Uncertainties in the model of soil organic matter pools dynamics ROMUL

Transcription

1 Romul.exe Uncertainties in the model of soil organic matter pools dynamics ROMUL A.Komarov 1, S.Bykhovets 1, O.Chertov, A.Mikhailov 1 1 Institute of Physicochemical and Biological Problems in Soil Science RAS Pushchino, Moscow region, Russia // Biological Institute, St. Petersburg State University, St. Petersburg, Russia

2 Flow chart of the EFIMOD system of models of forest growth and elements cycling PAR Climate Initialisation Available PAR for trees, ground vegetation and natural regeneration T R E E S 1 3. n Ground vegetation Natural regeneration Redistribution of soil available nitrogen Model of soil organic matter ROMUL Forest manager Data viewer Graph interface 3D visualisation

3 Modelling soil as a decomposing part of biological turnover in forest ecosystem : main problems How to split the soil organic matter into different pools, which are decomposing and humifying with different rates? How to evaluate the decomposition and humification rates for those fractions? How to assess the robustness of these coefficients concerning the accuracy of evaluation? What uncertainties arise at initialization of the model in relation to the suggested pools?

4 Theoretical basis of the ROMUL model Formalization of the classical concept of humus type (Humusform) The idea of three complexes of organismsdestructors influencing decomposition organic matter mineralisation and humification in dependence of soil temperature, soil moisture, litter nitrogen and ash content, C/N ratio in mineral topsoil specification of rate variables for above and below ground litter cohorts calculation of dynamics of organic matter and nitrogen during the decomposition with gross CO and available N evaluation //

5 Humus types formation

6 Humus types formation

7 Humus types formation

8 ROMUL flow chart k i 1L k i L k i 3L L i - Aboveground litter fall L i - Undecomposed litter on soil surface F.i - complex of humus substances with undecomposed debris (humified organic layer) Soil surface K j 1S K j S K j S K j L K j 5S L j u - Belowground litter fall L j u - undecomposed litter in mineral topsoil F j u - complex of humus substances with undecomposed debris in mineral topsoil ( labil humus ) H - humus bonded with clay minerals K K j 3S K j 5S

9 ROMUL input parameters: Above ground and belowground litter fractions (including CWD) - amount (by time step), nitrogen and ash concentrations Soil temperature and soil moisture (in mass per cent) Soil hydrological characteristics Pools of soil organic matter and soil nitrogen in organic horizons and mineral topsoil (1 cm layer): SOM of forest floor, labile SOM in mineral soil, stable SOM (humus) in mineral soil

10 Model realisation dl i /dt = L i -(ki 1L + k i 3L )Li, dl j u /dt = L j u -(kj 1S + k j 3S )L j u, df i /dt = k i 3L L i -(k i L + k i L + k i 5S )F i, df j u /dt = k j 3S L j u -(kj S (H)+ k j S + k j 5S )F j u,, dh/dt = δ Bact (Σ m k i L N i F +Σ n k j S N j Fu ) + δ Lumb (Σ m k i 5S N i F + Σ n k j 5S N j Fu ) - k H

11 Experimental basis of the model A set of laboratory experiments on the rate of decomposition of plant debris and soil organic matter of definite chemical composition in controlled conditions: previously published data and results of specially performed experiments in the Lab of Soil Biochemistry, St. Petersburg State University Pinus Sylvestris The coefficients have been 1. evaluated using exact solution Undecomposed matter, rel..... Fitted curve Experimentall points of the system of differential equations and Markquart method Then dependencies of the coefficients on nitrogen and ash contents, soil temperature and moisture have been found and. 1 3 inserted into the ROMUL model Time, days

12 Verification example: comparison of experimental data (Emmer, 1995) with ROMUL simulation 13 Litter CHS Humus Emmer's data SOM pools, years

13 Inputs: parameters and initial values We distinguish model parameters and state variables. In terms of computer code model parameters and initial values of state variables represent inputs of a model, whereas outputs consist of values of state variables at final and some intermediate state of the modeled system. Calibration is an adjustment (often determined by statistical estimation techniques ) of parameter values so that the model output estimates are close to the measured system output values. Additional problem is to find minimal number of parameters for calibration of initial data and reducing uncertainties. Their number must be small to avoid formal regression approximations. The calibration parameters usually are external in relation to the model and cannot be evaluated inside the model. In ROMUL they are proportion of labile and stable SOM in mineral soil and rate of decomposition of stable humus. Mentioned parameters could be evaluated in more comprehensive model, in our case in the model of carbon and nitrogen turnover in the forest-soil system EFIMOD.

