Using satellite fluorescence data to drive a global carbon cycle model: Impacts on atmospheric CO2.

Friday, 19 December 2014: 5:00 PM
George James Collatz1, Joanna Joiner1, Stephan R Kawa1, Alvaro Ivanoff2,3, Yuping Liu4, Yasuko Yoshida5, Joseph A Berry6 and Grayson M Badgley7, (1)NASA Goddard SFC, Greenbelt, MD, United States, (2)NASA Goddard Space Flight Center, Lanham, MD, United States, (3)ADNET Systems Inc. Greenbelt, Greenbelt, MD, United States, (4)SSAI, Greenbelt, MD, United States, (5)NASA Goddard Space Flight Center, Greenbelt, MD, United States, (6)Carnegie Inst Washington, Washington, DC, United States, (7)Carnegie Institution for Science, global ecology, Washington, DC, United States
Atmospheric CO2 variability is markedly influenced by biospheric fluxes (photosynthesis and respiration) from the land surface at seasonal, to annual, to decadal time scales. Process models of photosynthesis and respiration have considerable uncertainty as only the sum of these fluxes can be constrained on the bases of atmospheric CO2 measurements alone. An independent proxy for photosynthesis or gross primary productivity (GPP) has recently become available from measurement of solar induced fluorescence (SIF). We report here on the first (to our knowledge) simulations of global atmospheric CO2 concentration driven by GPP estimated from observations of SIF. A baseline model uses satellite derived FPAR, incident solar radiation, temperature, and moisture stress scalars to estimate net primary productivity (NPP). The fluorescence driven model uses only fluorescence from GOME-2 scaled to the mean annual NPP at every grid cell and assumes a constant NPP/GPP ratio. Respiration was modeled identically in the two simulations. This preserves the spatial distribution of production capacity but allows for independent seasonal cycle and interannual variability from the baseline model. The flux models were run at ½ degree monthly resolution for 2007-2012 and fluxes were reaggregated along with fossil fuel and ocean fluxes to 3-hourly, 1 x 1.25 degree resolution for the atmospheric transport model. Here, we compare the model’s skill at predicting CO2 variability at 40 NOAA CO2 flask network sites. The baseline model shows good skill at matching the seasonal cycle at the flask sites but is not as good at producing monthly and interannual anomalies. The fluorescence model shows similar (or even improved) performance even though solar radiation, FPAR, precipitation and temperature effects on GPP are not included in the simulation. The results demonstrate the capability of the fluorescence data to integrate physiological and biophysical controls on GPP into a single measured parameter.