**2.2 Satellite imagery**

The in-situ measurements provide detailed observations on CDOM optical properties. However, the spatial and temporal resolutions of in-situ observations are always limited by high costs in the survey implementation. The satellite imagery can offer better temporal and spatial coverage with high efficiency and low cost. Therefore, space-born and airborne sensors have been used to monitor spectral properties of natural waters for over four decades, and ocean color remote sensing has become an important technology to study water environment.

Nevertheless, the applicability of satellite observation depends on the reliability of CDOM ocean color algorithms. The spectral properties of water bodies may change substantially with geographical locations and time. The color (spectrum) of water can be used to estimate concentrations of optically active constituents in water, such as the phytoplankton, suspended solids, and CDOM based on effective ocean color algorithms.

A regional algorithm for CDOM absorption estimation from satellite ocean color imagery is developed for the spring PRE (Eqs. (3)–(5)), which derives CDOM UV absorption and spectral slope from the visible remote sensing reflectance [Rrs(λ)], which is established based on the in-situ measured above water remote sensing reflectance (**Figure 3**). The algorithm is with an overall mean absolute percentage difference (MAPD) of ~30 and ~5% for the estimation of CDOM absorption and the spectral slope over 250–450 nm, respectively.

$$\mathbf{a}\_{\mathrm{g}}(290) = \mathbf{108.2R\_{\mathrm{rs}}(596) - 0.5324} \tag{3}$$

$$\mathbf{R\_{rs\\_Gradient}} = \left[\mathbf{R\_{rs}}(\text{max}) - \mathbf{R\_{rs}}(\text{min})\right] / \left[\lambda(\text{max}) - \lambda(\text{min})\right] \tag{4}$$

*Variations of the Absorption of Chromophoric Dissolved Organic Matter in the Pearl River… DOI: http://dx.doi.org/10.5772/intechopen.90765*

$$\text{S}\_{\text{g}}(250-400) = 0.01187 \text{R}\_{\text{rs\\_Gradient}}^{\text{-0.1/41}} \tag{5}$$

Using all available satellite image data obtained by four ocean color sensors with different spatial and spectral resolutions: the Visible Infrared Imaging Radiometer Suite (VIIRS) on board Suomi National Polar-orbiting Partnership (Suomi NPP), the Ocean and Land Colour Instrument (OLCI) on board Sentinel-3A (S3A), the Hyperspectral Imager for the Coastal Ocean (HICO) integrated in the International Space Station (ISS) Window Observational Research Facility (WORF), and the Operational Land Imager (OLI) on board Landsat 8 (LS8) (**Table 1**), a time series of CDOM absorption and spectral slope in the PRE and the Hong Kong waters in spring from 2012 to 2018 is produced. Relevant factors related to the temporal variation of CDOM absorption and spectral slope are analyzed.

To match the season of the in-situ observations in this study, a set of satellite images acquired in spring (April and May) from 2012 to 2018 is selected for the application of the developed algorithm. For the OLI data with a 16-day revisit cycle, the time constraint is relaxed to March and June. The study area covers 22° and 23°N, 113.5° and 114.5°E. The images with cloud coverage below 20% are downloaded (see **Tables 2** and **3** for detailed information of the available images).

#### **Figure 3.**

*Normalized in-situ water surface reflectance in 380–900 nm measured by an Ocean Optics 4000 spectrometer using above-water method.*


#### **Table 1.**

*The properties of ocean color remote sensors utilized in this paper.*

#### *Estuaries and Coastal Zones - Dynamics and Response to Environmental Changes*


#### **Table 2.**

*Satellite images used for ocean color application.*


#### **Table 3.**

*Acquisition time of the available images.*


#### **Table 4.**

*The criterion used to match Rrs(596) with the bands of four sensors.*

Eq. (3) is applied to derive ag(290) from the selected satellite images. The Rrs(596) is matched to the bands of the four sensors by the criteria listed in **Table 4**. When retrieving Sg(250–400), considering the available spectral range of the satellite imagery and the performance of atmospheric correction, the lower ends for Rrs gradient calculation (the λmin in Eq. (4)) are set as 445, 415.5, 400, and 443 nm for the VIIRS, HICO, OLCI and OLI data, respectively. The range for the Rrs maximum is limited below 700 nm for all sensors. Sg(250–400) is afterward derived from Rrs gradient by Eq. (5).
