Algae are a diverse group of simple organisms consisting of chlorophyll-a (Chl-a), and an important part of the food web that can produce oxygen, remove nutrients from water, and stabilize sediments (Shalaby, 2011). Excessive algal biomass, which is measured by Chl-a concentration can be a sign of water quality degradation as a result of raised nutrients’ levels in the water (USEPA, 2022), proper temperature, and light conditions (Li et al., 2017). They can limit sunlight penetration through water and create the hypoxic condition that causes ecosystem degradation and economic damage to various sectors such as water supply and treatment, fisheries, and tourism (Hafeez et al., 2019, Stauffer et al., 2019). While algal bloom last from a few days to months and cover an area from a few square meters to several square kilometers, technical and finical constraints can limit the applicability of field and laboratory measurements for monitoring spatial and temporal changes of algae, especially in large water bodies (Hafeez et al., 2019). In contrast, satellite remote sensing can simultaneously cover a vast area and provide real-time data at a reasonable cost (Hossain et al., 2021).
Despite numerous studies on the estimation of Chl-a concentration in oceans and estuaries by remote sensing, inland water bodies have received less attention (Topp et al., 2020). Most researchers have found a high correlation between the field data of Chl-a and the ones obtained by remote sensing in lakes, dam reservoirs, and coastal areas (Hossain et al., 2021). However, applying remote sensing still faces several challenges such as the effect of weather and air conditions (e.g cloud cover and air pollution) on the quality of received images, and limited depth of measurement (Hafeez et al., 2019). Moreover, spatial, radiometric, spectral, and temporal resolutions of the sensors can affect performance of satellite measurements (NASA, 2022). These challenges can limit remote sensing applications, especially in small water bodies (Shi et al., 2022). Also, in shallow lakes, the effect of bottom reflection can make errors in the prediction of Chl-a concentration (Allan et al., 2007).
Satellite Ocean color sensors, such as MODIS and MERIS, meet the requirements for monitoring water bodies with 1–3 days temporal resolution and 12-bits radiometric resolution. While the spatial resolution of these sensors (300–1000 m) is appreciated for large water bodies, small water bodies need higher spatial and spectral resolutions. Landsat 1–7 series satellites provide better spatial resolutions (30–79 m) but, their radiometric resolutions (6–8 bits) are limited. Alternatively, Landsat 8 OLI/TIRS with 12-bits radiometric and 30 meters spatial resolutions, has temporal resolutions of 1–16 days which may restricts its application in water quality monitoring. In contrast, the Sentinel-2 (MSI) multispectral imager, launched in 2015, has spatial resolutions of 10, 20, and 60 meters, which make it a suitable option for small water bodies. It can provide data in 13 spectral bands with 12-bit radiometric resolution and five-day temporal resolution (Toming et al., 2016).
Chl-a has reflectance-absorption patterns at specific wavelengths that correspond to the reflectance-absorption patterns of algal pigments in water (Ha et al., 2017). For example, Chl-a shows high absorption around the blue and red regions and high reflectance in the green and near-infrared spectrum (NIR) regions of the electromagnetic spectra. These features have been used in several studies to develop Chl-a quantification algorithms (Buma & Lee, 2020). Among the various algorithms, the ones based on the relationship between Chl-a concentration and reflectance at the red and NIR spectra have shown the strongest correlations even in waters with high turbidity and colored dissolved organic matter (Gitelson, 1992). However, the best algorithm to estimate Chl-a concentration depends on the optical properties of water body, its type (ocean, coastal, inland water), etc. (Ha et al., 2017). Also, spectral responses and effective absorption and reflectance of wavelengths depending on the Chl-a concentration (Hafeez et al., 2019). Therefore, band ratio algorithms to estimate the Chl-a concentration should be evaluated separately for each water body.
For freshwater bodies, most Chl-a estimation approaches are based on empirical or semi-empirical relationships between the narrowband reflections (Ritchie et al., 2003), the ratio of sensor bands, and Chl-a (Topp et al., 2020). For example, in Lake Chad, among four common algorithms of 2BDA, 3BDA, NDCI, and FLH using World view-3, Landsat-8, and Sentinel-2 images, 3BDA and NDCI algorithms demonstrated the best performance for estimation of Chl-a concentration (Buma & Lee, 2020). Also, Mishra & Mishra (2012) showed the three band ratios and the NDCI index (the reflectance data of MERIS sensor) can predict the Chl-a concentration in high turbidity waters with appropriate performance. Moreover, Ha et al. (2017) used the green-to-red two-band ratios of Sentinel-2A images in an exponential function to estimate Chl-a concentration in a tropical freshwater lake and they found it the best model with an error of 5%. On the other hand, Shi et al. (2022) used Sentinel-2 and Gaofen-6 satellite images and applied three semi-empirical models of 3BDA, NDCI, and YA10 to estimate Chl-a concentration in small water bodies and found the 3BDA model as the best in both satellites. However, in recent research, machine learning algorithms such as artificial neural networks, support vector regression, and random forest that take the advantages of all bands’ reflections have been considered to find the best algorithm to quantify the water quality parameters (Hafeez et al., 2019; Yang et al., 2022).
Water quality monitoring of recreational waters such as Chitgar Lake, which are often shallow freshwater waterbodies, is particularly important because of their possible hygiene, health, and environmental impacts. High phytoplankton population and algal blooms can cause low oxygen and dangerous conditions for aquatic life, inappropriate views, and adverse effects on recreation activities such as the release of unpleasant odors and algal toxins (Bayat et al., 2019). Also, the low depths of shallow lakes make them more sensitive to climate changes and affect water mixing, Chl-a concentration, and water turbidity (Bohn et al., 2018). Thus, effective management of these water bodies often requires high-frequency monitoring of water quality over a large area, which usually requires a lot of human resources, financial resources, and equipment. Considering the importance of water quality management and the requirement for monitoring with high temporal and spatial resolution, this study was conducted to investigate the efficiency of remote sensing technology in estimating Chl-a concentration in the Chitgar Persian Gulf Martyrs Lake, Tehran, Iran.