Use of different model + shell combinations for Inverse Modelling with the same data set. Bernhard Gottesbüren (BASF) Results and Resume

Transcription

1 Shell+Model Combinations for a data set Bernhard Gottesbüren (BASF) Use of different model + shell combinations for Inverse Modelling with the same data set Data set PEST+PELMO SUSE + PESTRAS Results and Resume S2 L13+16

2 Description of the Data-set and shell+model DATASET 4 lysimeters at site A (years 1-3) + 2 lysimeters at site B (years 4-5) same crop, same soil, but different weather, different application rate SHELL+MODEL COMBINATION PEST (Levenberg Marquardt)+ PELMO 3.0 SUSE (Simplex) + PESTRAS 3.1

3 Simulation of the water recharge with PELMO Kumulative Sic ke rwasse rbildung (Simulation und Lysime te r) PELMO Sta nd a rd Lysim e t e r 1 5 PELMO O p timie rt Lysim e t e r 1 6 Lysim e t e r 1 2 Lysim e t e r /07/92 31/01/93 19/08/93 07/03/94 23/09/94 11/04/95 28/10/95 Ze it SITE A Sickerwasser L/m² 1200 Kumulative Sic ke rwasse rbildung (Simulation und Lysime te r) SITE B Lysim e t e r 1 3 Lysim e t e r 1 6 PELMO PELMO Sta nd a rd Tage nach Beginn der Simulation

4 Simulation of the water recharge with PESTRAS accumulated water recharge [mm] measured lys. 1 measured Lys. 2 measured Lys. 3 measured Lys. 4 simulated with PESTRAS SITE A Simulation days after 01/ water recharge [mm] measured lysimeter A measured lysimeter B Simulated with PESTRAS SITE B

5 PEST + PELMO example Measured at Site A Simulated with optimised parameters at Site A Da ys a fter sta rt of sim ulation exp. opt. 0 Concentration [µg/l]

6 PEST + PELMO example Measured at Site B Simulated with optimised parameters at Site B exp. opt. 0 Wirkstoffkonzentration [µg/l]

7 PEST + PELMO cross validation Measured at Site A Simulated with optimised parameters at Site A Days after start of simulation exp. opt. IUCT13 S2 L13+16 IUCT16 0 Concentration[µg/L] Cross validation with optimised parameters from site B

8 Optimisation Step (concentrations single applied lysimeter) measured (L12) measured (L16) uncalibrated (Lab. Para.) optimised Parameter Concentration in Percolate [µg/l] Days after 01/01

9 Optimisation Step (fraction percolated) measured L12 measured L16 simulated lab. parameter simulated with optimised parameters Fraction percolated [ - ]

10 Validation Step ( independent repeated applied lysimeter) measured repeated applied lysimeter simulated lab. Parameter simulated inv. mod. Parameter to Lys. A simulated inv. mod. Parameter to Lys. B Concentration in Percolate [µg/l] days after 01/01

11 Example for a validation/extrapolation step of inverse modelled parameters measured L13 measured L16 simulated Lab. parameters simulated inv. mod. Parameters DT 50 /KOC = 23.3/80.7 simulated inv. mod. Parameters DT 50 /KOC = 28.4/98.5 Sickerwasserkonzentrationen [µg/l] Tage nach

12 PEST + PELMO vs. SUSE+PESTRAS optimised parameters Parameter DT50 KOC [d] [dm³/kg] Laboratory Parameter PESTRAS PELMO

13 Resume simulations with PELMO At both experimental sites nearly identical KOC values but different DT50 were obtained Water recharge could be calibrated fairly well Optimised KOM-values were similar in all lysimeters but Optimised DT50-values were different In 50% of the studies the peak concentrations in the first recharge period could not be simulated correctly with PELMO (although water percolation pattern seemed to be correct) Crossvalidation was only partially successful -> Problem to simulate exact breakthrough curves of reactive substances with capacity type of model -> the required fine resolution of water transport on a daily basis can not be expected from a capacity model, (which on the other hand provides good prognoses of the long-term behaviour of pesticides)

14 Resume simulations with PESTRAS Optimised DT50 are in the range of measured field DT50 Optimised KOM are higher than measured K OM Adsorption but equal to measured K OM Desorption. The water recharge and the breakthrough of the compound could be simulated fairly well in all 6 lysimeter studies The laboratory parameter combinations were unsuitable The BTC of the compound in lysimeters at site B could be simulated very well with optimised parameters from site A