Elsevier

Remote Sensing of Environment

Volume 112, Issue 9, 15 September 2008, Pages 3582-3593
Remote Sensing of Environment

A simple semi-analytical model for remote estimation of chlorophyll-a in turbid waters: Validation

https://doi.org/10.1016/j.rse.2008.04.015Get rights and content

Abstract

Accurate assessment of phytoplankton chlorophyll-a (chla) concentrations in turbid waters by means of remote sensing is challenging due to the optical complexity of case 2 waters. We have applied a recently developed model of the form [Rrs 1(λ1) − Rrs 1(λ2)] × Rrs(λ3) where Rrs(λi) is the remote-sensing reflectance at the wavelength λi, for the estimation of chla concentrations in turbid waters. The objectives of this paper are (a) to validate the three-band model as well as its special case, the two-band model Rrs 1(λ1) × Rrs(λ3), using datasets collected over a considerable range of optical properties, trophic status, and geographical locations in turbid lakes, reservoirs, estuaries, and coastal waters, and (b) to evaluate the extent to which the three-band model could be applied to the Medium Resolution Imaging Spectrometer (MERIS) and two-band model could be applied to the Moderate Resolution Imaging Spectroradiometer (MODIS) to estimate chla in turbid waters.

The three-band model was calibrated and validated using three MERIS spectral bands (660–670 nm, 703.75–713.75 nm, and 750−757.5 nm), and the 2-band model was tested using two MODIS spectral bands (λ1 = 662–672, λ3 = 743–753 nm). We assessed the accuracy of chla prediction in four independent datasets without re-parameterization (adjustment of the coefficients) after initial calibration elsewhere. Although the validation data set contained widely variable chla (1.2 to 236 mg m 3), Secchi disk depth (0.18 to 4.1 m), and turbidity (1.3 to 78 NTU), chla predicted by the three-band algorithm was strongly correlated with observed chla (r2 > 0.96), with a precision of 32% and average bias across data sets of − 4.9% to 11%. Chla predicted by the two-band algorithm was also closely correlated with observed chla (r2 > 0.92); however, the precision declined to 57%, and average bias across the data sets was 18% to 50.3%. These findings imply that, provided that an atmospheric correction scheme for the red and NIR bands is available, the extensive database of MERIS and MODIS imagery could be used for quantitative monitoring of chla in turbid waters.

Introduction

Remote sensing of water-constituent concentrations is based on the relationship between the remote-sensing reflectance, Rrs(λ), and the inherent optical properties, namely, the total absorption (a) and the backscattering (bb) coefficients (e.g., Gordon et al., 1988):Rrs(λ)γbb(λ)a(λ)+bb(λ)where λ is the wavelength (nm) and γ is dependent on the geometry of the light field emerging from the water body. a(λ) is the sum of the absorption coefficients of phytoplankton pigments (apigm), colored dissolved organic matter (aCDOM), tripton (atripton, non-algal particles), and pure water (awater):a=apigm+aCDOM+atripton+awater

To retrieve chla concentrations from spectral reflectance, one needs to isolate the chla absorption coefficient, achla, which is a part of apigm. In open ocean waters where water and pigments are the primary components of Eq. (2), chla is typically derived using the blue and green spectral regions (e.g., Gordon and Morel, 1983, Morel and Prieur, 1977). In turbid waters, however, these spectral regions cannot be used for retrieving chla because of the overlapping, uncorrelated, and much larger absorptions of dissolved organic matter and tripton (e.g., Dall'Olmo et al., 2003, GKSS, 1986, Gitelson, 1992, Gons, 1999, Gitelson et al., 1987, Gitelson et al., 2007).

A variety of algorithms have been developed for retrieving chla in turbid waters. All are based on the properties of the reflectance peak near 700 nm (e.g., Gitelson et al., 1985, Gitelson, 1992, Gower et al., 1999, Stumpf and Tyler, 1988, Vasilkov and Kopelevich, 1982). These include the ratio of that reflectance peak to the reflectance at 670 nm (the red chla absorption band), or the ratio R705/R670 (Dekker, 1993, Gitelson et al., 1985, Gitelson and Kondratyev, 1991, Gitelson, 1992, Han and Rundquist, 1997, Mittenzwey et al., 1992), and the position of this peak (Gitelson, 1992). Using vector analysis, Stumpf and Tyler (1988) showed that the ratio of the near infra-red (NIR) and the red bands of AVHRR and CZCS can identify phytoplankton blooms and has the potential to provide estimates of chla above 10 mg m 3 in turbid estuaries. Gons (1999) and Gons et al. (2000) used the ratio of reflectances at 704 nm and 672 nm and the absorption and backscattering coefficients at these wavelengths to assess widely ranging chla concentrations. In order to obtain bb, the backscattering coefficient in Eq. (1), reflectance at 775 nm was also used. Later, this algorithm was adapted for use with MERIS satellite imagery (Gons et al., 2002). This MERIS algorithm was calibrated using spectral data taken from shipboard measurements and validated for various inland and coastal waters covering a chla concentration range of 3–185 mg m 3. Alternatively, good correlations have been found between chla and the band ratio R725/R675 (Dall'Olmo and Gitelson, 2005, Hoge et al., 1987, Pierson and Strömbäck, 2000, Yacobi et al., 1995). Ruddick et al. (2001) analyzed how errors in reflectance measurements affect chla retrievals for a NIR to red reflectance ratio algorithm with a general choice of wavelengths. They have found that the effect on chla retrieval depends strongly on the choice of the NIR wavelength if the error is spectrally neutral, suggesting a new type of algorithm, where the NIR wavelength used for retrieval is chosen dynamically for each spectrum to be processed. A prerequisite for application of the algorithm is the availability of reflectance data with sufficient spectral resolution in the 700–740-nm range and good wavelength accuracy to enable the critical NIR wavelength to be accurately located.

