Elsevier

Journal of Environmental Management

Volume 151, 15 March 2015, Pages 416-426
Journal of Environmental Management

Integrated satellite data fusion and mining for monitoring lake water quality status of the Albufera de Valencia in Spain

https://doi.org/10.1016/j.jenvman.2014.12.003Get rights and content

Highlights

  • Remote sensing provides intensive monitoring capacity.

  • Image fusion and data mining are key to success of environmental monitoring.

  • Near real-time monitoring can help improve lake management policy.

Abstract

Lake eutrophication is a critical issue in the interplay of water supply, environmental management, and ecosystem conservation. Integrated sensing, monitoring, and modeling for a holistic lake water quality assessment with respect to multiple constituents is in acute need. The aim of this paper is to develop an integrated algorithm for data fusion and mining of satellite remote sensing images to generate daily estimates of some water quality parameters of interest, such as chlorophyll a concentrations and water transparency, to be applied for the assessment of the hypertrophic Albufera de Valencia. The Albufera de Valencia is the largest freshwater lake in Spain, which can often present values of chlorophyll a concentration over 200 mg m−3 and values of transparency (Secchi Disk, SD) as low as 20 cm. Remote sensing data from Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat Thematic Mapper (TM) and Enhance Thematic Mapper (ETM+) images were fused to carry out an integrative near-real time water quality assessment on a daily basis. Landsat images are useful to study the spatial variability of the water quality parameters, due to its spatial resolution of 30 m, in comparison to the low spatial resolution (250/500 m) of MODIS. While Landsat offers a high spatial resolution, the low temporal resolution of 16 days is a significant drawback to achieve a near real-time monitoring system. This gap may be bridged by using MODIS images that have a high temporal resolution of 1 day, in spite of its low spatial resolution. Synthetic Landsat images were fused for dates with no Landsat overpass over the study area. Finally, with a suite of ground truth data, a few genetic programming (GP) models were derived to estimate the water quality using the fused surface reflectance data as inputs. The GP model for chlorophyll a estimation yielded a R2 of 0.94, with a Root Mean Square Error (RMSE) = 8 mg m−3, and the GP model for water transparency estimation using Secchi disk showed a R2 of 0.89, with an RMSE = 4 cm. With this effort, the spatiotemporal variations of water transparency and chlorophyll a concentrations may be assessed simultaneously on a daily basis throughout the lake for environmental management.

Introduction

Eutrophication in lakes and coastal waters has long been a common environmental problem in many countries around the world. Eutrophication mostly results from an anthropogenic supply of nutrients, mainly nitrogen and phosphorus, from urban and agricultural runoff (Bricker et al., 1999). These nutrients enhance algal growth, thereby reducing water transparency and deteriorating water quality (Bricker et al., 1999). Remote sensing reflectance bands can be used in aquatic environments to detect these pollution impacts. Chlorophyll a concentrations (Chl-a hereafter), an indicator of phytoplankton biomass that exhibits a positive correlation with a nutrient increase, are normally determined from bio-optical observations in relatively open waters (O'Reilly et al., 1998).

Remotely sensed Chl-a is particularly challenging in turbid, highly productive waters (Siegel et al., 2000, Gitelson et al., 2007, Gitelson et al., 2008), such as the coastal lake Albufera de Valencia (Spain). In many turbid water bodies, total suspended solids (TSS), either organic (mainly phytoplankton) or inorganic (sediment particles), determine water column optics. Whereas the latter can be depicted by water transparency, the former can be reflected by Chl-a. Monitoring water transparency informs the eutrophic status of the ecosystem because the growth of phytoplankton decreases light penetration. Paired monitoring of water transparency and Chl-a can thereby reveal the general condition of the trophic status in a lake when the lack of transparency is mostly due to phytoplankton. With these indicators, water bodies can be identified through the oligotrophic to the hypereutrophic range with accuracy. The Organization for Economic Co-operation and Development (OECD, 1982) proposed different ranges of the trophic classification determined in terms of total phosphorus (TP), Chl-a and water transparency measured with Secchi disk (SD), which have been widely accepted.

