| Since the industrial revolution,the increase of anthropogenic carbon dioxide concentration in the atmosphere has become an important cause of environmental problems such as global warming and ocean acidification.At present,ocean acidification is changing the chemical structure of seawater and the marine environment on which marine organisms depend at an unprecedented rate.The pH value of surface seawater has been reduced from 8.21 before the industrial revolution to 8.1 now,and its scientific and socio-economic impact has attracted more and more attention.The ocean is considered to be a buffer zone for ocean acidification,and the gas exchange between its surface and atmosphere plays an important role in the study of marine carbon cycle,mitigation of greenhouse effect and regulation of climate change.With the increasingly serious problem of ocean acidification,it is particularly important to accurately evaluate the temporal and spatial changes of ocean acidification and its impact mechanism.On-site sampling can obtain accurate and reliable seawater acidification data,but this method has limitations such as lack of historical data,small space-time coverage and low resolution.There is a response relationship between seawater pH value,TA and seawater physicochemical parameters.The seawater pH value and TA can be obtained by parametric modeling.However,the measured data are limited by time and space,which makes it difficult to establish seawater pH value and TA model.Remote sensing data has the advantages of wide coverage,strong timeliness and high spatial resolution,which provides a way to expand the observation and analysis of the temporal and spatial changes of ocean acidification.The Northwest Pacific has more tropical cyclones than other sea areas,which is the most active sea-air interaction and the highest annual average temperature of the global ocean.Therefore,in this paper,the Northwest Pacific Ocean is taken as the research area,and the measured data(surface seawater pH value,total alkalinity(TA))and remote sensing data(sea surface temperature(SST),sea surface salinity(SSS),sea surface chlorophyll a concentration(Chla))are combined.Linear regression,BP neural network and random forest are used to construct the remote sensing inversion model of surface seawater pH value and TA.The model is evaluated qualitatively and quantitatively and its applicability is analyzed.The best model is used to study the temporal and spatial variation characteristics of surface seawater pH and TA in the Northwest Pacific Ocean.The interannual and seasonal variations of surface seawater pH and TA from 2003 to2020 were analyzed and their influencing mechanisms were briefly discussed.The main research contents and results are as follows.(1)Quantitatively,the model based on BP neural network algorithm is the best.The determination coefficient of pH is 0.6952,the correlation coefficient is 0.8338,and the root mean square error is 0.0182.The determination coefficient of TA is 0.8687,the correlation coefficient is 0.9320,and the root mean square error is 9.4755.The model based on random forest method was the second.The R~2,R and RMSE of pH modeling were 0.647,0.8126 and 0.019,respectively.The R~2,R and RMSE of TA modeling were0.851,0.9243 and 10.387,respectively.The inversion model based on linear regression method has the lowest accuracy.The linear equation corresponding to pH modeling is8.209-0.006SST+0.001SSS-0.001Chla,R~2 is 0.6356,R is 0.7349,RMSE is 0.0310.The linear equation corresponding to TA modeling is 304.995-1.105 SST+57.713SSS-0.256Chla,R~2 is 0.8542,R is 0.8960,RMSE is 11.7167.(2)Qualitatively,the pH value of the surface water in the Northwest Pacific Ocean estimated by the BP neural network algorithm has the smallest deviation from the pH value of the Copernicus website and the highest degree of agreement,followed by the random forest.The pH value of the Copernicus website based on the linear regression method has the largest difference in the spatial and temporal distribution;the surface seawater TA in the Northwest Pacific Ocean estimated by BP neural network algorithm has the best agreement with the TA in the Ocean SODA-ETHZ dataset,with the smallest deviation,followed by linear regression,and the TA estimated by random forest method has the worst effect.In addition,in winter and summer,the distribution characteristics of pH and TA retrieved by the optimal BP neural network model are consistent with the existing research.In summary,it is feasible to integrate the measured and remote sensing data into the estimation of ocean acidification,and the pH and TA models based on BP neural network algorithm have the highest accuracy and are more suitable for the Northwest Pacific Ocean.(3)The best BP neural network model was used to invert the Northwest Pacific Ocean from 2003 to 2020,and it was found that the surface seawater pH and TA had obvious temporal and spatial variation characteristics.Temporally,the surface seawater pH and TA showed obvious interannual and seasonal variation characteristics.Over the years,the pH value showed a downward trend,as follows:winter>spring>autumn>summer;the overall trend of TA showed a gentle upward trend,and the season also showed winter>spring>autumn>summer.Spatially,pH is affected by temperature,salinity and chlorophyll concentration.In spring,autumn and winter,it shows a ladder-like distribution of high in the north and low in the south.In summer,the center is slightly lower than the surroundings.TA is mainly controlled by the conservative parameter salinity.The spatial distribution of TA in spring,summer,autumn and winter is relatively consistent,showing a gradual increase from west to east,and a decreasing trend from 30°N to north and 16°N to south.The maximum vortex appears near 25°N,about 2350μmol/kg.(4)In this paper,the spatial and temporal distribution of average sea surface temperature,salinity,chlorophyll,pH and TA from 2003 to 2020 and the correlation between pH,TA and temperature,salinity and chlorophyll in the ocean center,equatorial sea area and offshore sea area are analyzed.It can be seen that the occurrence and development of surface seawater acidification in different sea areas are affected by different environmental factors and show different spatial distribution characteristics.In the center of the ocean,due to the relatively stable nature of seawater,the correlation between seawater pH and temperature is the strongest;in the equatorial and offshore waters,the correlation between pH and salinity is the strongest.Due to the main control of salinity,TA has the strongest correlation with salinity in each sea area. |
