Assesment of СО2 fluxes in the typical steppe of southern Western Siberia

Abstract

For the first time, net ecosystem exchange (NEE) and ecosystem respiration (Reco) were measured using the chamber method within the typical steppe of the Kulunda Plain, West Siberia, (Altai Krai), and gross primary production (GPP) was calculated. Тo identify the relationship between CO₂ fluxes and environmental factors, the following parameters were also measured: aboveground phytomass (AGP), photosynthetically active radiation (PAR), air temperature, soil temperature and soil moisture. Тhe median (1Q, 3Q) NEE, Reco, and GPP were -103 (-152, -66), 90 (74, 105), and -200 (-251, -151) mg C m-2 h-1,

respectively. Тhe key factors of variability in NEE and GPP were aboveground phytomass, while for Reco - soil moisture content in the 100 cm layer. Based on the obtained data, a neural network was successfully trained using the Levenberg-

train

 

valid

 

test

 

Marquardt algorithm (R2         0.9, R2              0.9, R2            0.8 MSEtrain       335,

MSEvalid   647, MSEtest   360 Ntrain   50 Nvalid    11 Ntest    11), which will be used in the future to scale carbon fluxes in space and time.