Elsevier

Science of The Total Environment

Volumes 601–602, 1 December 2017, Pages 998-1007
Science of The Total Environment

Estimating the biomass of unevenly distributed aquatic vegetation in a lake using the normalized water-adjusted vegetation index and scale transformation method

https://doi.org/10.1016/j.scitotenv.2017.05.163Get rights and content

Highlights

  • Remote sensing models for estimating biomass of aquatic vegetation were developed.

  • A new normalized water-adjusted vegetation index was established.

  • A new field biomass scale transformation method was proposed.

  • The proposed modeling scheme had a high accuracy in biomass estimation of aquatic vegetation.

Abstract

Satellite remote sensing is advantageous for the mapping and monitoring of aquatic vegetation biomass at large spatial scales. We proposed a scale transformation (CT) method of converting the field sampling-site biomass from the quadrat to pixel scale and a new normalized water-adjusted vegetation index (NWAVI) based on remotely sensed imagery for the biomass estimation of aquatic vegetation (excluding emergent vegetation). We used a modeling approach based on the proposed CT method and NWAVI as well as statistical analyses including linear, quadratic, logarithmic, cubic, exponential, inverse and power regression to estimate the aquatic vegetation biomass, and we evaluated the performance of the biomass estimation. We mapped the spatial distribution and temporal change of the aquatic vegetation biomass using a geographic information system in a test lake in different months.

The exponential regression models based on CT and the NWAVI had optimal adjusted R2, F and Sig. values in both May and August 2013. The scatter plots of the observed versus the predicted biomass showed that most of the validated field sites were near the 1:1 line. The RMSE, ARE and RE values were small. The spatial distribution and change of the aquatic vegetation biomass in the study area showed clear variability.

Among the NWAVI-based and other vegetation index-based models, the CT and NWAVI-based models had the largest adjusted R2, F and the smallest ARE values in both tests. The proposed modeling scheme is effective for the biomass estimation of aquatic vegetation in lakes. It indicated that the proposed method can provide a most accurate spatial distribution map of aquatic vegetation biomass for lake ecological management. More accurate biomass maps of aquatic vegetation are essential for implementing conservation policy and for reducing uncertainties in our understanding of the lake carbon cycle.

Introduction

There are 304 million natural lakes with an area greater than or equal to 0.1 ha on all continents. These natural lakes cover more than 2.8% of the land surface, not including temporary water bodies and wetlands (Downing and Duarte, 2009). Aquatic vegetation is an essential element in the life systems of most lakes and performs a number of useful functions in maintaining the productivity (Brothers et al., 2013) and biogeochemical cycles (Carpenter and Lodge, 1986) in lakes. It has been reported that macrophytes provide microhabitats for zooplankton (e.g., space and food resources, yielding positive relationships with zooplankton diversity) and that microhabitat structure determines the diversity and abundance of zooplankton communities (i.e., dry weight, species number, and plant type) (Folt and Burns, 1999, Padial et al., 2009, Choi et al., 2014). Macrophytes also provide habitat areas for aquatic insects, fish, and other resident aquatic and semiaquatic organisms and provide structure and food for mammals, birds, reptiles, and amphibians. Aquatic macrophytes can stabilize clear water conditions in shallow lakes (Scheffer et al., 1993, Hilt et al., 2011). Aquatic plants reduce the resuspension of sediment particles (Barko and James, 1998, Vermaat et al., 2000), the water turbidity (Sachse et al., 2014), the light availability for phytoplankton (Pokorný et al., 1984), the phosphorus and the nitrogen (Zuo et al., 2015). In addition, aquatic vegetation serves to anchor soft sediments, stabilize underwater slopes and soft lake bottoms and minimize shoreline erosion by dampening the effect of waves; macrophyte uptake from the interstitial waters is responsible for a significant loss of phosphorus from the sediment (Gudimov et al., 2015). Aquatic vegetation produces and stores significant amounts of carbon as biomass (Means et al., 2016) and represents a major component of the carbon balance due to its fast growth rates and high productivity (Piedade et al., 1991, Costa, 2005, Engle et al., 2008, Silva et al., 2009, Silva et al., 2010). All of these factors emphasize the importance of a healthy, natural aquatic plant community.

