?
On calibration of remote sensing retrievals of ecosystem respiration (Reco) with tower measurements over,Russian forests and wetlands
Shabanov N., Kuricheva O., Kurbatova J., Bartalev S., Endsley K. A., Kivalov S., Dyukarev E., Koshelev A., Kurganova I., Kuzenko A., Mamkin V., Maximov T., Miglovets M., Panov A., Petrov R., Prokushkin A., Varlagin A., Zagirova S., Zyryanov V.
The carbon balance of an ecosystem is the difference between Gross Primary Productivity (GPP) and Ecosystem Respiration (Reco) as expressed by Net Ecosystem Exchange (NEE). While remote sensing retrievals of GPP have reached maturity, Reco estimation remains underexplored and ultimately cast bias on NEE. Here we present an end-to-end multi-scale analysis of the mechanism of Reco, implemented at the level of eddy covariance tower measurements and scale it up to country level by means of remote sensing. Study has been performed over Russian forests and wetlands. Carbon flux data were collected from 20 RuFlux and 36 FLUXNET sites. We implemented three Reco regression models: 1) linear, forced by daily GPP, 2) Arrhenius exponential, forced by daily average Air Temperature, 3) linear combination of the above. Each model free parameters were parameterized as function of annual GPP, to account for spatial and temporal variability. Best performance is achieved by the third model, while the first two serve as components accounting for respiration of vegetation and soil, respectively. Daily and annual GPP/Reco/NEE products were generated from MODIS remote sensing and MERRA2 reanalysis data. Reco exhibits lower amplitude and more inert spatial variability compared to that of GPP. At the annual scale Russian forests and wetlands serve as carbon sink. All three annual carbon fluxes have tendency to increase (by absolute value) from north to south. WET have GPP close to Reco and therefore may change NEE sign, while DBF have GPP significantly prevail Reco to stay as a carbon sink.