Support for HRRR

[piyushrpt]

Description

• Support for HRRR model

Motivation and Context

• Get HRRR working with ray tracing

How Has This Been Tested?

• Download tested
• Subsetting tested
• Projection interpretation tested

[dbekaert]

@piyushrpt How are we currently validating ray-tracing. Comparing zenith projection to slant with slant directly? or providing zenith unit vector compared with zenith delay option?

[piyushrpt]

We are comparing zenith projected to slant - and checking magnitudes. The only way of really doing this is to build a synthetic weather model with constant values where the computed delay should turn out proportional to ray length. This will be a unit test that will be important and should be developed going forward. The zenith case does not test any aspects of spatial interpolation - so has limited value.

[piyushrpt]

Here is the example input file that I'm testing with right now:

look_dir: right

date_group:
  date_list:  [20181113]

time_group:
  time: "23:00:00"

weather_model: HRRR

aoi_group:
  bounding_box: 35.4 38.2 -76.6 -74.1

height_group:
  use_dem_latlon: False
  height_levels: "-500 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000"

los_group:
  ray_trace: False  # Use projected slant delay by default
  zref: 60000. # Integration height. Only used in raytracing.
  los_convention: isce  # can be "isce" or "hyp3", see *** for details
    # orbit_file: geom/S1A_OPER_AUX_POEORB_OPOD_20181203T120749_V20181112T225942_20181114T005942.EOF

runtime_group:
  verbose: True
  raster_format: nc # Can be any rasterio-compatible format
  output_directory: current/HR_HRRR
  weather_model_directory: current/HR_HRRR/weather_files
  cube_spacing_in_m: 2000.
  output_projection:  32611

Things look ok - we have reconciled reading HRRR with ERA5. But I seem to be encountering a whole bunch of nans here:

RAiDER/tools/RAiDER/models/weatherModel.py

Line 558 in 1585569

def _uniform_in_z(self, _zlevels=None):

All interpolate calls in that function receive valid data but return only NaNs. If there are assumptions being made about arrays being contiguous etc this makes sense. Extra set of eyes familiar with code here would help.

[piyushrpt]

Numerically, this should work. Things to check include

1. Use of correct constants for HRRR
2. Correct interpretation of units for pressure, temperature and geopotential height etc

Hydrostatic delay seems high but I suspect its because of units - possibly acceleration in geopotential height.

[jlmaurer]

Just a few comments, looks great!

[jlmaurer]

I'll go ahead and approve since they are pretty minor comments