Hi Dimosthenis,

with "calibrate" probabily you mean to convert the at sensor radiances to earth surface temperature. Am I right?

If so, with AVHRR and MODIS "easy" algorithms to apply are variations of the Split Window (SW) algorithm which exploit two TIR channels to compensate for the atmospheric effect and to calculate the surface temperature. A first step is to convert the radiance into brightness temperature using the Plank's law, then to apply the SW algorithm which however needs calibration and one more parameter: the emissivity of the surface considered.

Over the sea (sea surface temperature or SST), being the surface uniform, variants of the SW algorithms avoid the need of the emissivity knowledge, giving good results (e.g. you can apply the standard MODIS SST L2 product algorithm).

On the land (land surface temperature or LST) the calibration of the SW algorithm and the emissivity estimation is very critical for the accuracy to be obtained.

MODIS LST L2 product for example applies a Look Up Table associating fixed emissivity on the land cover types. Other solutions imply to retrieve emissivity from NDVI.

You can find a review of SW algorithms for LST sensors, including here: http://www.uv.es/ucg/articulos/2008/Publications_2008_6.pdf

Concerning Landsat you have only one TIR channel, so the SW cannot be applied and the atmospheric correction is more critical together with the still present emissivity problem. Here you can find an example from the same author before: http://www.uv.es/~ucg/articulos/2009/Publications_2009_1.pdf

In my experience the three sensors AVHRR, MODIS and Landsat have different sensitivity for the estimation of the surface temperature and so a direct numerical comparison (e.g. by resampling the latter) can give very different results. Moreover the temporal and spatial characteristics of AVHRR/MODIS in respect to Landsat makes then better for different applications.

Thank you very much for your answer. It was very clear. But i'd like als to know which channel of AVHRR can i use. Because, in thermal it has 3b, 4 and 5. What about that solution:

http://www.ncdc.noaa.gov/oa/pod-guide/ncdc/docs/klm/html/c7/sec7-1.htm

Thank you in advance

It depends on what surface temperature you would like to retrieve.

For Sea Surface Temperature (SST), the 3b band is used together with 4 and 5 ones for applying the Multi-Channel Sea Surface Temperature (MCSST) which can give very accurate results.

For Land Surface Temperature (LST) I believe that the most used algorithm is the Split-Window one in different variants, which uses bands 4 and 5.

Concerning the calibration of the L1b data into brightness temperature (i.e. blackbody temperature), the page you cited tells you how to do it, but consider that in most EO processing software, both commercial and free (e.g. SeaDAS, BEAM), this calibration procedure is included and very simple to apply.