Information on the areal biomass and distribution of aquatic vegetation is necessary for the monitoring, management and understanding of lake aquatic ecosystems (Vis et al., 2003). However, determining the properties of aquatic plants such as biomass is difficult at both small and large spatial scales because of the spatial heterogeneity of aquatic plant communities (Duarte and Kalff, 1990, Vis et al., 2003, Silva et al., 2008). Traditional field-based mapping and monitoring of aquatic vegetation biomass at large spatial extents presents several challenges, including the inaccessibility of some areas for field sampling; rapid changes in aquatic plant location, extent, and density; and high costs and travel time. Furthermore, such mapping and monitoring is labor intensive and destructive to sensitive lake ecosystems (Zhang et al., 1997, Jakubauskas et al., 2002), resulting in reduced sampling effort and/or incomplete data sets with limited spatial and temporal coverage (Vis et al., 2003). Given these limitations, biomass estimation through satellite remote sensing is a feasible alternative (Costa, 2005, Silva et al., 2010, Goetz and Dubayah, 2011, Byrd et al., 2014). Remote-sensing methods offer the ability to continuously monitor growth and phenology on the same individuals and can reduce the effort involved in biomass harvesting, but these methods depend on strong relationships between the predictor variables and plant biomass (Armstrong, 1993, Peñuelas et al., 1993, Daoust and Childers, 1998, Zhang, 1998, Silva et al., 2010). Estimating the biomass of aquatic vegetation using remote sensing and statistical regression involves two important parameters: the biomass values of the field sampling-site quadrats and the spectral data. These two parameters directly determine the accuracy and precision of the biomass estimation of aquatic vegetation obtained through satellite remote sensing. Due to the large spatial heterogeneity and uneven distribution of aquatic vegetation in lakes, areas of continuous and uniformly distributed aquatic vegetation that are 30 × 30 m or larger are uncommonly found in field investigations. Thus, the biomass values of field sampling-site quadrats cannot be regarded as identical to those of field sampling-site pixels. If the biomass values of field sampling-site quadrats were directly used to build a model, serious errors or large deviations would result in the biomass estimation based on remote sensing. Therefore, it is necessary to apply a scale transformation to the biomass values of the field sampling-site quadrats before building a statistical model of biomass based on remote sensing. The other parameter used for biomass estimation is the spectral data of the remote-sensing imagery, which mainly include the bands or band combinations (ratios, indices) (Armstrong, 1993, Peñuelas et al., 1993, Zhang, 1998, Silva et al., 2008). To date, published studies (e.g., Peñuelas et al., 1993, Zhang, 1998, Payton, 2001, Schweizer et al., 2005, Ma et al., 2008, Robles, 2009, Byrd et al., 2014, Massicotte et al., 2015) on the use of vegetation indices to estimate the biomass of aquatic vegetation are limited to the normalized difference vegetation index (NDVI), the enhanced vegetation index (EVI), and the normalized difference green index (NDGI). However, these vegetation indices were established mainly for terrestrial vegetation, and the question remains as to whether a new aquatic vegetation index can be established that is more suitable for the remote-sensing estimation of aquatic vegetation biomass in lakes.

The remote-sensing estimation of aquatic vegetation biomass in lakes remains rare, and few studies have used scale transformation and the aquatic vegetation index in the remote-sensing estimation of aquatic vegetation (including floating-leaf and submergent vegetation and excluding emergent vegetation, as described below) biomass in a lake. In the following, we proposed a scale transformation (CT) method of converting field sampling-site biomass from the quadrat to the pixel scale for remote-sensing biomass estimation and a new normalized water-adjusted vegetation index (NWAVI) based on remote-sensed imagery for the biomass estimation of aquatic vegetation. Then, we tested whether it was possible to estimate the biomass of aquatic vegetation in a lake using the proposed CT- and NWAVI-based method and evaluated the performance of the biomass estimation. Finally, we mapped the spatial distribution and temporal change of the aquatic vegetation biomass in a test lake in different months.

Section snippets

Study area

Lake Taihu, the third largest freshwater lake in China, is located in the south of Jiangsu Province, China (119°53′49″–120°35′25″E and 30°55′32″–31°32′50″N) (Fig. 1). It has a surface area of 2338 km2 and average and maximum water depths of 1.89 m and 2.6 m, respectively. Before the 1960s, Lake Taihu had a large area of aquatic vegetation (Nanjing Institute of Geography, Chinese Academy of Sciences, 1965). According to recent field surveys, the distribution of aquatic vegetation was widest for

Field data: quadrat- and pixel-scale biomass

In May 2013, the biomass values of the thirty-five field sampling-site quadrats ranged from 0.17 to 6.64 kg/m2 with an average of 1.75 kg/m2 and a standard deviation of 1.34 kg/m2; the biomass values of the thirty-five field sampling-site pixels ranged from 0.03 to 3.29 kg/m2 with an average of 0.73 kg/m2 and a standard deviation of 0.75 kg/m2. Compared with the values from the corresponding quadrats, the biomass values of the thirty-five field sampling-site pixels were reduced from − 34.59% to − 

Importance of field biomass-scale transformation

In most studies estimating the biomass of aquatic vegetation using remote sensing (e.g., Peñuelas et al., 1993, Zhang, 1998, Payton, 2001, Schweizer et al., 2005, Ma et al., 2008, Robles, 2009, Byrd et al., 2014, Massicotte et al., 2015), measurements of aboveground biomass of aquatic vegetation made during field surveys were used to model the relationship between the field biomass and the spectral data of the remote-sensing imagery. Byrd et al. (2014) averaged the biomass from multiple plots

Data accessibility

The remote-sensing data used in this manuscript have been archived on the website of the China Center For Resources Satellite Data and Application (http://www.cresda.com).

