Scenario 1F Water Temperature Progress

[Bobfrat]

Progress: 1F Temperature

Successes

• Near real time observed water temperature data were discovered and accessed through CO-OPS stations and NERACOOS buoys
• Catalog-driven search using OGC CSW allowed for dynamic updating as new datasets become available (or inactive).
• Data from both OPeNDAP-CF and SOS endpoints were discovered, accessed and plotted (Figure 1). The accessing of SOS endpoint leveraged IOOS driven tools (PYOOS). The observed data is available in high resolution (6 mins), making it useful to support recreational activities like surfing.
• Sea surface temperature from NOAA's nowCOAST project was added as a WMS layer to a map of the observation stations (Figure 2).
• Forecast model datasets were discovered, accessed, and plotted for points near the observation stations. The COAWST model grid point were overlaid on an interactive map to show the spatial resolution of the model and the observation data (Figure 3)
• Modeled and observation data were compared via a time series plot (Figure 4). The observed data was down sampled for easy comparison.

Figure 1: Time series plot of water temperature

Figure 2: Interactive map of the observation stations (CO-OPS stations are blue, NERACOOS buoys are purple) with WMS Sea Surface Temperature data overlaid.

Figure 3: Map shows all the observation stations within a bounding box along with COAWST model grid points.

Figure 4: Observed and modeled water temperature data.

Easy Challenges: Identified and Fixed

• A specialized function was written to extract and parse NERACOOS buoy data via the SOS service.

  Longer-term Challenges Identified

None

[rsignell-usgs]

@bobfrat, in regard to the longer-term challenge of USEAST not working, this works for me:

import iris
url='http://omgsrv1.meas.ncsu.edu:8080/thredds/dodsC/fmrc/us_east/US_East_Forecast_Model_Run_Collection_best.ncd'
var='sea_water_potential_temperature'
cube = iris.load_cube(url,var)

does that work for you, or is there another problem?

[Bobfrat]

Works for me now too. Thanks @rsignell-usgs

[rsignell-usgs]

@Bobfrat , the issue with Iris loading elevation values from FVCOM should have been fixed here: https://github.com/SciTools/iris/issues/1084, https://github.com/SciTools/iris/pull/1159

I notice that you have

iris.FUTURE.netcdf_promote = True

commented out in the code. Does it work if you uncomment it?

[ocefpaf]

@rsignell-usgs I think @Bobfrat meant 2D sigma_coordinates and not 2 sigma coordinates, right? If that is the case it is not solved by iris.FUTURE.netcdf_promote = True. But maybe this can help:

def z_coord(cube):
    """Heuristic way to return the dimensionless vertical coordinate."""
    try:
        z = cube.coord(axis='Z')
    except CoordinateNotFoundError:
        z = cube.coords(axis='Z')
        for coord in cube.coords(axis='Z'):
            if coord.ndim == 1:
                z = coord
    return z

z = z_coord(cube)
if z:
    positive = z.attributes.get('positive', None)
    if positive == 'up':
        idx = np.unique(z.points.argmax(axis=0))[0]
    else:
        idx = np.unique(z.points.argmin(axis=0))[0]
    c = cube[idx, ...].copy()
else:
    idx = None

Let me know if that works for you, I have not test it extensively yet.

[Bobfrat]

iris.FUTURE.netcdf_promote = True does not work for me because I am currently using iris 1.6.0. I can install iris 1.7.0 (on Mac OSX 10.7.5) but I get an import error: ImportError: No module named iris. I thought it might be tied to UDUNITS but i made sure to set the environmental variable. Any advice?

[Bobfrat]

@rsignell-usgs, @ocefpaf thanks for the tips. I upgraded Iris and now FVCOM can be extracted. http://nbviewer.ipython.org/github/Bobfrat/system-test/blob/temp_nb/Theme_1_Baseline/Scenario_1F_Temperature/Temperature.ipynb

[rsignell-usgs]

Which tip specifically got you going?

[rsignell-usgs]

@Bobfrat , I just noticed that it appears that this notebook pulls observational temperatures from various depths, but does not use those depths to extract data from the model. The data from the model seems always to be the surface temperature.

If this is correct, then it's no wonder some of the comparisons are poor. We should be interpolating the vertical temperature profile at the [i,j] grid cell, or at least picking the closest layer.

[Bobfrat]

Adding @ocefpaf 's method seemed to do the trick.

Great point about the depths... I suppose I'll need to do a DescribeSensor request for each of the stations to get the depth.

[jkupiec]

Because no further work has been done on this scenario in about a month, I am closing this.