of the diamond is smaller than before.  This is due to both higher skill for this week compared to previous and to the fact that our forecast is close to the longer-term average.

I don't have a good sense for weather the warming will continue or whether we will remain close to the long term average.  NOAA's Climate Prediction Center is forecasting below average air temperatures over the next few weeks, so perhaps we'll hold steady with a slightly delayed season.

2014 Lobster Forecast--Update 1

| No Comments
Seascape has a long tradition of trying to develop ecosystem forecasting.  It started with copepods and whales, and last year, we made our first stab at forecasting lobsters, specifically, the date when the Maine lobster landings begin their rapid summer increase.  Our  forecasts were motivated by the extremely warm year in 2012 and the havoc that it caused in the Maine lobster fishery.  

Here is our first attempt at a forecast for the 2014 fishing season:

Forecast2014_04.07.jpg

Just a reminder that this is the forecast for the start of the "high catch period" when landings really begin to increase.  They are based on the temperature at 20m at NERACOOS Buoy E operated by the University of Maine Physical Oceanography Group.  Full details are at the end of the post.

There's a lot on this figure (we're still experimenting with the best way to present these forecasts).  The blue diamonds are the forecasts, with the width of the diamond representing the 95% confidence interval.  Each diamond is the forecast generated on a specific date.  The forecast at the top is the most current, the one on the bottom was made using data in early February.  Our current forecast is for the landings to start picking up around the Fourth of July, about 5 days later than usual (usual is indicated by the black line).

The sticks and the text at the top indicate the start date from the past.  Our current forecast is for the timing of the 2014 season to look a lot like 2003, which was a pretty cold year (indicated by its blue color).  Years that were warmer than normal are indicated in red, years that are near the average temperatures are in black.  Gray indicates that buoy temperatures were not available in those years.

You'll notice that we were forecasting an even later start of the season (about 8 days delayed) back in the end of March.  This coincided with the coldest temperature anomalies. The warming over the last few weeks has been greater than normal, and water temperatures are almost to the 2002-2011 average for this time of year.  This has caused the forecasts to move back towards the center.  I would not be surprised if this trend continues and that our forecast at the end of April is close to the middle.  

We will try to update these forecasts over the next few weeks.  We also have a proposal pending at NASA that would allow us to improve and expand these forecasts in terms of what we forecast (timing and volume of landings by zone, hard/soft-shell mix) and how early we issue them.

Methods
We downloaded temperature data from the 20m sensor on NERACOOS buoy E.  The best place to look at the current conditions at E is through the NERACOOS Climatology Tool, but if you want the data, UMaine provides netCDF files.  We created a time series of 8-day average temperatures.  The dates on the left side of the figure are the middle of the 8d period used to generate the forecasts.

For each 8d period, we used historical temperature and landings data to build a simple linear model relating temperature on that date to the date when landings begin to increase.  We generated the lobster landings "start date" using monthly landings for Maine. For each year, we divide the monthly landings by the total landings for the year, and then use linear interpolation to find the point where normalized landings increase to 0.08.  These are the dates indicated by the sticks at the top.

Predicting Temperature and Lobster Phenology

| No Comments
We will hopefully have a paper coming out soon that lays out the 2012 Gulf of Maine story--expect a link and more info when it's ready.  As you recall, 2012 was really warm.   Lobsters started moving inshore and began molting very early.  This led to an increase in lobster landings but a collapse in prices. 

One of the statements we make in the paper is that "Many biological responses to the warmer 2012 conditions could have been predicted using data that are routinely collected by ocean observing systems."  This statement is meant to be a challenge to the community to start developing ecosystem forecasts using the observing systems. After all, many of these systems were sold based on their ability to provide information useful to resource management. 

