Red Tide Remote Sensing Monitoring Based On MODIS And HJ-1 Data

With quickly accessing, wide coverage, syncing with the Earth and other advantages, MODIS and other remote sensing data have been good used in ocean dynamic monitoring, ocean productivity evaluation, and related areas. In recent years, small-scale red tides frequented in Chinese coast. MODIS image has 1 kilometer spatial resolution, and it can’t meet our needs to monitor small-scale red tides. HJ-1 image has a spatial resolution of 30 meters, so it can make up for the shortcomings of MODIS image. In this paper, experiments were carried out in the area of Xiamen and Wenzhou coast with MODIS and HJ-1 data. The study was developed on small-scale red tide monitoring. The main contents were summarized as follows:(1) Extraction of red tide water characteristics and analysis on temporal variation of red tide water. Using temporal MODIS data before and after the red tide to extract chlorophyll a concentration (chl-a), Sea surface temperature (SST), the ratio of reflectivity (Rrs) and normalized water-leaving radiance (nLw). Then we analyzed the temporal variation of these red tide characteristics. Chl-a、SST、the ratio of Rrs and nLw in red tide water are bigger than non-red tide water. Water characteristic parameters changed significantly when red tide happened, with a clear inflection point. Chl-a remained relatively unchanged before red tide, increased rapidly in red tide and steadily declined after red tide. SST changed smoothly during red tide. The ratio of Rrs and nLw had maximums when severe red tide occurred. Chl-a is the best feature in red tide monitoring. Followed by the ratio of Rrs and nLw, SST can only be used to auxiliary judge red tide.(2) Construct of chl-a inversion model for Xiamen coast based on HJ-1 satellite images and model verification. Experimental data includes HJ-1 data and measured data of water spectral and water chl-a. We estabished chl-a inversion localized model: chla=100.15-3.48r,R=log10 (b1/b2), referred to ten exponential model. We carried out accuracy verification with the measured data of water chl-a, correlation coefficient was 0.8147. By this model, we inversed the chl-a of 2011 in time-series and it provided a higher spatial resolution data for red tide monitoring.(3) Comparison of chl-a and spectral normalization. We compared chl-a of MODIS with chl-a of HJ-1 CCD. The chl-a of MODIS data by experience inverse model was higher than actual value. The chl-a of HJ-1 CCD data by ten exponential model was quite accurate, but it had a incomplete time-series. We used spectral normalization to eliminate the difference of two data. But it was ineffective in improving the consistency of the two data.(4) Feature level fusion of chl-a and analysis. We used adaptive weighted averaging method with enhanced spatial and temporal adaptive reflectance fusion model (E-STARFM) to fuse chl-a of MODIS and HJ-1 CCD data. Adaptive weighted averaging method improved effective spatial coverage of data, but it failed to improve data spatial resolution. The result of E-STARFM basically remained the spatial resolution of HJ-1 CCD. It also maked up the missing time of HJ-1 CCD. We extracted information of red tide based on chl-a in time-series. Results showed red tide information was consistent with historical events in occurrence time, fade time, location, and scale of red tide.