Acknowledgments

This research was financially supported by the Major Science and Technology Program for Water Pollution Control and Treatment (Grant numbers 2012ZX07501-001-03 and 2012ZX07506-001). The authors would like to thank the China Center for Resources Satellite Data and Application (CCRSDA) for their significant efforts in developing and distributing the remotely sensed satellite data and products and for their kindness in providing online accessibility to download this information.

References (52)

  • O. Mutanga et al.

    High density biomass estimation for wetland vegetation using WorldView-2 imagery and random forest regression algorithm

    Int. J. Appl. Earth Obs. Geoinf.

    (2012)
  • J. Peñuelas et al.

    Assessing community type, plant biomass, pigment composition, and photosynthetic efficiency of aquatic vegetation from spectral reflectance

    Remote Sens. Environ.

    (1993)
  • J. Pokorný et al.

    Production-ecological analysis of a plant community dominated by Elodea canadensis Michx

    Aquat. Bot.

    (1984)
  • R. Sachse et al.

    Extending one-dimensional models for deep lakes to simulate the impact of submerged macrophytes on water quality

    Environ. Model. Softw.

    (2014)
  • M. Scheffer et al.

    Alternative equilibria in shallow lakes

    Trends Ecol. Evol.

    (1993)
  • T.S.F. Silva et al.

    Assessment of two biomass estimation methods for aquatic vegetation growing on the Amazon floodplain

    Aquat. Bot.

    (2010)
  • W.J. Timmermans et al.

    An intercomparison of the surface energy balance algorithm for land SEBAL and the two source energy balance TSEB modeling schemes

    Remote Sens. Environ.

    (2007)
  • C.J. Tucker

    Red and photographic infrared linear combinations for monitoring vegetation

    Remote Sens. Environ.

    (1979)
  • P. Villa et al.

    Aquatic vegetation indices assessment through radiative transfer modeling and linear mixture simulation

    Int. J. Appl. Earth Obs. Geoinf.

    (2014)
  • C. Vis et al.

    An evaluation of approaches used to determine the distribution and biomass of emergent and submerged aquatic macrophytes over large spatial scales

    Aquat. Bot.

    (2003)
  • D. Zhao et al.

    Remote sensing of aquatic vegetation distribution in Taihu Lake using an improved classification tree with modified thresholds

    J. Environ. Manag.

    (2012)
  • S. Zuo et al.

    Effect of allelopathic potential from selected aquatic macrophytes on algal interaction in the polluted water

    Biochem. Syst. Ecol.

    (2015)
  • R.A. Armstrong

    Remote sensing of submerged vegetation canopies for biomass estimation

    Int. J. Remote Sens.

    (1993)
  • J.W. Barko et al.

    Effects of submerged macrophytes on nutrient dynamics, sedimentation, and resuspension

  • S. Brothers et al.

    Plant community structure determines primary productivity in shallow, eutrophic lakes

    Freshw. Biol.

    (2013)
  • P.S. Chavez

    Image-based atmospheric correction-revisited and improved

    Photogramm. Eng. Remote Sens.

    (1996)
  • Cited by (25)

    • Impact of upstream landslide on perialpine lake ecosystem: An assessment using multi-temporal satellite data

      2020, Science of the Total Environment
      Citation Excerpt :

      Monitoring capabilities demonstrated by the presented approach were amplified by the specific characteristics of Sentinel-2, in terms of data revisit (5–10 days), spatial resolution (up to 10 m) and spectral content extending from visible to shortwave infrared range, that meet the requirements for effective, operational ecosystem monitoring (Kutser et al., 2006; Song et al., 2009). Operational mid-resolution satellite data have recently been used for environmental applications on aquatic systems focusing, for example on: riparian vegetation classification (Stratoulias et al., 2015), mapping water quality and submerged macrophyte cover (Dörnhöfer et al., 2016; Fritz et al., 2019), estimating aquatic vegetation biomass (Gao et al., 2017), assessing the spatio-temporal evolution of primary producers (Pinardi et al., 2018; Ghirardi et al., 2019), analysing the seasonal dynamics of wetland plant communities (Villa et al., 2018), and mapping cyanobacteria blooms (Sòria-Perpinyà et al., 2020). Our work further shows the capabilities of mid-resolution satellite data and enlarges the scope with respect to previous studies, by providing a comprehensive application that joins water quality (turbidity) and aquatic to wetland vegetation (submerged macrophytes and helophytes) monitoring.

    View all citing articles on Scopus
    View full text