Towards Bayesian uncertainty quantification for forest models used - PowerPoint PPT Presentation

Towards Bayesian uncertainty quantification for forest models used in the U.K. GHG inventory for LULUCF Marcel van Oijen & Amanda Thomson Centre for Ecology and Hydrology Edinburgh, U.K.

Contents 1. Current methodology used to make the LULUCF inventory 2. Research on alternatives: Forest models that include the effects of • climate & soil conditions Bayesian uncertainty quantification • 3. Conclusions & outlook

Primary information flows GHG emissions / C stock LUC soil C stock changes Forest C stock changes changes (Tier 3) (Tier 3) (Tier 2)

Primary information flows Country-specific data on EFs & activities: • Deforestation • Liming • Afforestation rates Land-use areas: • Lowland drainage (FC, NIDA) • CS, MLC • Peat extraction • Yield tables (FC) • NICS, AC & FS • Non-forest biomass F F C C G G S S F F EF F F C C G G S S C C F F EF G G C C Data Data S S G G S S Parameters Dynamic Parameters: soil model • Production CFLOW • Turn-over GHG emissions / C stock LUC soil C stock changes Forest C stock changes changes (Tier 3) (Tier 3) (Tier 2)

Uncertainty propagation Country-specific data on EFs & activities: • Deforestation Uncertain inputs (pdf’s) • Liming • Afforestation rates Land-use areas: • Lowland drainage (FC, NIDA) • CS, MLC • Peat extraction • Yield tables (FC) • NICS, AC & FS • Non-forest biomass F F C C G G S S F F EF F F C C G G S S C C F F EF G G C C Data Data S S G G S S Parameters Dynamic Parameters: soil model • Production CFLOW • Turn-over GHG emissions / C stock LUC soil C stock changes Forest C stock changes Uncertain outputs changes (Tier 3) (Tier 3) (Tier 2)

CFLOW Yield tables & Yield tables & expansion factors expansion factors • Afforestation rates (FC, NIDA) thinning thinning Wood Wood Woody Woody Non-woody Non-woody products products biomass biomass biomass biomass • Yield tables (FC) and and harvesting harvesting Natural mortality Natural mortality Thinnings Thinnings Harvest debris Harvest debris Data Woody Woody Non-woody Non-woody litter litter litter litter Parameters: Transfer of Transfer of residues to soil residues to soil • Production CFLOW Soil organic Soil organic • Turn-over matter matter Forest C stock changes (Tier 3)

CFLOW • Afforestation rates (FC, NIDA) CFLOW forest model : • Yield tables (FC) • Simple C-pools model with first order flows • Robust Data • Coarse (no effects climate & soil conditions) • Uncertainty quantification … Parameters: • Production CFLOW • Turn-over Forest C stock changes (Tier 3)

CFLOW: Uncertainty quantification by MC 1. Quantify pdf’s for all inputs 0.20 0.20 0.20 0.20 0.20 0.20 Initial C soluble Initial C soluble Initial C soluble Initial C starch Initial C starch Initial C starch 0.15 0.15 0.15 0.15 0.15 0.15 • Afforestation rates 0.10 0.10 0.10 0.10 0.10 0.10 0.05 0.05 0.05 0.05 0.05 0.05 (FC, NIDA) 0.00 0.00 0.00 0.00 0.00 0.00 • Yield tables (FC) -0.05 -0.05 -0.05 0.00 0.00 0.00 0.05 0.05 0.05 0.10 0.10 0.10 0.15 0.15 0.15 0.20 0.20 0.20 0.25 0.25 0.25 0.30 0.30 0.30 0.35 0.35 0.35 -0.05 -0.05 -0.05 0.00 0.00 0.00 0.05 0.05 0.05 0.10 0.10 0.10 0.15 0.15 0.15 0.20 0.20 0.20 0.25 0.25 0.25 0.30 0.30 0.30 0.35 0.35 0.35 2. Take sample Data from pdf’s Parameters: 3. Run model • Production CFLOW 1000 times • Turn-over Forest C stock changes (Tier 3)

