AWIPS-GOES Data Utilization Page

July 1997 Edition - Dan Birkenheuer

NOAA's Privacy Policy

Introduction

This is a page designed to offer some insights in the application of GOES satellite data to AWIPS . It illustrates some of the work that is being conducted at the Forecast Systems Laboratory (FSL). Most of our satellite research at FSL is targeted toward using data on workstations. The GOES data that will be getting to the forecast office of the future will be coming through AWIPS and its own satellite broadcast network (SBN). SBN should not be confused with the GVAR or "raw data" down-link transmission. Rather the data on SBN is designed as"display ready" images, SBN also contains model data and other nationally disseminated data that is needed in the forecast office. The SBN image data come in standard projections and contain 8-bit pixels. These 8-bit gray scale images span 0 to 255 bits where 0 is dark and 255 is pure white. The images are of fixed dimension and since they have been remapped to standard projections, a specific location in each will always reference the same earth location (latitude and longitude). This makes it very easy to work with this type of image; one does not have to be concerned with remapping each pixel, they are already remapped.

All is not a bed of roses however, sacrifices are made for this convenience. First, only image data is currently available on SBN (GVAR has sounder data also). Second, the image data are only 8-bits whereas GVAR data have 10-bits offering more dynamic range in the images. Third, since the AWIPS image is remapped we lose track of which detector is responsible for a given pixel. Even so, there is useful radiometric information in the AWIPS imagery, and it is possible to use the data in objective analysis. One such analysis system is LAPS (Local Analysis and Prediction System).

The key to using AWIPS formatted image data for objective purposes is to derive radiance information from the degraded 8-bit image data. Here are some useful tools that help accomplish this.

GOES-8 image enhancement curves

To obtain brightness temperatures from AWIPS formatted 8-bit data, it is helpful to know about 10-bit GVAR GOES Imager Calibration. Typical 8-bit display enhancements are linear with respect to brightness temperature. The enhancements are subjectively designed and easily changed. Sometimes it is good to have a number of different enhancements for different applications. Generally, these enhancements are set once and rarely change, but the user needs to be aware they can change. Once the temperature of the pixel is known, with some minor assumptions you can work backwards to determine radiance.

Here are the enhancement curves that SBN uses for image display. Each plot below corresponds to one of the GOES IR channels (2-5). Plotted for each channel are the reference brightness temperature (K) and 8-bit display counts (range: 0 ---> 255) as a function of GVAR 10-bit counts. Note that the enhancement is designed for showing cold cloud tops as white and warm pixels as dark. Click on any of the small plots for a larger version (one that you can read).

Channel 2Channel 3Channel 4Channel 5

The enhancements are also available in numeric form for your convenience. Tables for the GVAR counts, radiance, brightness temperature, and 8-bit enhancements can be viewed or downloaded in ASCII:

It should be noted that the relationship between GVAR 10-bit count and radiance is accurate for all channels and all detectors. However, detector sets are not perfectly identical, their passbands differ slightly. In computing these tables, the central passband wavenumbers have been averaged. Thus, the brightness temperatures plotted and tabulated represent channel averages and are NOT to be used if exact GVAR 10-bit/brightness temperature relationships are required. For 8-bit imaging and certain analysis purposes, the approximation is usually satisfactory.

If one is working from 8-bit images and desires brightness temperatures, it is better to use simple linear relationships instead of the above tables. Here are the equations to use for this problem. In all cases C = 8-bit count value and T = brightness temperature in K. We also reserve the count value of 255 to represent missing or bad data.

Channel 2

Channel 3

Channels 4 and 5

Finally, if you need radiance from 8-bit counts, the best approach would be to compute brightness temperature using the above equations and then interpolate to radiance using the tables.

Special Products

In addition to using AWIPS imagery for LAPS, FSL has also explored creating special imagery from GOES IR data. The idea here is that a central facility would make special products and then disseminate these in the SBN. In image form, the special product would enable it to be used subjectively on a workstation and also objectively by analysis. One example of this type of image is a derived image. Derived imagery are not unique to FSL. Other organizations (i.e., the University of Wisconsin -- Madison) are actively researching derived imagery.

Last year FSL demonstrated a derived image for total precipitable water. This image was created from AWIPS formatted 8-bit imagery (the type described above). We found this image to be inadequate for our LAPS analysis. Therefore it has been sidelined for the time being and we have moved on to using variational techniques utilizing GOES data.

An example of this image is shown here:

A GOES-8 IR derived image showing 3-D water vapor

Caption

Variational Techniques

The variational analysis application of GOES data is new to LAPS. Since April of 1996 we continue to research the application of the AWIPS image data to LAPS upper level moisture. This technique is showing greater promise than using the data to describe total precipitable water as we attempted to do in our composite image trial. It has consistently demonstrated a 50-70% reduction in RMS error when comparing LAPS with and without GOES data to the Denver RAOB. The current algorithm focuses on the 6.7 micron channel (imager channel 3) but also relies on channels 4 and 5.

Ongoing research strives to extend the technique to more levels (the current technique impacts the atmosphere above 500 hPa) and will eventually use variational techniques to simultaneously analyze more than one data source.

Further information on the current variational technique is available here.

Back to Dan Birkenheuer's homepage