Despite intensive research, carbon cycle is still under-investigated and not fully understood for many ecosystems, especially the ones which are of minor direct importance in terms of food and wood production. Karst ecosystems can be included in this group of ecosystems. The study was conducted at the Podgorski Kras plateau (400 - 430 m.a.s.l.), which was in the past subjected to agriculture land abandonment, in the sub-mediterranean region of Slovenia (SW Slovenia). Within the study area two study sites were chosen: grassland and forest succession. At both sites, in July 2008, an open-path Eddy covariance (EC) was installed at 15 m and 2 m height for Succession site and Grassland site, respectively. Also Rs and other auxiliary measurements were performed. For the observed period (July 1th 2008 - November 30th 2012) no major differences were measured between Grassland site and Succession site concerning air temperature and precipitation. Concerning the EC data quality for the observed period, 59.7% and 33.7% of expected data have not been discarded for Succession site and Grassland site, respectively. On the average annual basis Succession site was net sink of carbon (NEE = -184 Ž 19 gCm-2y-1) while Grassland site was a source of carbon (NEE = 293 Ž 34 gCm-2y-1). Based on the eddy covariance measurements it can be concluded that overgrown area increased sink activity compared to the extensive grassland in observed period. After B4 correction was applied on our datasets cumulative NEE fluxes changed. For observation period Succession site shifted to weak sink (-28 gCm-2y-1) of carbon, while the Grassland site remained a source (456gCm-2y-1). Applying SISC correction NEE changed to -127 gCm-2y-1 and 362 gCm-2y-1 for Succession site and Grassland site, respectively. Our measurements showed the need for self-heating correction also for our ecosystems. It is more appropriate to use site specific self-heating correction based and developed on own measurements than those suggested by Burba et al. (2008). Concerning the Rs measurements it can be concluded that knowledge of temporal variability can be greatly improved with an automatic system. Corresponding measurements of soil temperature and moisture together with manual Rs measurements, improved the knowledge about spatial variability of Rs.