Encouraged by the enforcement of the European Water Framework Directive (WFD) in 2000, which establishes that all EU countries have to monitor the ecological status of their surface water bodies in order to achieve the good ecological status, there is a growing interest in using remote sensing techniques to monitor water quality variables. Remote sensing may allow a highly consistent monitoring to come with essential spatial coverage and temporal resolution when compared to the sporadic and intermittent field sampling work. Lake Albufera de Valencia, located in eastern Spain, is a hypertrophic water body that presents average Chl-a concentrations of up to 100 mg m−3, and peaks are higher than 200 mg m−3. Therefore it requires intensive monitoring to retrieve its water quality conditions, leading to the generation of specific measures for environmental management (Vicente and Miracle, 1992). Right after the Albufera de Valencia and its surroundings were classified as a Natural Park in 1985. López-García and Caselles (1987) carried out the first remote sensing study in this water body. The main goal of this study was to monitor the Chl-a concentration, seston concentration, and water transparency with the aid of Thematic Mapper (TM) sensor onboard Landsat-5 satellite (López-García and Caselles, 1987). To retrieve the general water quality information, they applied regression techniques for estimating each parameter.

Since then, many water quality monitoring studies have focused on mapping different water quality parameters using remote sensing technologies, with varying purposes (e.g. Serrano et al., 1997, Härmä et al., 2001, Peña et al., 2004, Duan et al., 2006, Allan et al., 2011, McCullough et al., 2012). Landsat imageries were extensively used in those studies because of its relatively high spatial resolution of 30 m and its long period of archived data, which allows one to conduct multitemporal change detection of targeted water bodies (e.g., Mayo et al., 1995, Allee and Johnson, 1999, Olmanson et al., 2008). Yet, the sensor onboard Landsat has a limited spectral resolution and this, together with the low temporal resolution (16 days), makes it difficult to determine some of the water quality parameters accurately and timely. Sensors such as MODIS and MERIS, that have higher temporal-spectral resolution, are more suitable to monitor different water quality parameters with time-sensitive reflective spectral signatures (Peña et al., 2004; McCullough et al., 2012); but the spatial resolution of their images is too coarse to analyze medium or small size water bodies (Mancino et al., 2009).

To study highly eutrofied lakes, such as the Albufera de Valencia, the development of the integrated data fusion and mining (IDFM) algorithm with the involvement of both Landsat and MODIS imageries might be a promising tool to get through the hurdle of near real-time monitoring (Chang et al., 2014a, Chang et al., 2014b). A comparison between the main features of both sensors (i.e., Landsat and MODIS) gives rise to some insight about the possibility of sensor fusion, data fusion, or even information fusion (Table 1). To fuse the images collected by MODIS and Landsat, the data fusion algorithms working at the pixel level, such as the Temporal Adaptive Reflectance Fusion Model (STAR-FM) algorithm, may be adopted to enhance spatial, spectral, and temporal properties (Gao et al., 2006). Nevertheless, bio-optical or empirical algorithms, such as statistical regression and computational intelligence algorithms, are required to help classify and interpret fused remote sensing data linking the reflectance over radiance bands of the sensors with ground truth data (Peña et al., 2004, Allan et al., 2011, Domínguez et al., 2011, McCullough et al., 2012, Alonso-Fernández et al., 2013, Chang et al., 2014a, Chang et al., 2014b).

The aim of this paper is to apply the IDFM algorithm developed by Chang et al., 2014a, Chang et al., 2014b to estimate and assess the dynamics of Chl-a concentration and water transparency at the Albufera de Valencia, addressing the water quality status under the agricultural runoff impact. With such understanding of the changing water quality status, water management policy may be examined for sustainable development. In this context, fused STAR-FM Landsat–MODIS images were analyzed by a suite of genetic programming (GP) models, in order to retrieve the water quality status in the Albufera de Valencia. This endeavor leads us to explore the following science questions: 1) Can data fusion techniques be used to fill in the temporal data gaps left by Landsat and capture the spectral features of water transparency and Chl-a concentrations simultaneously during the study period? 2) Can we perform data mining via GP models to effectively retrieve water transparency and Chl-a concentrations based on the fused images at the same time? We hypothesized that the IDFM technique can estimate multiple water quality parameters at the same time, improving the spatial and temporal resolution of the water quality maps for holistic, near real-time environmental monitoring and assessment.

