Towards accurate L4 ocean colour products: Interpolating remote sensing reflectance via DINEOF

Abstract

Listen

Translate article

Ocean colour (OC) remote sensing benefits society by providing continuous biological and ecological parameters relevant to sustainable marine resource exploitation. It enhances our understanding of climate change and allows us to monitor oceanographic phenomena over various scales of variability. However, significant data gaps occur daily due to cloud cover, atmospheric correction failures, sun-glint contamination, and satellite coverage limitations. Level 4 (L4) gap-free images are generally created by averaging over specific periods (e.g., weekly, monthly, seasonal) or re-gridding data with coarser resolution to overcome these limitations. These approaches, however, often fail to capture anomalous events or fine-scale resolution processes, calling for more advanced methods. The Data Interpolating Empirical Orthogonal Function (DINEOF) method has proved effective in reconstructing missing OC data and capturing smaller-scale features in noisy fields. To the best of authors knowledge, DINEOF is here used for the first time to interpolate multispectral Remote Sensing Reflectance (Rrs) to produce a consistent and gap-free L4 Rrs dataset, minimizing errors in inferred ocean products, such as Chlorophyll-a (Chl), the most widely used proxy for phytoplankton biomass. Specifically, using a multivariate approach, we assessed the DINEOF technique’s capability to reconstruct Rrs, focusing on six bands (412, 443, 490, 510, 555, and 670 nm) and validating the results using extensive in situ datasets. Our outcomes show that this “upstream interpolation” method can generate a consistent Rrs dataset, thereby improving the accuracy of L4 Chl predictions when used as input in algorithms for remote Chl estimation. We anticipate further improvements in L4 Rrs accuracy using richer spectral information from upcoming hyperspectral satellite missions. This study highlights the effectiveness of using Rrs as a standalone dataset for DINEOF interpolation. Operationally, it can derivate various gap-free and consistent biogeochemical parameters with reduced uncertainty, thus providing a more reliable and versatile method.