Using flux networks to discover long term controls of ecosystem productivity

Climate, potential biota, topography, geological parent material, and time. The state factor-interactive-controls hypothesis is a pervasive concept in ecosystem ecology that could explain long term controls of eddy covariance estimates of gross primary productivity. The hypothesis is adopted by ecologists whenever a gradient analysis is used (a chrono-sequence or a climate gradient). In this framework the state factors of climate, potential biota, topography, geological parent material, and time control ecosystem processes but are not themselves influenced by ecosystem processes at local scales. Interactive controls like realized plant functional types, soil resources, microclimate and disturbance frequency influence and are influenced by ecosystem processes. Flux networks represent whole ecosystem process measurements and while much has been learned from analyzing short term controls, longer term controls have been investigated less often. The last two decades have seen the growth of the Ameriflux network in North America; similar measurements of ecosystem carbon, water and energy exchange made across a wide range of ecosystem types. We tested whether gross primary productivity, estimated using the eddy covariance method across more than 40 ecosystems in North America conformed to the State-Factor-Interactive-Controls hypothesis. To estimate state factors we combined satellite observations, digital elevation models, geological and soil maps and climate re-analysis. By limiting our analysis to sites with more than 10 years of data we were able to remove the effect of short-term direct controls (light, temperature, moisture etc) on gross primary productivity. We found significant interactive effects of climate and geological substrate and a strong direct effect of climate on average gross primary productivity. We also found a strong effect of biota on the variation that was not explained by state factors. Comparing these patterns to predictions from an Earth System Model we found contrasting results. These findings provide support for the state factors-interactive-controls hypothesis and suggest new opportunities for ecological synthesis using networks of ecological data.