Developing MODIS-based retrieval models of suspended particulate matter concentration in Dongting Lake, China

https://doi.org/10.1016/j.jag.2014.03.025Get rights and content

Highlights

  • Dongting Lake is an internationally important wetland with a rich biodiversity.

  • An exponential CSPM retrieval model of MODIS red minus mid-infrared band was developed for Dongting Lake.

  • The developed model is helpful for understanding, managing and protecting Dongting ecosystem.

Abstract

To case-II waters, suspended particulate matter (SPM) is one of the dominant water constituents, SPM concentration (CSPM) is a key parameter describing water quality, and developing remote sensing-based CSPM retrieval models is foundation for obtaining its spatiotemporal distributions. This study aimed to develop moderate resolution imaging spectroradiometer (MODIS)-based CSPM empirical retrieval models in Dongting Lake, China. The 95 CSPM measurements on 31 August 2012 and 14 June 2013 and their corresponding MODIS Terra images were used to calibrate models, and the model calibration results showed that the 250 m MODIS red band obtained better fitting accuracies than the near infrared band; the quadratic and exponential models of single red band explained 75% (estimated standard errors (SE) = 6.19 mg/l) and 71% (SE = 6.54 mg/l) of the variation of CSPM; and the quadratic and exponential models of red minus shortwave infrared (SWIR) band at 1240 and 1640 nm explained 72–73% (SE = 6.43–6.48 mg/l) and 68–69% (SE = 6.83–6.96 mg/l) of the variations of CSPM, respectively. The quadratic and exponential models of red band and red minus SWIR band were applied to the MODIS Terra image on 16 September 2013 to estimate CSPM values. By comparing the estimated CSPM values on 16 September 2013 and the measured ones on 17 September 2013 at 40 sampling points for model validations, the results indicated that there exited significantly strong correlations between the measured and estimated CSPM values at a significance level of 0.05 for all models, and the exponential model of red minus SWIR band at 1240 nm achieved the best estimation result within all models. Such result provided foundation for obtaining the spatiotemporal distribution information of CSPM from MODIS images in Dongting Lake, which will be helpful for understanding, managing and protecting this ecosystem.

Introduction

Suspended particulate matter (SPM, including organic and inorganic matter) is one of the main constituents of case-II waters, and it greatly affects water quality and aquatic ecosystems, such as transporting nutrients and contaminants, declining water clarity and reducing light transmission through water column (Kirk, 1994, Davies-Colley and Smith, 2001, Cigizoglu and Kisi, 2006, Giardino et al., 2010, He et al., 2013, Long and Pavelsky, 2013, Xing et al., 2013). SPM concentration (CSPM) is an important parameter describing water quality (Zhang et al., 2003, Pozdnyakov et al., 2005, Uddin et al., 2012), and obtaining its spatiotemporal distribution information is thus necessary for understanding, managing and protecting aquatic ecosystems.

Remote sensing techniques have been widely applied to obtain the spatiotemporal information of CSPM since the first Landsat satellite's launch in 1972. Moderate resolution imaging spectroradiometer (MODIS) is a key instrument aboard the Terra and Aqua satellites of the National Aeronautics and Space Administration (NASA). With their advantages of medium spatial resolution, daily coverage, high sensitivity and cost-free distribution (Li and Li, 2004, Miller and McKee, 2004), MODIS images have been frequently employed to retrieve CSPM values during the past decade (Table 1). Furthermore, the two MODIS sensors have collected more than 10-year massive data of the Earth's surface, and such an image archive provides great opportunity to monitor and analyse the long-term spatiotemporal dynamics of the Earth's surface parameters like CSPM.

There are three kinds of retrieval models generally applied to estimate the water quality parameters (including CSPM) of case-II waters: empirical, semi-empirical and semi-analytical models. Empirical model is based on the bi-variate or multivariate regressions between remote sensing data and measured water quality parameters; semi-empirical model is generally be used when the spectral characteristics of the parameters of interest are known, and such spectral characteristics are included in the statistical analysis by focusing on well-chosen spectral areas and appropriate wavebands; and analytical model determines the water quality parameter values from remote sensing data based on the relations of water quality parameter values, inherent optical properties, apparent optical properties and the top-of-atmosphere radiance captured by remote sensing sensor (Giardino et al., 2007).

