Modelling gross primary production in semi-arid regions: effects on carbon uptake of intensive agriculture in southern Kenya

Sammanfattning: Background and aim: Gross primary production (GPP) is the largest global carbon (C) flux and an important component for counteracting anthropogenic CO2 emissions, understanding vegetation dynamics, and sustaining universal human standards. Africa plays a prominent role in the global C cycle, though our understanding of GPP dynamics is largely hampered by a paucity of ground-based observations. By using flux data from a newly installed eddy covariance (EC) flux tower in southern Kenya and high-resolution Sentinel-2 satellite data, the aims of this thesis were to i) determine the net C flux and climatic factors governing GPP at a high-intensively cropped dryland farm, ii) use the dataset to test different remote sensing driven GPP models and the MOD17 GPP product (collection 6), and iii) assess the difference in annual GPP among croplands and shrublands, grasslands and open forests. Results: The studied cropland had a net ecosystem exchange (NEE) of 170 g C m-2 between March and December 2019, thus acting as a source of C during the measurement period. Climatic factors governing GPP were vapour pressure deficit (VPD), soil moisture and air temperature (Tair). Using the 2-band enhanced vegetation index (EVI2), VPD and Tair as input in a multiple linear regression yielded the best results to in situ GPP (R2=0.65, RMSE = 2.02 g C m-2 d-1). Croplands had a higher mean annual GPP compared to grasslands, shrublands and open forest in the sub-humid to semi-arid agro-climatic zone, and a similar annual GPP to open forest in the semi-arid agro-climatic zones. The studied field at the Ausquest farm had however a considerably lower GPP compared to both natural lands and other croplands. Conclusion: Based on these results, land cover changes occurring over the vast savannah landscape in southern Kenya are not likely to reduce GPP in the region, however, more data is needed to make a more robust assessment. The studied cropland at the Ausquest farm emitted considerable amounts of C during the measurement period, indicating the importance of accounting for respiration in land-cover change studies. The ambiguity of the GPP models can be attributed to the persistent cloud cover during the growing season, as well as difficulties in accounting for the influence of anthropogenic effects on GPP, and to little ground-based data for calibration and validation.