Algorithms that establish relationships between variables obtained through remote sensing and geographic information system (GIS) technologies are needed to allow the scaling up of site-specific CO2 flux measurements to regional levels. We obtained Bowen ratio-energy balance (BREB) flux tower measurements during the growing seasons of 1998-2000 above a grassland steppe in Kazakhstan. These BREB data were analyzed using ecosystem light-curve equations to quantify 10-day CO2 fluxes associated with gross primary production (GPP) and total respiration (R). Remotely sensed, temporally smoothed normalized difference vegetation index (NDVIsm) and environmental variables were used to develop multiple regression models for the mapping of 10-day CO2 fluxes for the Kazakh steppe. Ten-day GPP was estimated (R 2 = 0.72) by day of year (DOY) and NDVIsm, and 10-day R was estimated (R2 = 0.48) with the estimated GPP and estimated 10-day photosynthetically active radiation (PAR). Regression tree analysis estimated 10-day PAR from latitude, NDVIsm, DOY, and precipitation (R2 = 0.81). Fivefold cross-validation indicated that these algorithms were reasonably robust. GPP, R, and resulting net ecosystem exchange (NEE) were mapped for the Kazakh steppe grassland every 10 days and summed to produce regional growing season estimates of GPP, R, and NEE. Estimates of 10-day NEE agreed well with BREB observations in 2000, showing a slight underestimation in the late summer. Growing season (May to October) mean NEE for Kazakh steppe grasslands was 1.27 Mg C/ha in 2000. Winter flux data were collected during the winter of 2001-2002 and are being analyzed to close the annual carbon budget for the Kazakh steppe. ?? 2004 Springer-Verlag New York, LLC.