The semi-analytical models, due to their relatively good theory, might be more unified for different water bodies compared with the empirical and semi-empirical ones (Sipelgas et al., 2009, Binding et al., 2010, Morozov et al., 2010); however, their applications are limited because of the difficulties or inaccuracies in obtaining model-driving parameters, such as the absorption and backscattering coefficients of main water constituents (Ma et al., 2010). Thus, empirical and semi-empirical models are now frequently employed to estimate CSPM values (Table 1) since they are simple and easy to be developed. Most MODIS-based empirical and semi-empirical CSPM models apply red and/or near infrared band, however, their types (e.g. linear, cubic and exponential) and fitting accuracies (determination coefficient (R2) = 0.58–0.92) varied at different water bodies or different seasons (Table 1). These studies indicate that an empirical or semi-empirical model might generally need to be calibrated when it is applied to different water bodies or seasons, possibly due to their different water constituents and different optical properties. Therefore, it is necessary to develop suitable MODIS-based CSPM retrieval models for a specific water body.

Dongting Lake is the second largest freshwater lake in China. To date, few remote sensing techniques have been employed to study the water quality of this lake from our knowledge and literature review, such as the Hyperion image was applied to analyse the spatial variation of water quality parameters in East Dongting Lake (Mo et al., 2013). This study aimed to develop MODIS-based CSPM empirical retrieval models of Dongting Lake, which is theoretically and practically meaningful for exploring the potential of MODIS images in retrieving the CSPM of Dongting, for obtaining its spatiotemporal information and further for understanding, managing and protecting this ecosystem.

Section snippets

Study area

Dongting Lake (110°40′–113°10′E, 28°30′–30°20′N) is located at the south of the middle Yangtze River (Fig. 1), and it is a large and shallow freshwater lake in China (Ding and Li, 2011). The lake area fluctuates from <500 km2 in the dry seasons (e.g. December–February) to around 2500 km2 in the flood seasons (e.g. July–August) (Huang et al., 2012). Dongting Lake is divided into three parts (East, South and West), in which the East Dongting is the biggest one. The waters of Dongting Lake are

Results

The statistics of CSPM measurements on 31 August 2012, 14 June 2013 and 17 September 2013 are shown in Table 2. The CSPM values between the two years have statistically significant difference at a significance level of 0.05 (z-test for mean difference: mean CSPM on 31 August 2012 = 12.58 ± 13.75 mg/l, mean CSPM on 14 June 2013 = 10.83 ± 10.42 mg/l, z = 2.45, p = 0.014); their difference, however, was not very large (mean difference < 5 mg/l, the highest value difference < 5 mg/l). Compared with the CSPM values

Discussion

Scale is an important factor affecting the developments of water quality retrieval models. In developing models, one image pixel covers a certain area (about 250 × 250 m2 for MODIS red and near infrared bands), while it corresponding water quality parameter value is generally derived from the water sample collected at one certain location. There exists a scale gap between image pixel and water quality parameter measurement, and thus it is generally assumed that the water quality parameter values

Conclusion

An exponential model of the MODIS red minus SWIR band at 1240 nm was developed (CSPM = 3.04 exp(20.23(Rrs(645)  Rrs(1240))), R2 = 0.68, SE = 6.96 mg/l, F = 231.10), and it obtained acceptable result for estimating CSPM values in Dongting Lake. The model lays foundation for obtaining the spatiotemporal distribution information of CSPM from historic MODIS image achieve, and will be helpful for understanding, managing and protecting Dongting ecosystem.

Acknowledgements

This study was supported by the National Natural Science Foundation of China (Grant Nos. 41171290 and 40971191).

References (50)

  • M. Ondrusek et al.

    The development of a new optical total suspended matter algorithm for the Chesapeake Bay

    Remote Sens. Environ.

    (2012)
  • D. Pozdnyakov et al.

    Operational algorithm for the retrieval of water quality in the Great Lakes

    Remote Sens. Environ.

    (2005)
  • L. Sipelgas et al.

    Operational monitoring of suspended matter distribution using MODIS images and numerical modelling

    Adv. Space Res.

    (2006)
  • J.J. Wang et al.

    Estimation of suspended sediment concentrations using Terra MODIS: An example from the Lower Yangtze River, China

    Sci. Total Environ.

    (2010)
  • G. Wu et al.

    Comparison of MODIS-based models for retrieving suspended particulate matter concentrations in Poyang Lake China

    Int. J. Appl. Earth Observ. Geoinform.