Section snippets

Study area

Albufera de Valencia is a shallow lake located in the Mediterranean coast of Valencia, Spain (39° 20′N, 0° 20′W) (Fig. 1). It is the largest natural water body on the Iberian Peninsula, and its surface is 23.2 km2 with an average depth of 1.2 m (Romo et al., 2008). In 1985, the lake and its surroundings were recognized as a Natural Park. In 1989, the lake was included in the Ramsar international list of protected wetlands. After that, the lake has been a popular site for numerous limnological

Data fusion

For the purpose of demonstration, we show the fused image in September 2006 (days of year – DOY – 269) in Fig. 3. Note that Landsat 7 images have scan line errors resulting in strips which require pretreatment before fusion. To solve the ETM+ gap problem, we applied a bilinear interpolation method for pretreatment (ESRI, 2013). For the purpose of comparison, the pre-condition date is the DOY 262 with and without pretreatment (Fig. 3a, d), respectively, and the post-condition date is the DOY 278

Conclusions

Our study demonstrates that the IDFM method fusing the Landsat and MODIS images, with the aid of STAR-FM algorithm, is useful to improve the temporal resolution of Landsat images. This solves the lengthy data gap posed by Landsat, providing more intensive Earth system observations. The GP models were produced for the estimation of Chl-a concentrations and water transparency simultaneously, based on the synthetic Landsat reflectance values. Satisfactory results were obtained for the estimation

Acknowledgments

This study was jointly supported by the “Ministerio de Economía y Competitividad” by aid granted projects CGL2010-17577-CLI, CGL2013-46862-C2-1/2-P, PROMETEUII/2014/086 to VC and CGL2012-38909 to AC. The authors would like to thank Mr. Yan-Oing Dong for providing the code to solve the issue of Landsat-7 data.

References (44)

  • S.B. Bricker et al.

    National Estuarine Eutrophication Assessment, Effects of Nutrient Enrichment in the Nation's Estuaries

    (September 1999)
  • N.B. Chang et al.

    Integrated data fusion and mining techniques for monitoring total organic carbon concentrations in a lake

    Int. J. Remote Sens.

    (2014)
  • N.B. Chang et al.

    Comparative sensor fusion between hyperspectral and multispectral satellite sensors for monitoring microcystin distribution in Lake Erie

    IEEE J. Sel. Top. Appl. Earth Obs.

    (2014)
  • J.A. Domínguez et al.

    Remote sensing as a tool for monitoring water quality parameters for Mediterranean Lakes of European Union water framework directive (WFD) and as a system of surveillance of cyanobacterial harmful algae blooms (SCyanoHABs)

    Environ. Monit. Assess.

    (2011)
  • J.A. Domínguez-Gómez et al.

    Estudio de aguas continentales mediante teledetección

    (2012)
  • C. Doña et al.

    Empirical relationships for monitoring water quality of lakes and reservoirs through multispectral images

    IEEE J. Sel. Top. Appl. Earth Obs.

    (2014)
  • H. Duan et al.

    Assessment of chlorophyll-a concentration and trophic state for Lake Chagan using Landsat TM and field spectral data

    Environ. Monit. Assess.

    (2006)
  • J. Dwyer et al.

    The MODIS reprojection tool

  • Environmental System Research Institute (ESRI), 2013....
  • European Water Framework Directive (WFD). European Commission. Directive 2000/60/EC of the European Parliament and of...
  • D.F. Francone

    Discipulus Software Owner's Manual

    (1998)
  • F. Gao et al.

    On the blending of Landsat and MODIS surface reflectance: predicting daily Landsat surface reflectance

    IEEE Trans. Geosci. Remote Sens

    (2006)
  • Cited by (84)

    • Water quality classification using machine learning algorithms

      2022, Journal of Water Process Engineering
    View all citing articles on Scopus
    View full text