To test our assertion, I put together some very simple statistical forecasts for weekly surface temperatures at Buoy E in the Gulf of Maine based on the temperature conditions during the first week of May.  This is basically an extension of my prediction from a few weeks ago. I also developed a simple statistical model that predicts the change in the start of the high-catch period of the lobster fishery.  A description of the models is at the end of the post. Here's what we get for 2013:

LobsterTempPred_2013.jpg

The blue region is the mean temperature cycle surrounded by 95% confidence bounds.  The gray region is the 95% confidence intervals surrounding the prediction (black line).  The red crosses are the actual observations. At the start of May, we were just under a degree warmer than average, and the model predicts a slightly above normal summer.  Note that the most recent observation was actually a bit below normal.  It will be interesting to see how this progresses, given that NOAA is still predicting an above average summer using their more complex models.  I predict a slightly earlier start to the lobster season.

To check the validity of this approach, I went through the full forecasting process for each year from 2002-2012.  For each year, I didn't include any data from that year when I fit the temperature and lobster models.  Thus, the correspondence between the red crosses (the observations) and the forecast (black line) should be a good test of the approach.  Here's how we would've done in 2012:


LobsterTempPred_2012.jpg

Given how anomalous 2012 was, it's a bit surprising that this is one of the best forecasts. However, this might actually make some sense, especially for the lobsters.  The lobster model only includes temperature effects--it doesn't include other processes that might influence landings like the price of fuel or bait, the number of lobstermen, or biological factors.  Since the 2012 temperature signal was so strong, it likely overwhelmed all of these other factors.  Here is the complete set of hypothetical forecasts:

200220032004200520062007200820092010201120122013

So, it looks like there is some potential to forecast aspects of the lobster fishery, and potentially summer water temperatures as well.  The accuracy of the forecasts could be improved within the year by assimilating information (temperature and lobsters) as it comes in or by incorporating other factors into the models.

The Models
The temperature model predicts the sea surface temperature in week j (SST_j) using the previous week's temperature (SST_j-1) and the temperature at 20m in week 16 (T16, first week in May).  For each weekly period (after week 16), I fit the model:
SST_j=a_j*SST_j-1 + b_j*T16 + c_j
using all data from that weekly period.  My very quick assessment is that the fits are pretty good and that including T16 is an improvement over the SST model, but this will need to be more carefully analyzed.  To make the forecasts, I use SST_16 and T16 to compute SST_17.  I then feed SST_17 into the model for week 18 and keep rolling forward.  It would be easy to substitute the observed temperatures.  

The lobster model is even simpler.  It is just a vanilla linear regression between the start date and T16.  The trick is how to define the start date.  We grabbed monthly landings data from Maine DMR and NOAA. For each year, we divide the monthly values by the total landings for the year.  I found the day (using linear interpolation) when landings should first reach 8% of the total.  Of course, the fishery runs year-round, but this captures the start of the really intense summer period.  I have not looked at the end date.  For the figures, I just applied the shift in the start date to the end date (assuming an early year is also a late year).  This is more to show how we might visualize the duration of the season.

Summer Water Temperature Forecast: Warm, but not Hot

| No Comments
Last weekend, the Portland Press Herald ran a story on how lobstermen in Maine are worried about a repeat of last year's warm temperatures and the havoc that the warming caused on the lobster fishery (early inshore migration and ramp-up of high catch period, glut of soft-shells, record catch, collapse in price, etc.).  The paper used the NERACOOS "Ocean and Weather Climate Display" tool to conclude that temperatures at 50 m (locations not specified) are running 1-2° above the average. My colleague Rick Wahle was quoted "We aren't likely to see as an extreme of an event as last year, but I wouldn't be surprised if we see an earlier-than-normal shed."  This prompted me to wonder whether conditions in April are a good predictor of temperatures in the rest of the year?

I grabbed the data from NERACOOS Buoy E.  This buoy has been in the water since 2001, and I think it's location makes it a good indicator of conditions along the coast of Maine.  I then averaged the 0, 20, and 50m temperatures into 8 day bins.  Then, I looked at the correlation between the bin starting on April 15 with the temperatures later in the year:
BuoyE_persistence.jpg
Not surprisingly, at all depths, the correlation with April 15 decreases the further you go on the year.  For the surface temperatures, the correlations decline quite rapidly, but at 20 and 50m, the correlation is quite strong, even into December.  The rapid decline in the correlations at the end of the year reflects the influence of winter mixing and cooling.  My hunch is this continues to decline until late March, when the minimum temperature is reached. 

