Evening dialogues, history, and path forward

Late last night, even after the Yanks were celebrating their victory, the OSSE dialogue continued.  I am compiling many of the thoughts and language of the dialogue.  It provides a great summary of why we are excited, where we have been, and let's us dream about where we are going.

I begin again with Yi's thoughts after yesterday's telecon:

Yi was excited about a feature observed to the North in the MAB.  He saw an outflow feature from the NY harbor,(satellite images of the features and the model realizations are posted in one of the blogs from last night).  Yi's observation was that all four models are warmer in that outflow region, suggesting a weak or  no outflow from the NY harbor, leading him to want some salinity data and he suggested we send a glider up to verify. This suggestion illcited a comment from the Stevens crowd and began an email dialogue.  I am posting that dialogue.

Nickitas chimed in with some thoughts about the to what degree we can believe the SST for identifying small differences.  His point was SST from satellite has inherent errors of a degree Celsius in our region and he cautioned that it can't be taking as absolute truth.  Nickitas looked at the NDDBC buoys and saw NYHOPS is currently tracking COLDER by 1 degree Celsius, not warmer, at NDBC buoy 44009, and it is right on the money at NDBC 44025 and NDBC 44065 just outside NY/NJ Harbor, and several other stations not shown here but in the www.stevens.edu/maritimeforecast pages. These stations are inside the continental shelf.

As Yi pointed out later in the night, The errors for the various satellite data sets are in the range of 0.5 to 1.0 degree.  Merging them together is certainly a challenge, probably will introduce  more errors, so the SST uncertainty is probably over 1.0 degree. While we cannot say much with those regions with smaller differences, there are areas where the difference is significantly more than 1.0 degree, suggesting the discrepancies among the various models.  As was the theme of the discussion in the cool room yesterday, this was viewed as exciting as it provided something as close to an objective "road map" of where to fly.  He also pointed out here was not look for model model differences but to test how our systems are integrated by the cyberinfrastructure.  He asked for comments, and John Orcutt soon chimed in that while there are examples of real-time assimilation into models (e.g.   SCCOOS and Monterey) this is, he asked it this was the first time that the data  and subsequent models are used to steer gliders.

Yi and Pierre followed.  I add to their list.

In 1998-2001, the HyCODE/COMOP efforts at LEO manually ran models and twice a week provided a four day ensemble of forecasts.  All field assets (REMUS, plane, boat, glider) used those forecast to plan their experiments.  Those experiments were a collection of distributed efforts serving many needs, so scientists scattered "Much like Monty Python's race for people with no sense of direction.

In 2003 when a large experiment was conducted.  The  AOSN experiment in Monterey Bay, data are processed  and models are run manually.  Every day, people have to be in the "control room" in  MBARI giving briefings.  Yi had to fly from LAX to MRY a few times a week.  One Sunday night,  they lost Yi's bag, and he had  give the briefing the second day in day-old T-shirt and short.  Many in the room remember that forecast.....  🙂

Progress evolved during the e ASAP experiment in 2006, scientists were  distributed in a virtual mode.  Models were used to assimilate data collected from the field.  While people were looking at model forecasts, but the end-to-end loop was being developed.  FDuring this effort however models were used to guide the gliders and AUVs, the MB06 group even planned glider paths in real-time as well as horizontal and vertical yoyo's for the AUVs (done for PLUSNet, in real-time).

In 2007, with Pierre & Henrik's teams,  guided the Kayaks directly from model forecasts that posted waypoints in web file on a website that the Kayaks read via iridium, with no human in the loop. These efforts continue  in 2008 and 2009.

In 2008, there was the  MB08 experiment, and the network did collect some EO-1 Hyperion images,  but it was not coordinated with gliders and models.  Scientists  guessed where the was interesting feature (the north bay) was  interesting, so Steve Chien's team uploaded commands to steer the satellite imager toward that general direction.

As you said, this is the first time that we complete the sensor web loop end-to-end, including  in situ and satellite sensors, assimilative and forecasting models, and use 3D forecasted current to  optimize the glider trajectories/waypoints.  The next is couple that to the Hyperion overflight for Saturday.

The take home message the vision of distributed end to end system, all in the team have been working is evolving rapidly.  On a personal note, thinking back to the 1998 experiments, it is exciting to see this evolution, and I would submit this history and present state should be a focus of manuscript by this team.  We should think a nice venue to share that exciting history!

Comments are closed.