Can MODIS Land Reflectance Products be Used for Estuarine and Inland Waters?

Although designed for land surfaces, MODIS Aqua surface reflectance data products (MYD09, termed as R_Land in this work) have also been used for water applications. Yet to date their uncertainties and general suitability in such applications have rarely been documented. In this study, R_Land products of two regions (Chesapeake Bay and Taihu Lake) between July 2002 and December 2015 are evaluated against in situ measurements and against reflectance products derived by the MODIS Ocean Team using atmospheric correction schemes specifically designed for water applications, namely the default atmospheric correction method based on the near‐infrared (NIR) bands (denoted as R_NIR, data products available from NASA) and alternative atmospheric correction method based on the shortwave‐infrared (SWIR) bands (denoted as R_SWIR, data products not available from NASA but require customized processing by the user), respectively. Results suggest high accuracy in R_Land(645) and R_Land(645/555) for both Chesapeake Bay and Taihu Lake in terms of daily spatial distributions, seasonality, and long‐term trends. A sensitivity test also shows improved data quality in R_Land(645/555) when data are binned by 7 × 7 pixels in space and 32 days in time. Improved data quality can also be obtained for R_Land(645) when data are only binned in time to minimize the patchiness noise in R_Land daily images. Given the fact that most users do not have the capacity to process low‐level data to obtain R_SWIR and the standard NASA R_NIR products often lack coverage over inland waters because they are optimized for global oceans instead of inland waters, this study provides a general guide on the applicability of the widely available R_Land products in inland and estuarine water applications in the absence of customized R_NIR or R_SWIR data products for local regions. Key Points MYD09 (R_Land) products were evaluated using in situ data and reflectance derived using ocean color based NIR and SWIR atmospheric corrections R_Land(645) and R_Land(645/555) demonstrated high accuracy in terms of daily spatial distributions, seasonality, and long‐term trends Improved data quality in R_Land(645/555) was found when data are binned by 7 × 7 pixels in space and 32 days in time