| Carbon is the most important component of ecosystem,algae carbon refers to the carbon in the cells of planktonic algae,photosynthetic carbon sequestration of planktonic algae in inland lakes is one of the important process of carbon sink and is the main embodiment of the biological pump function,the strength of the carbon cycle and the process has the specificity and complexity.In the ecosystem,algae carbon has an irreplaceable role in the carbon cycle and mitigation of climate warming,and is crucial to the development of the Earth system model.Therefore,the accurate estimation of algae carbon’s current amount helps to better understand,evaluate,and monitor the carbon dynamics of water bodies,which is of great significance for studying the temporal and spatial changes of primary water production.Marine primary productivity drives marine bio-carbon pumps,while stimulating other marine biological productivity.China is the representative of water resource country,the algae carbon in inland water is of great significance to regional ecological system and climate change.The development of remote sensing technology makes it possible to monitor inland lake waters in a macro,rapid and real-time manner,which can reflect the spatio-temporal difference and seasonal variation of algae carbon distribution.Taking Taihu Lake,a typical inland eutrophic lake as an example,the relationship between the algal carbon and the intrinsic optical quantity(phytoplankton absorption coefficient)was studied.On the basis of this,a semi-analytical model of algae carbon concentration estimation based on Sentinel-3 OLCI image was constructed.At the same time,this paper constructed an empirical model of algal carbon remote sensing estimation based on the relationship between algal carbon and the apparent optical quality(remote sensing reflectance).The main research contents and conclusions include the following three aspects:(1)Construction of semi-analytical model for retrieval of algae carbon concentrationAlgae carbon is derived from phytoplankton and is part of the particulate organic carbon POC,while the remote sensing inversion of POC is relatively mature.Therefore,we try to separate algae carbon from POC based on the phytoplankton absorption coefficient.At 443 nm as a result of chlorophyll and carotenoid,the absorption coefficient of planktonic algae in the whole band show that the maximum value,form a shoulder peak,while under the influence of least pigment absorption at 550 nm,form a valley,difference of the two can maximize the instructions planktonic algae content,also can indirectly indicate the carbon content of planktonic algae,and the total particulate matter at 675 nm absorption coefficient can signify the content of organic matter.Therefore,the optical index(aph(λ1)-aph(λ2))/ap(λ3)was constructed for algae carbon:POC content.Based on the absorption coefficient of planktonic algae,the optical index of different forms was constructed and the band was iterated,Finally,the optimal optical index of algae carbon:POC was found to be(aph(443)-aph(550))/ap(675).Finally,use an improved QAA inversion algorithm which Mishra et al(2014)proposed for turbid blue-green algae waters to invert optical indices(aph(443)-aph(550))/ap(675),combining the inversion models of algae carbon:POC and POC,the root mean square error of the semi-analytical model for inverting algae carbon concentration was 0.78 mg/L,and the average relative error was 41.13%.The data set of the semi-analysis model is too small and the error propagation of each sub-model will be the two most important reasons leading to the poor final inversion effect.For the Taihu Lake in the study area,firstly increase the data volume,For the empirical parameters in the model,a relatively stable value is determined.Second,the steps and parameters in the semi-analytical model need to be simplified,and a more succinct and stable model is established.(2)Construction of empirical model for algae carbon concentration inversionThe field remote sensing reflectance can be simulated to Sentinel-3 OLCI image.The relationship between algal carbon concentration and simulated remote sensing reflectance was analyzed.The correlation coefficient between algal carbon concentration and remote sensing reflectance in 400-900nm ranges from-0.43 to 0.07.Near 490 nm,the algal carbon showed the highest negative correlation with remote sensing reflectance,with a correlation coefficient of-0.43.Taking into account that the algae carbon is derived from phytoplankton,The simulated remote sensing reflectance were operated by a combination of band ratio,NDVI,the fluorescence line height FLH,Maximum chlorophyll index MCI,triband,etc.The results showed that the inversion effect of the fluorescence line height FLH is the best,the root mean square error of the model inversion is 0.40mg/L,and the average relative error is 39.36%.Choose simulated remote sensing reflectance in 490 nm,510 nm,560 nm,620 nm,665 nm,681 nm which had good correlation with algae carbon concentration and the fluorescence line height(FLH)was selected as the input parameter of random forest.The root mean square error of model inversion after training is 0.28 mg/L and the average relative error is 31.41%.(3)Spatial and temporal distribution of algae carbon concentration in Taihu Lake regionThe random forest algorithm can be used to estimate carbon concentration in Taihu lake region,The results showed that the distribution of algae carbon concentration in Taihu Lake in spring showed a relatively high trend in the north and the center of the lake,lower in the south;The concentration of algae carbon in the Taihu Lake in the fall shows a high spatial distribution in the northwest and low in the southeast. |
