Inversion Of K_d(PAR) And Euphotic Zone Depth Of Typical Water Bodys In Northeast China With Remote Imagery

Inland waters plays an important role in human life and production.Lake eutrophication generally damages the aquatic ecosystem of inland waters. Euphotic zone depth(zeu) is an improtant property for aquatic ecosystem. It is the depth where 1% of the photosynthetic available radiation(PAR) just beneath the water surface remains, and is often quantified by the the diffuse attenuation coefficient of PAR(K_d(PAR)). It reflects the penetration of the light into the water, thus, determines the type and distribution of algae species and hydrophytes. Large amounts of lakes distributed in various geographical environment in Northeast China. They are eutrophic at different degrees. Regional scale monitoring of zeu with in situ measured PAR is expensive and time-consuming, also it is impossible to sample water body frequently at large extent. Based on the relationship between water leaving radiance and zeu, it can be estimated by remotely sensed imagery, which is cost-efficient. Landsat and MODIS have played a significant role in earth system observation. Thus, they can be applied to aquatic environment monitoring. With the sampling data achieved from several field experiments conduct in Northeast China during April to September, 2015, the determinant optical active constituents(OACs) for K_d(PAR) was determined through gray system relative analysis. Later, therelationship between OACs and K_d(PAR) was obtained by linear analysis.In order to derive the zeu from remote imagery at region scale, the images of Landsat/TM/ETM+/OLI and MODIS daily surface reflectance(MOD09GA ~500m, Bands1-7) were used to build the inversion model.The band combinations used to establish the model were selected by multiple stepwise regression. After applying the models to Landsat and MYD09 GA data that imaging on the same date, respectively, and calculating the average K_d(PAR) of different lakes. The consistency of the derived K_d(PAR) between Landsat and MODIS model was analyzed. The consistency between MYD09 GA and MODIS Aqua 8-Day composed surface reflectance(MYD09A1) was also assessed in the same way but by applying MODIS model to both MYD09 GA and MYD09A1. At last, the spatial distribution of zeu in Northeast China, September, 2015 was obtained by Landsat8/OLI, and the temporal variation of several large lakes’ K_d(PAR) in 2015 was depicted by MYD09A1. Some conclusions from this paper are as follows:(1) Total suspended material(TSM) is the main factor that affects the K_d(PAR). Three kinds of gray relevance calculated between K_d(PAR) and TSM are all bigger than the chlorophyll-a(Chl-a) concentration and chromophoric dissolved matter(CDOM). The linear regression between OACs and K_d(PAR) indicates that even though the relationship between K_d(PAR) and TSM is not the best for some particular lakes, it is generallywell for all lakes(R2=0.907, RMSE=0.702). The results demonstrate that TSM accounts for most of the lakes’ K_d(PAR) in Northeast China.(2) The landsat and MODIS all performed well in the derivation of K_d(PAR). The difference of Blue and Red combined with the ratio of Red and NIR were used in Landsat model, and the difference and ratio of Blue and Red were used in MODIS model. The accuracy of the two models was evaluated by performing 10 times 10-fold cross validation, and the results are R2=0.831±0.012, RMSE=0.952±0.017 and MRE=0.334±0.097 for Landsat, and R2=0.860±0.016, RMSE=0.910±0.024 and MRE=0.189±0.053 for MOD09 GA. The consistency of the K_d(PAR) derived from Landsat and MODIS model was evaluated and the result indicates that the MODIS model calculated a slight bigger K_d(PAR) than Landsat with slope=1.20 and R2=0.972. However, the K_d(PAR) derived from MYD09 GA and MYD09A1 was consistent with slope=1.04 and R2=0.966. The MYD09A1 is suitable for inversion of K_d(PAR) at regional scale.(3) The spatial distribution of K_d(PAR) in Northeast China derived from Landsat8 OLI revealed that there is a large difference of K_d(PAR)between different regions. The 8511 farm reservoir and Qingnian reservoir located in Sanjiang Plain have the biggest K_d(PAR) due to their high concentration of TSM. The K_d(PAR) of the lakes located in west of Songnen Plain is generally bigger than other regions. While the reservoirs in the eastern mountainous area of Northeast have a little K_d(PAR). Thetemporal changes of K_d(PAR) derived from MYD09A1 indicates that there is no consistent regularity of the dynamics. Some lakes have a lower K_d(PAR) at the wet season, while some lakes are on the contrary. And others have no obvious changes in one year. This may attribute to the different influence of the geographical environment factors.