Effects of broad bandwidth on the remote sensing of inland waters: Implications for high spatial resolution satellite data applications

Abstract

High spatial resolution satellite sensors provide opportunities to observe spatial variations of biogeochemical properties of small- and medium-sized inland water bodies. However, high spatial resolution sensors are usually equipped with wider spectral bandwidth (>50 nm) that diminishes the features of the spectrum. Therefore, the effects of the border bandwidth issue need to be evaluated prior to application in aquatic environments. Based on the in situ optical data [remote sensing reflectance (Rrs) and absorption coefficients] and the radiative simulations of hyperspectral remote sensing reflectance and using band specifics of common sensors (e.g., OLCI, VIIRS, MSI, OLI, ETM+ and WFV) as examples, the effects of bandwidth on optical properties of inland waters were analyzed. The results showed the followings. (1) The difference between values at center-wavelength and band-averaged values increased with increasing bandwidth for Rrs and the absorption coefficients. The difference was wavelength-dependent. The difference of Rrs at the visible band was within 0.25% but greater than 0.5% for the spectral bands near 710 nm and 665 nm. (2) The accuracy of the total absorption coefficient derived from QAA-750E, spectral match technique (SMT) and deep neural network (DNN) decreased with increasing bandwidth. The QAA-750E was more sensitive to bandwidth than SMT and DNN. Otherwise, the empirical algorithms for estimating chlorophyll-a (Chla) concentrations were significantly affected by bandwidth. The performance of algorithms for estimating cyanobacterial phycocyanin (PC) and suspended particulate matter (SPM) concentrations changed slightly with a wider bandwidth. Finally, the maximum bandwidth requirement for optical remote sensing in inland waters was proposed. For bandwidth options, it should be within 20 nm for 700–710 nm, ∼30 nm maximum for ∼560 nm and ∼665 nm, 60 nm for ∼620 nm, and ∼80 nm for ∼443 nm and ∼490 nm, respectively. The difference between the Rrs of narrow bands (10–20 nm) and the Rrs of the bands with the recommended bandwidth was within 0.25%. The corresponding bandwidth from MSI and OLI sensors meet this criterion for Chla and SPM. However, the lack of spectral coverage near 700–710 nm may present a challenge to retrieve Chla concentration from OLI images. This study provided helpful theoretical and practical references for the retrieval of inland water parameters by high spatial resolution satellite sensors and its prospective development.