The strong correlation between April and summer temperatures can be used to forecast conditions this summer.  Unfortunately, the 20m and 50m instruments at Buoy E seem to be offline.  However, nearby Buoy B is running 1°C above its mean temperature, so we'll use that.  Note: the 1-2° increase over the mean reported by the PPH seems to be driven a layer of very warm, salty water in Jordan Basin.  This is potentially an intrusion of temperate, perhaps even Gulf Stream, water.  Worth keeping an eye on this, but my suspicion is that it is an isolated pocket of water that will not impact the more coastal areas we're interested in.  So, if I plug a 1°C anomaly into the equations relating April temperature to the rest of the year, I can predict the temperatures for this summer:
2013_20mprediction.jpg
The blue line is the average temperature at 20m at Buoy E, and the red dashed line is the prediction.  The shaded area indicates the uncertainty around the prediction.  Based on the uncertainties, there's a 75% chance that bottom temps will be warmer than average this summer and a 50% chance that they'll be more than 0.5 degrees warmer.  There is only a 1 in 10 chance that we'll get temperatures as warm as last summer. 

It's worth noting that this prediction assumes that atmospheric conditions will be normal.  During my temperature-cod analysis, I was able to estimate 20m and 50m temperatures using temperatures from the surface.  This suggests that while the 20 m waters might be separated from the surface, the isolation is not complete.  Weather conditions will have a weak effect below 20m, even in the summer, and likely, the deeper layers accumulate the changes coming from the atmosphere.  If the next few months are warmer (or cooler) than average, then I would expect the 20m predictions to run slightly warmer (or slightly cooler).  Right now, NOAA's Climate Prediction Center says there is a 33% chance that New England will be warmer than the 1981-2010 average:
off01_temp.gif
If this plays out, I would expect the summer 20m temperatures to be even warmer than I predict, but a repeat of 2012 is still unlikely.  

The Phenomenal Weirdness of 2012

| 4 Comments
Yes, I know that 2012 is still in progress (BTW, today is 10/11/12), but I think it's very clear that 2012 has been a remarkable year.  We had a massive drought in the midwest (following in the footstep of last year's whopper drought), Arctic sea ice reached its lowest level ever recorded, and the Red Sox stunk.  2012 is also emerging as an exceptional year in the North Atlantic.  Below is an image of how sea surface temperature this summer compared to the 1981-2011 average.

2012SSTanomaly.001.jpg

Of course, the most notable feature of the image is the giant red blob extending from Cape Hatteras to Iceland and penetrating into the Labrador Sea.  This event is larger in than the midwest drought, and like the drought, it has impacted ecosystems and people.  However, because it took place in the ocean, it will be several years before we know the full extent of its impact.  While it is tempting to conclude that this event is caused by global warming (the logic of Jim Hansen's recent paper would suggest it is), I am more interested in using this event to understand how our ecosystems, fisheries, and coastal economies will fare in a warming world.

Painting Monets

| No Comments

The concept for this entry comes from a presentation by Jeffrey Runge.  While practicing a talk at GMRI in front of a small audience, Jeff brought up an interesting analogy for modeling.  He explained that modeling is like impressionistic painting.

blogmonet.png
blogharbor.png

There are many ways to interpret the past and the future.  Models (whether ecosystem, population dynamics, or couple natural and human interactions) are an attempt to portray past time periods and predict future ones.  Jeff describes models with, "Their predictions will not be precise, but more like impressions, but with enough information to discern whether the future state is like a sunny day on the seascoast or a smoggy day in an industrial harbor."  I thought it was a very cool way of describing what a lot of us are working toward.  Beautiful!


About this Archive

This page is an archive of recent entries in the Ecosystem Forecasting category.

Cruise July 2010 is the previous category.

General is the next category.

Find recent content on the main index or look in the archives to find all content.