Recently, Dall'Olmo et al. (2003) provided evidence that a three-band reflectance model, originally developed for estimating pigment contents in terrestrial vegetation (Gitelson et al., 2003, Gitelson et al., 2005), could also be used to assess chla in turbid waters. The model relates pigment concentration to reflectance R(λi) in three spectral bands λi (Gitelson et al., 2003):Pigmentconcentration[R1(λ1)R1(λ2)]×R(λ3)

Reciprocal reflectance in the first spectral band λ1, R 1(λ1) (achla  +aCDOM + atripton + awater + bb) / bb, should be maximally sensitive to achla; this means that λ1 has to be restricted within the range of 660 to 690 nm (Dall'Olmo & Gitelson, 2005). However, in addition to absorption by chla, R 1(λ1) is also affected by absorption by tripton, CDOM, and water as well as backscattering by all particulate matter bb. The effect of (atripton +aCDOM) and bb (in the denominator of Eq. (1)) can be minimized using a second spectral band, where R 1(λ2) is minimally sensitive to absorption by chla (i.e., achla(λ2) << achla(λ1)) and (atripton + aCDOM) in band λ2 is quite close to that in band λ1. Thus, this band position should meet two conflicting requirements: (a) to be quite close to λ1 in order to meet the requirement atripton(λ1) ~ atripton(λ2) and aCDOM(λ1) ~ aCDOM(λ2) and (b) it should be quite far from λ1 to provide achla(λ2) << achla(λ1). It was shown that λ2 should be in the range from 710 to 730 nm (Dall'Olmo and Gitelson, 2006, Gitelson et al., 2007). The underlying assumption is that both (atripton + aCDOM) and bb are spectrally flat in the red-NIR region. The difference, R 1(λ1) − R 1(λ2) [achla (λ1) + awater(λ1) − awater(λ2)] / bb, however, is still affected by bb; if backscattering varies between samples, the model output would be different for the same chla. To account for this, a third spectral band λ3 has been used, where reflectance is minimally affected by achla, atripton, and aCDOM, and the total absorption at λ3 is a measure of the absorption by water, i.e. a(λ3) ~ awater. The NIR range, where abb and RNIR ∝ bb, meets these conditions (e.g., Babin and Stramski, 2002, Gons, 1999). Thus, the three-band model [R 1(λ1) − R 1(λ2)] × R(λ3) ∝ achla (λ1); it allows one to isolate the chla absorption coefficient. The two-band model R 1(λ1) × R(λ3) is a special case of the conceptual model (Eq. (3)) when achla(λ1) ≫ bb, and achla(λ1) ≫ atripton(λ1) + aCDOM(λ1) (Dall'Olmo & Gitelson, 2005).

The algorithm for chla estimation was tested using observations from lakes and reservoirs with variable optical properties in Nebraska and Iowa (Dall'Olmo & Gitelson, 2005). It was found that the variability in the quantum yield of chla fluorescence, and the chla specific absorption coefficient, among other factors, considerably affect the accuracy of remote chla estimation (Dall'Olmo & Gitelson, 2006). Instead of re-parameterization of the models (adjustment of the coefficients) for water bodies with specific optical properties, Dall'Olmo and Gitelson (2006) suggested tuning the spectral band positions in order to minimize these effects.

Thus, to use this model for chla estimation over a wide range of optical properties, trophic conditions and geographical locations, the positions of the spectral bands and their widths must be identified. A wide bandwidth offers the advantage of improved signal-to-noise ratios, but the use of wide bandwidths needs to be assessed with respect to the impact on the ultimate accuracy achieved in determining the chla concentration when employing a three-band model. Once the spectral configuration is determined and understood, one should be able to specify the requirements for a simple three-band radiometer having enough sensitivity to measure accurately the low values of water reflectance in the red and near infra-red regions.

