Interannual variability in terrestrial carbon exchange using an ecosystem-fire model and inverse model results Sergey Venevsky (1), Prabir K. Patra (2), Shamil Maksyutov (3), and Gen Inoue (3) (1) Obukhov Institute for Atmospheric Physics, Moscow , Russia (2) Frontier Research Center for Global Change/JAMSTEC, Yokohama , Japan (3) National Institute for Environmental Studies, Tsukuba , Japan
Outline of the presentation 1.Modelling of land-atmosphere carbon fluxes at global and regional scale – dynamic global vegetation model SEVER. 2.Description of human and lightning induced fires at global scale –SEVER-FIRE 3.Monthly-mean CO 2 fluxes from 11 land regions using a time-dependent inverse (TDI) model. 4.Comparison of the regional net carbon exchange fluxes simulated by TDI and SEVER.
SEVER dynamic global vegetation model (Venevsky and Maksyutov, 2005) Precip Temp, SWR H2OH2O Precip convect CO 2 Plant functional types distribution NEE=Rh+C fire-NPP C fire Rh NPP Daily time step 0.5°x0.5° grid cell Vegetation composition FPC Annual time step
Technical realisation of SEVER – modification of the LPJ DGVM to a daily time step. State-of-the-art dynamic global vegetation model LPJ DGVM (Sitch, et.al., 2003, Thonicke et.al., 2001, Venevsky et.al., 2002) was taken as a basis for technical realisation of SEVER. Advantages of the LPJ DGVM: modular stucture with identified processes of vegetation dynamics and soil/biosphere biogeochemistry successfuly reproduces vegetation composition and vegetation/soil carbon pools and fluxes at global scale Code is avaiable on the Internet Disadvantage: Pseudodaily approach Temp day T month 1 GPP day 1 gpp midmonth DESIGN OF SEVER (Venevsky, Maksyutov, 2005): Modification of LPJ modules from month/mid-month to a daily step New radiation routine New soil temperature routine New fire model
SEVER/DGVM – Technical details Temperature Precipitation Shortwave radiation SOIL classification data: 9 classes at 0.5°x0.5° (Sitch et.al., 2003) CLIMATE data: 6-hr NCEP/NCAR reanalysis data for at T62 resolution were interpolated to 0.5°x0.5° with correction to elevation. a) No fire emissions considered:b) complete set Minimum temperature Maximum temperature Precipitation Shortwave radiation Convective precipitation
Comparison of SEVER DGVM with LPJ DGVM (new version) Correlations (r) between the observed and calculated monthly-mean NEE for all sites SEVER: r 2 =0.75 LPJ (new version) : r 2 =0.51
SEVER-FIRE global mechanistic fire model Day Fire moisture extinction level lat lon Fire weather danger Flammability threshold Spread and termination Lightning and human ignition Carbon fire emission
Human induced ignitions: conceptual scheme & Wealth status * Population density * Urban/ rural * Timing * accessibility
Lightning ignitions: conceptual scheme Cumulonimbus Elevation Moisture Duff Moisture Duff LPC/LCC flashes Smoldering probability
Stepwise validation of SEVER- FIRE model (1) Number of cloud-to-ground (CG) flashes is validated using data of the Optical Transient Detector for the continents (Christiensen et.al, 2002) Number of lightning/human fires is validated using data for Canada (Stocks, et.al., 2002) and Spain (Vasques, et.al.,2002)
Step validation of SEVER- FIRE model (2) Areas burnt for Canada (Stocks et.al., 2002), Spain (Vasques et.al., 2002), Africa (Barbosa et.al., 1998). Examples of complete step validation for North-West Alberta, Canada (Wielgolaski et.al., 2002), for lightning fires:
Annual Flux ( ) Flux Anomaly ( ) Lightning fires Human induced fires
Comparison SEVER-FIRE vs CASA estimates, based on satelitte derived area burnt and CO SEVER CASA (van der Werf et.al., 2004)
Comparison of SEVER/DGVM fire flux with MODIS fire counts (seasonal cycle and spatial pattern) gC/m2/mon
Simulated global CO 2 fire emission during (human and lightning cases) El Niño Total averaged for annual fire emissions 3581 TgC (3530 TgC for ,Randerson, 2005)
Time-dependent Inverse Model (64 regions): Monthly fluxes (S) and associated covariance (C S ) are calculated as: (1) (2) G = Transport model operator, D = atmospheric CO2 data, C D = Data Covariance Patra et al., GBC, October 2005 Rayner et al., 1999 Gurney et al., 2004
Overview of TDI Results – Global and hemispheric scale CO2 flux anomaly Patra et al., GBC, October 2005 Patra et al., GBC, July 2005 Anomaly = monthly fluxes – mean seasonal cycle
Climate control on regional CO2 flux anomaly (Patra et al., GBC, 16 July 2005) Flux Anomaly Region/PC Climate Oscillation IndicesMeteorology ENSONAORainTemp. Temp North America−0.41 (1)−0.47 (2)−0.32 (3)0.33 (4) Trop South America0.48 (0)−0.20 (0)−0.47 (4)0.57 (0) Tropical Africa0.71 (3)−0.17 (0)0.42 (4)0.36 (0) Boreal Asia0.21 (5)−0.36 (5)0.17 (5)0.23 (0) Tropical Asia0.61 (2)−0.18 (0)−0.63 (5)0.41 (2) Europe−0.18 (0)0.45 (5)0.18 (5)−0.47 (1)
Comparison of CO2 flux anomalies – TDI vs SEVER/DGVM ABAB
Regional Flux Anomaly ( ) : Europe Ciais et al., 2005 : 0.5 Pg-C for 2003
Processes associated with interannual CO 2 flux variability– TDI vs SEVER/DGVM ABAB
Seasonal Cycle and long-term means of CO2 fluxes – TDI vs SEVER/DGVM
Conclusions 1.Human induced fire, increased in last 20 yrs, carbon emission exceeds that from lightning, with a ratio 68/32, despite of small number of lightning fires (1/20 of human fire). 2.We simulated greater human induced fires during the El Niño event, and the emissions are highest in the tropics. 3.The ecosystem model simulated flux anomalies are fairly in phase and amplitude with those estimated using inverse modelling atmospheric CO 2. 4.However, there still exists significant disagreements between the inversion and ecosystem flux amplitudes at the regional scale.