Arctic ecosystems play a key role in the terrestrial carbon cycle. Our aim was to combine satellite-based normalized difference vegetation index (NDVI) with field measurements of CO2 fluxes to investigate changes in gross primary production (GPP) for the peak growing seasons 1992–2008 in Rylekærene, a wet tundra ecosystem in the Zackenberg valley, north-eastern Greenland. A method to incorporate controls on GPP through satellite data is the light use efficiency (LUE) model, here expressed as GPP = ɛpeak × PARin × FAPARgreen_peak; where ɛpeak was peak growing season light use efficiency of the vegetation, PARin was incoming photosynthetically active radiation, and FAPARgreen_peak was peak growing season fraction of PAR absorbed by the green vegetation. The ɛpeak was measured for seven different high-Arctic plant communities in the field, and it was on average 1.63 g CO2 MJ−1. We found a significant linear relationship between FAPARgreen_peak measured in the field and satellite-based NDVI. The linear regression was applied to peak growing season NDVI 1992–2008 and derived FAPARgreen_peak was entered into the LUE-model. It was shown that when several empirical models are combined, propagation errors are introduced, which results in considerable model uncertainties. The LUE-model was evaluated against field-measured GPP and the model captured field-measured GPP well (RMSE was 192 mg CO2 m−2 h−1). The model showed an increase in peak growing season GPP of 42 mg CO2 m−2 h−1 y−1 in Rylekærene 1992–2008. There was also a strong increase in air temperature (0.15 °C y−1), indicating that the GPP trend may have been climate driven.