BASFOR: process-based forest model BASFOR • Parameters (many!) Atmosphere Atmosphere inputs: • Input variables (many!) N-deposition N N C C H 2 O H 2 O BASFOR CO 2 N N Tree Tree Radiation H 2 O H 2 O N N C C H 2 O H 2 O Temperature Data Soil Soil Rain N N C C H 2 O H 2 O Humidity Subsoil Subsoil Wind speed CFLOW Forest C stock changes (Tier 3)

BASFOR: process-based forest model BASFOR forest model : • Parameters (many!) • Input variables (many!) • Preprocessor for CFLOW (effects of climate, N-deposition, [CO2]) BASFOR • … or replacement of CFLOW? • But is BASFOR robust enough ? • More inputs & parameters => more Data demanding uncertainty quantification CFLOW Forest C stock changes (Tier 3)

Process-based models BASFOR forest model : • Parameters (many!) • Input variables (many!) • Preprocessor for CFLOW (effects of climate, N-deposition, [CO2]) BASFOR • … or replacement of CFLOW? • But is BASFOR robust enough ? • More inputs & parameters => more Data demanding uncertainty quantification CFLOW Bayesian calibration & uncertainty quantification Forest C stock changes (Tier 3)

The Bayesian approach Uncertainty propagation Country-specific data on Country-specific data on EFs & activities: EFs & activities: • Deforestation • Deforestation Uncertain inputs (pdf’s) • Liming • Liming • Afforestation rates • Afforestation rates Land-use areas: Land-use areas: • Lowland drainage • Lowland drainage (FC, NIDA) (FC, NIDA) • CS, MLC • CS, MLC • Peat extraction • Peat extraction • Yield tables (FC) • Yield tables (FC) • NICS, AC & FS • NICS, AC & FS • Non-forest biomass • Non-forest biomass F F F F C C C C G G G G S S S S Bayesian F F F F F F F F C C C C G G G G S S S S EF EF C C C C F F F F EF EF G G G G C C C C Data Data S S S S Data Data G G G G S S S S calibration Parameters Parameters Dynamic Dynamic Parameters: Parameters: soil model soil model • Production • Production CFLOW CFLOW • Turn-over • Turn-over GHG emissions / C stock LUC soil C stock changes Forest C stock changes Uncertain outputs changes (Tier 3) (Tier 3) (Tier 2)

The Bayesian approach Uncertainty propagation Country-specific data on Country-specific data on EFs & activities: EFs & activities: • Deforestation • Deforestation Uncertain inputs (pdf’s) • Liming • Liming • Afforestation rates • Afforestation rates Land-use areas: Land-use areas: • Lowland drainage • Lowland drainage (FC, NIDA) (FC, NIDA) • CS, MLC • CS, MLC • Peat extraction • Peat extraction • Yield tables (FC) • Yield tables (FC) • NICS, AC & FS • NICS, AC & FS • Non-forest biomass • Non-forest biomass F F F F C C C C G G G G S S S S Bayesian F F F F F F F F C C C C G G G G S S S S EF EF C C C C F F F F EF EF G G G G C C C C Data Data S S S S Data Data G G G G S S S S calibration Parameters Parameters Dynamic Dynamic Parameters: Parameters: soil model soil model • Production • Production CFLOW CFLOW • Turn-over • Turn-over GHG emissions / C stock LUC soil C stock changes Forest C stock changes Uncertain outputs changes (Tier 3) (Tier 3) (Tier 2) Prior pdf for the parameters Likelihood of the data P( ϑ |D) = P( ϑ ) P(D| ϑ ) / P(D) Scaling constant Posterior pdf for the parameters ( = ∫ P( ϑ ) P(D| ϑ ) d ϑ )

The Bayesian approach P( ϑ |D) = P( ϑ ) P(D| ϑ ) / P(D) Bayes’ Theorem implemented using MCMC (Metropolis algorithm)