Albert, what do you mean by NPP?

General  approach to estimate any bio- or physical or statistical parameneters is regression based. Except satellite data you should have ground measurements (data) for several reference points, and build a regression dependency between satellite data as regressor and parameters you are interested in (particularly, NPP).

See, for example

Conference Paper Validation of Global EO Biophysical Products at JECAM Test S...

I realise you are looking to apply this to higher resolution imagery.  You may be aware that it has been done at 1km using MODIS data.  The below document the algorythms used and may be useful to you in attempting a similar calculatiion on higher-resolution imagery.  Good Luck!

http://modis.gsfc.nasa.gov/data/atbd/atbd_mod16.pdf

Hello - Here are a few references on the topic that may be useful to add to the insights already offered. The Goetz paper uses Landsat; the Hilker paper provides a review that might be informative on the overall topic area. 

Article The use of remote sensing in light use efficiency based mode...

Article Remote sensing of net primary production in boreal stands