14 Sources of uncertainties at simulation: Ambiguous splitting of SOM into pools Imperfect model structure Errors at coefficients and their dependencies evaluation Variability and uncertainty of initial soil data Variability of soil climate Variability and uncertainty of litter fall data: amount and quality, including fine roots production and litter

15 A soil climate generator is used in the ROMUL model for two purposes: as a method of evaluation of soil temperature and moisture using measured standard meteorological longterm data; For statistical simulation of realisations of long-term series of necessary input climate data with known statistical properties. 1 T S, C а 5 w weight,% б дек.7 дек.75 дек.7 дек.77 дек.7 дек.79 дек. дек.1 дек. дек.3 дек. дек.5 дек. дек.7 дек.75 дек.7 дек.77 дек.7 дек.79 дек. дек.1 дек. дек.3 дек. дек.5 дек. LONG-TERM DYNAMICS OF MEASURED (1) AND CALCULATED () SOIL TEMPERATURE (a) AND GRAVIMETRIC MOISTURE (б)

16 Central Forest Reserve Climate change scenarios from different models Air temperature Precipitation 3.75 E 5.5 N 3.75 E 5.5 N 1 13 obs erved Air temperature (annual) observed CGCM-A1Fi CGCM-A CGCM-B1 CGCM-B CSIRO-A1Fi CSIRO-A CSIRO-B1 CSIRO-B Ha dcm3-a1fi Ha dcm3-a Ha dcm3-b1 Ha dcm3-b PCM-A1Fi PCM-A PCM-B1 PCM-B Precipitation (annual), mm 1 CGCM-A1Fi CGCM-A CGCM-B1 CGCM-B CSIRO-A1Fi CSIRO-A CSIRO-B1 CSIRO-B HadCM3-A1Fi HadCM3-A HadCM3-B1 HadCM3-B PCM-A1Fi PCM-A PCM-B1 PCM-B Comparison of stable climate and HADCM3 scenario Central Reserve site Comparison of SOM total at maximal variation Central Reserve site Comparison of CHS above at maximal variation stat warm SOM of forest floor SOM of forest floor warm 1,5 1 15,5 15, years years

17 Uncertainties of litter fall data Central forest reserve (Tver region, Russia) uneven-aged spruce forest no variation Central forest reserve (Tver region, Russia) uneven-aged spruce forest litter 5% variation Forest floor Labil humus Humus Forest floor Labil humus Humus Months Months Uncertainties of climate variation Central forest reserve (Tver region, Russia) uneven-aged spruce forest no climate variation Central forest reserve (Tver region, Russia) uneven-aged spruce forest climate variation Forest floor Labile SOM Stable SOM months Forest floor Labile SOM Stable SOM months 55

18 Variability and uncertainty of initial soil data Central forest reserve (Tver region, Russia) uneven-aged spruce forest forest floor 1% variation Central forest reserve (Tver region, Russia) uneven-aged spruce forest humus % variation Forest floor Labil humus Humus Forest floor Labil humus Humus Months Months

19 Variability of the model coefficients of mineralisation and humification Central forest reserve (Tver region, Russia) uneven-aged spruce forest coefficients of humification, 3% variation Forest floor Labile SOM Stable SOM months Central forest reserve (Tver region, Russia) uneven-aged spruce forest coefficients of forest floor transformation, 3% variation Forest floor Labile SOM Stable SOM months Central forest reserve (Tver region, Russia) uneven-aged spruce forest rate of mineralisation of stable SOM, 3% variation Forest floor Labile SOM Stable SOM months Central forest reserve (Tver region, Russia) uneven-aged spruce forest all coefficients, 3% variation Forest floor Labile SOM Stable SOM months

20 Conclusions: From ecosystem point of view soil is an exponential filter, which smoothes outliers and forced oscillations at different scales Ordinary climatic and litter input variations do not affect model output The initialization of mineral soil variables and evaluation of rates of humification and mineralization of pool of stable SOM are main sources of uncertainties in model output Initial values of pools of SOM in forest floor and mineral soil could be evaluated on data for forest types (prehistory of a plot could be a crucial point in uncertainties arising) The relation between labile and stable SOM in mineral soil can be evaluated from data about coarse and fine roots productivity (or by Bayesian estimation) The rate of mineralization of stable SOM can be used as a calibrating parameter at forest ecosystem level: it must control rate of stable humus accumulation at early successive stages of forest ecosystem development and lead to steady state at climax stage, it could be done using known durations of stages of succession

21 Thank you for your attention!