Further Study On Calibration Of Dimethysulfide Model Parameters In Barents Sea

In recent decades, due to large amounts of greenhouse gases, such as carbon dioxide emissions into the atmosphere, the Arctic sea surface temperatures(SST)rises, sea ice melt acclerated. Dimethysulfide(DMS) produced by marine biology emitted into the atmosphere through a series of changes will weaken the greenhouse effect. The paper studied distributions of Ice Cover(ICE), Wind Speed(WIND), Cloud Cover(CLOUD), Sea Surface Temperatures(SST), Chlorophyll(CHL)and, Mixed Layer Depth(MLD) in Barents Sea(30°E-40°E,70°N-80°N)during 2003-2009, MLD data is calculated based on the CTD data in Arctic Ocean. DMS observation data for year 2003-2005 is obtained by using BP neural network time series prediction data based on year 1998-2002 history DMS data in the study region. Simo's DMS formula is also used for comparisons. DMS model parameters calibration used genetic algorithm technique and obtained the optimal parameters of the region. Sensitivity analysis is then carried out for the same parameters. Finally, predictions the change of DMS flux used input data from CMIP5(5th Coupled Model Intercomparison Project)for 24 ×CO compared with 21×CO.The main conclusions are as follows: From the beginning of 2004, ICE strated to decline, in 2004-2007 years from August to September, the ice cover was reduced to 0%. In the southern region(30°E-40°E,70°N-75°N), it was almost ice free after 2004. CLOUD was more than 80% in general, only down to 75.4% in year 2004. WIND usually reached to minimum in July(about 5m/s), and maximum in winter(up to 12.5m/s). SST reached maximum in August. SST had a declining trend in the 7 years. Mean SST reached to the maximum(8.2?) in 2003, and down to the peak value of 5.9? in 2009. CHL reached to the peak(about 4mg/m3) in year 2009. MLD was more shallow in summer and deeper in winter. MLD had double peaks in January and March.Using BP neural network predicted DMS reached to the peak in year 2004( 11.77--12dumolm). Calibrated results fit better if using BP neural network predicting DMS time seires as observation data(comparing with using Simo's formula calibrated results). The peak of CHL precedes the peak of DMS 8-16 days, DMS-flux and Kw(transfer velovity) had their peaks almost at the same time. The sensitivity test showed that maximum sensitivity parameter is ?(phytoplankton S(DMSP): N ratio), followed by 28k(DMS consumption rate by Bacteria), Compared to 21×CO, the growth rates in 24 ×CO for SST, WIND and MLD are 233.684%?22.744%?28.232%. ICE would reduce 11.986%. CLOUD would only decrease by 1.322%. Under the condition of 24 ×CO, DMS air-sea flux would increase 148.840% and 249.679% with and without ICE respectively. The results are valuable for Arctic ecosystem and climate research.