Recently, several issues involving phytoplankton communities have emerged in Lake Biwa. However, it is difficult to estimate detailed dynamics of phytoplankton distribution by only ship observation in Lake Biwa, the largest lake in Japan. Therefore, in this study, Moderate-resolution Imaging Spectroradiometer (MODIS) satellite data were used to estimate chlorophyll a concentrations in Lake Biwa. Moreover, we developed a method for improving the accuracy of chlorophyll a concentration (Chl_sate) estimates using MODIS data. In-situ chlorophyll a concentrations and remote sensing reflectance values were measured from the littoral zone to offshore in the northern basin of Lake Biwa from 2012 to 2018. Remote sensing reflectance data (Rrs_sate: Level-2) from the MODIS instrument aboard the Aqua platform were used to estimate Chl_sate. The Chl_sate data obtained using the ocean chlorophyll three-band algorithm for MODIS (OC3M) tended to be overestimated and inaccurate (NMB: 708.7%, RMSE: 76.7mgm^－3), mainly because Rrs_sate tended to be underestimated in the short-wavelength region. These results indicated that the standard NASA atmospheric correction algorithm could not be employed to estimate the chlorophyll a concentrations in the inland water body of Lake Biwa. Further, the reduced accuracy of the Chl_sate estimates could also be attributed to an error in the OC3M algorithm. We therefore developed a correction method for Rrs_sate (488) and Rrs_sate (547) and optimized the coefficients of the OC3M algorithm using in-situ data. Consequently, the accuracy of the corrected Chl_sate (NMB: 0.79%, RMSE: 2.14mgm^－3) values estimated using our method were greatly improved compared to the uncorrected Chl_sate values.

Many studies have been conducted on satellite derived bathymetry in shallow water, and most of them use training bathymetry data to obtain a regression formula for depth estimation. However, it is not efficient in practical use to collect training bathymetry data for each satellite image. In this study, we developed on a generalized depth estimation model. A depth estimation model which is widely applicable was created using satellite images and training bathymetry data in variable coastal waters. The accuracy of the estimated depth derived by the model was evaluated using the cross-variation method.

In order to accurately and inexpensively fix the location of ground control points for drone remote sensing, we tested the precision of post-processing static (PPS) calculations using a compact GNSS receiver (Reach RTK) and an open-source GNSS post-processing program package (RTKLIB). GNSS fix tests were carried at 3rd order survey points and other ground markers in Joso City and Tsukuba City in Ibaraki Prefecture, Japan. Tests using GNSS antennas of different reception sensitivity found that the most inexpensive antenna provided with Reach RTK was sufficient for precise positional measurements. This paper compares two methods for obtaining the final PPS calculation. In method 1, the last fix calculated was used as the final fix position. In method 2, the final fix position was calculated as the median value of all intermediate fix estimates. The tests found that method 2 was able to reduce GPS fix variability during observation because it could avoid the effects of temporary fix errors or erroneous calculations. In 60-minute observation sessions, the RMSE around the true position using method 2 was 2.7cm in the horizontal direction, and 1.2cm in the vertical direction. The GNSS devices used in this test achieved errors at levels of a few centimeters, more than sufficient for fixing the position of ground markers used for georeferencing drone remote sensing imagery. The method used here showed superior start-up costs (low equipment purchase costs) and running costs (no access costs for GNSS post-processing data) compared to existing GNSS devices or GNSS-equipped ground markers, and can serve as relatively inexpensive and powerful tools for supporting research using drone remote sensing.