Satellite remote sensing of turbid waters has been hindered by the lack of an atmospheric correction that does not assume zero water-leaving radiance in the NIR. An alternative correction scheme based on the short-wave infra-red bands of the Moderate Resolution Imaging Spectroradiometer (MODIS) has been recently shown to retrieve accurate water-leaving radiance over optically complex waters (Wang & Shi, 2005). On the other hand, the Medium Resolution Imaging Spectrometer (MERIS) is provided with a greater spectral resolution than MODIS in the red and NIR. Encouraged by the availability of these satellite products, we studied the performance of a MERIS implementation of the three-band model and a two-band model that could be implemented using MODIS data. Those results are presented here.

The study had three objectives: (a) to locate the positions and widths of the spectral bands of the three-band model and develop an algorithm for accurate chla assessment, (b) to evaluate the accuracy of the three-band algorithm with wide spectral bands and with no re- parameterization (adjustment of coefficients), and (c) to test the three-band algorithm with the spectral bands of MERIS and to test the two-band algorithm with the spectral bands of MODIS.

Section snippets

Datasets used

To assess the accuracy of the model in predicting chla concentrations, five independent datasets containing the spectral optical properties of the water column were used. The first dataset was used for model calibration, while the second through fifth were used for model validation. The calibration dataset containing 145 samples was taken in 2001–2002 in lakes and reservoirs in Nebraska and Iowa. Two complementary types of Eastern Nebraska water bodies were sampled: (1) sandpit lakes with

Constituent concentrations and compositions

The datasets encompassed varying optical conditions and included a wide range of phytoplankton taxonomic groups including Chrysophyta, Chlorophyta, Cyanophyta, Cryptophyta, and Pyrrophyta (details in Dall'Olmo and Gitelson, 2005, Dall'Olmo et al., 2005). In each of the datasets taken in Nebraska and Iowa from 2001 through 2005, the concentrations of chla, TSS, OSS, and ISS, as well as the Secchi disk depth and turbidity, varied over two orders of magnitude (e.g., 1–240 mg chla m 3, 0.2–210 mg

Discussion

The three-band model (Eq. (3)) was calibrated using data taken over lakes and reservoirs in Nebraska and Iowa and the specific form of this model as expressed by the three-band algorithm in Eq. (9) was applied for predicting chla concentrations for four independent datasets taken in inland and estuarine waters. It was found that this algorithm does not require further optimization of spectral band positions and site-specific parameterization to accurately estimate chla in water bodies with

Acknowledgements

This research was supported partially by the US Environmental Protection Agency under grant R-828634501 to J.H., A.G., and D.R., NASA grant NNG06GA92G to T.F. and A.G, NASA grant NNG06GG17G to A.G., and NASA Earth System Science Fellowship Grant NGT5-NNG04GQ82H to G.D. A contribution of the University of Nebraska Agricultural Research Division, Lincoln. This research was also supported in part by funds provided through the Hatch Act. We would like to thank two anonymous reviewers for very

References (43)

  • Dall'OlmoG. et al.

    Effect of bio-optical parameter variability and uncertainties in reflectance measurements on the remote estimation of chlorophyll-a concentration in turbid productive waters: Modeling results

    Applied Optics

    (2006)
  • Dall'OlmoG. et al.

    Towards a unified approach for remote estimation of chlorophyll-a in both terrestrial vegetation and turbid productive waters

    Geophysical Research Letters

    (2003)
  • Dekker, A. 1993. Detection of the optical water quality parameters for eutrophic waters by high resolution remote...
  • GitelsonA. et al.

    Inland waters quality assessment from satellite data in visible range of the spectrum

    Soviet Remote Sensing

    (1985)
  • GitelsonA.A.

    The peak near 700 nm on reflectance spectra of algae and water: Relationships of its magnitude and position with chlorophyll concentration

    International Journal of Remote Sensing

    (1992)
  • GitelsonA.A. et al.

    Optical models of mesotrophic and eutrophic water bodies

    International Journal of Remote Sensing

    (1991)
  • GitelsonA.A. et al.

    Optical properties and non-destructive estimation of anthocyanin content in plant leaves

    Photochemistry and Photobiology

    (2001)
  • GitelsonA.A. et al.

    Remote estimation of canopy chlorophyll content in crops

    Geophysical Research Letters

    (2005)
  • GitelsonA.A. et al.

    Relationship between gross primary production and chlorophyll content in crops: Implications for the synoptic monitoring of vegetation productivity

    Journal of Geophysical Research

    (2006)
  • GitelsonA.A. et al.

    Remote estimation of phytoplankton density in productive waters, Archive Hydrobiological Special Issues

    Advances in Limnology

    (2000)
  • GitelsonA.A. et al.

    Assessing carotenoid content in plant leaves with reflectance spectroscopy

    Photochemistry and Photobiology

    (2002)
  • Cited by (511)

    • An improved algorithm for the column-integrated algal biomass retrieval in Lake Chaohu, a large eutrophic lake

      2024, International Journal of Applied Earth Observation and Geoinformation
    View all citing articles on Scopus
    View full text