    (2013)
  • C.E. Binding et al.

    Suspended particulate matter in Lake Erie derived from MODIS aquatic colour imagery

    Int. J. Remote Sens.

    (2010)
  • H.K. Cigizoglu et al.

    Methods to improve the neural network performance in suspended sediment estimation

    J. Hydrol.

    (2006)
  • L. Cui et al.

    Using remotely sensed suspended sediment concentration variation to improve management of Poyang Lake, China

    Lake Reserv. Manage.

    (2013)
  • R.J. Davies-Colley et al.

    Turbidity, suspended sediment, and water clarity: a review

    J. Am. Water Resour. Assoc.

    (2001)
  • D. Doxaran et al.

    Apparent and inherent optical properties of turbid estuarine waters: measurements, empirical quantification relationships, and modeling

    Appl. Opt.

    (2006)
  • L. Feng et al.

    Human induced turbidity changes in Poyang Lake between 2000 and 2010: observations from MODIS

    J. Geophys. Res.

    (2012)
  • C. Giardino et al.

    Remote sensing of suspended particulate matter in Himalayan Lakes a case study of Alpine Lakes in the Mount Everest Region

    Mount. Res. Dev.

    (2010)
  • S.F. Huang et al.

    Water surface variations monitoring and flood hazard analysis in Dongting Lake area using long-term Terra/MODIS data time series

    Nat. Hazards

    (2012)
  • X. Huang

    Eco-Investigation, Observation and Analysis of Lakes

    (1999)
  • H. Jiang et al.

    Monitoring of TSS concentration in Poyang Lake based on MODIS data

    Yangtze River

    (2011)
  • Cited by (31)

    • Variations of suspended particulate concentration and composition in Chinese lakes observed from Sentinel-3A OLCI images

      2020, Science of the Total Environment
      Citation Excerpt :

      For instance, dredging activities were found to increase the SPM in winter but decrease it in summer in Lake Hongze (Cao et al., 2017). Sand dredging activities are prevalent in three (Lake Poyang, Lake Dongting, and Lake Hongze) of the five largest freshwater lakes in China, and occur in Lake Luoma and the Yangtze River (Duan et al., 2019; Lai et al., 2014; Wu et al., 2014). In general, these driving forces jointly contribute to the spatial and temporal variations in the concentrations and composition of suspended particulate matter.

    • Fluvial and oceanographic influences on suspended sediment dispersal in the Magdalena River Estuary

      2020, Journal of Marine Systems
      Citation Excerpt :

      We selected the best model based on multiple statistics (mean quadratic error, coefficient of determination, p-value and Akaike Information Criterion AIC) (Wu et al., 2014). The root mean square error (RMSE) measures the error between the observed and estimated value (Wu et al., 2014). The coefficient of determination and p-value indicate the significance of the model, and the nature of the relationship between variables (Wu et al., 2014).

    • Evaluation of the suitability of Landsat, MERIS, and MODIS for identifying spatial distribution patterns of total suspended matter from a self-organizing map (SOM) perspective

      2019, Catena
      Citation Excerpt :

      Although Landsat TM/OLI sensors are not designed for high temporal resolution, their high spatial and radiometric resolution make them quite suitable for water system monitoring (Gerace et al., 2013). Even though these three sensors possess technical limitations, several investigators have applied them to the problem of modeling water quality parameters and TSM in coastal and estuary waters (Chen et al., 2015a; González Vilas et al., 2011; Kyryliuk and Kratzer, 2016; Wu et al., 2013; Wu et al., 2014). Numerous studies have explored the potential of Landsat MSS/TM/OLI 30-m, MODIS 250-m, and MERIS full resolution 290-m data for monitoring water qualities (Ambarwulan et al., 2012; Bernardo et al., 2016; Cai et al., 2015; Kyryliuk and Kratzer, 2016; Tarrant et al., 2010; Wu et al., 2013; Wu et al., 2014; Zhang et al., 2014; Zhou et al., 2006), especially with a focus on the measurement of TSM in lakes and coastal areas, which obtained a good performance for estimating TSM concentration (Ambarwulan et al., 2012; Chen et al., 2015a, 2015b; Emiyati et al., 2017; Luan, 2016; Miller and McKee, 2004; Wu et al., 2015; Xi and Zhang, 2010; Zhang et al., 2014; Zhou et al., 2006).

    View all citing articles on Scopus
    View full text