Brightness temperature

The data stored in the level 1b HDFs as delivered is radiance data, scaled by a multiplier (almost always 1000, but check the radsc vector in the HDF to be sure). Many people will wish to convert this to temperature. This requires a computation, of which there is an example below.

This guide assumes you have an ATM level 1b dataset, access to IDL and the IDL routines supplied by us (see attachments on this page).

Setting up

Make sure the supplied routines (get_atm_temp.pro, l_band.pro, lbb.pro, t_bb.pro) are in a directory seen by your IDL path. If necessary you can add to this path. For instance, if you put them in /users/me then on the IDL command line you could type "print,!path" to see what's in the path already, then "!path='/users/me:'+!path" to add a directory to the path.

Converting ATM band 11 to temperature (simplest method)

Run get_atm_temp on your ATM data by typing "get_atm_temp,input_file,output_file" with the appropriate filenames in IDL (inputs can be hdf or bip).

This creates output_file as a copy of input_file with band 11 replaced by temperature in Kelvin scaled by a factor of 100. So a DN of 27315 translates to a temperature of 273.15K - it's freezing out there!

Further details

The program was designed to find the skin temperature of water bodies, so by default it assumes a fixed emissivity of 0.98, a good approximation for water. If you want to input you own emissivity type "get_atm_temp,input_file,output_file,emissivity=X", or if you just want brightness temperature type "get_atm_temp,input_file,output_file,/bright" (equivalent to emissivity=1).

A word of warning - the program does no atmospheric correction so outputs the temperature that an object would have if it were just in front of the sensor. This may affect the accuracy of the retrieved temperatures. The sensor also only sees a few millimetres into the water, and this may not be the same as the bulk water temperature. In my images of water, some brightening can be seen towards the edge of the scan which is probably to do with atmospheric effects. Hence when producing composite maps, I favoured lower temperatures over higher ones. I also experimented with weighting pixels according to how close to the scan centre they were, and so the program also sets ATM band 1 (virtually useless anyway due to noise) to a sinusoidal function from zero at the scan edges to a maximum at the centre. This may be mapped along with band 11 and used to weight overlapping flightlines.