Elsevier

Science of The Total Environment

Volume 649, 1 February 2019, Pages 284-299
Science of The Total Environment

Evaluating temporal controls on greenhouse gas (GHG) fluxes in an Arctic tundra environment: An entropy-based approach

https://doi.org/10.1016/j.scitotenv.2018.08.251Get rights and content

Highlights

  • GHG fluxes show significant spatio-temporal variability in low-gradient Arctic tundra.

  • Variability in CO2 fluxes was governed by soil temperature and vegetation dynamics.

  • Variability in CH4 fluxes was governed by seasonal vegetation and thaw dynamics.

  • Entropy classification technique can be used to identify years with higher GHG fluxes (e.g., 2014).

  • Environmental managers can use entropy scheme as a generic tool for other data of interest.

Abstract

There is significant spatial and temporal variability associated with greenhouse gas (GHG) fluxes in high-latitude Arctic tundra environments. The objectives of this study are to investigate temporal variability in CO2 and CH4 fluxes at Barrow, AK and to determine the factors causing this variability using a novel entropy-based classification scheme. In particular, we analyzed which geomorphic, soil, vegetation and climatic properties most explained the variability in GHG fluxes (opaque chamber measurements) during the growing season over three successive years. Results indicate that multi-year variability in CO2 fluxes was primarily associated with soil temperature variability as well as vegetation dynamics during the early and late growing season. Temporal variability in CH4 fluxes was primarily associated with changes in vegetation during the growing season and its interactions with primary controls like seasonal thaw. Polygonal ground features, which are common to Arctic regions, also demonstrated significant multi-year variability in GHG fluxes. Our results can be used to prioritize field sampling strategies, with an emphasis on measurements collected at locations and times that explain the most variability in GHG fluxes. For example, we found that sampling primary environmental controls at the centers of high centered polygons in the month of September (when freeze-back period begins) can provide significant constraints on GHG flux variability – a requirement for accurately predicting future changes to GHG fluxes. Overall, entropy results document the impact of changing environmental conditions (e.g., warming, growing season length) on GHG fluxes, thus providing clues concerning the manner in which ecosystem properties may be shifted regionally in a future climate.

Introduction

Identifying key factors causing temporal variability in CO2 and CH4 fluxes has been the subject of considerable research over the past two decades (e.g., Arora et al., 2016b; Bousquet et al., 2006; Janssens et al., 2001; Schimel et al., 2001). Temporal variability in carbon fluxes has been linked to environmental factors such as snowmelt timing, growing season dynamics, water table variations and temperature fluctuations (e.g., Arora et al., 2013; Grant et al., 2017; Yabusaki et al., 2017). In particular, Zona et al. (2009) showed that CH4 fluxes in the growing season were strongly correlated with soil temperature and non-linearly correlated with water table depth. Harper et al. (2005) showed that decreasing the amount and increasing the timing between rainfall events decreased CO2 fluxes over four growing seasons (1998–2001). Changes in plant productivity have also been correlated with seasonal and annual variability in carbon fluxes (Janssens et al., 2001; Street et al., 2007).

While several drivers of temporal variability in GHG fluxes have been identified, the relationship between GHG fluxes and these drivers shows considerable variability in space and time, thereby contributing to significant uncertainties in estimating future changes to landscape-level carbon budgets. For example, Friborg et al. (2000) indicated that CH4 fluxes were related to soil temperature and water table in the late part of the summer, whereas the thickness of the thaw layer was the most important control in the early part of the season. Similarly, Grogan and Chapin III (1999) indicated that climate (temperature) had strong effects on belowground CO2 release in both summer and winter seasons while the type of vegetation only impacted summer CO2 efflux. Contrary to these findings, a separate study by Bubier et al. (2003) suggested that the effect of vegetation type on growing season CO2 efflux varied significantly between wet and dry years. Together, these studies suggest that different environmental factors can become important under different spatio-temporal settings. Moreover, recent studies have shown that temporal variability in environmental constraints may itself be unknown or masked by other variables. For example, Malhotra and Roulet (2015) showed that temperature sensitivity of CH4 increased with increasing thaw, but this trend was not found to be consistent and suggested confounding effects of substrate or water limitation on the apparent temperature sensitivity. It is thus important to understand the mechanistic and site-specific nature of relationships between greenhouse gas fluxes and environmental factors, and quantitatively attribute temporal variability to specific factors at a given site.

Understanding the variable nature of relationships between GHG fluxes and environmental factors is particularly important in Arctic tundra environments because of the vast amount of soil carbon stored in these regions and the potential of these regions to convert from a global carbon sink to a source under warmer conditions (Billings et al., 1982; Oechel et al., 2000; Sistla et al., 2013). These relationships can be especially complex and difficult to interpret in Arctic environments because shifts in the timing of snowmelt and plant phenology can strongly influence CH4 and CO2 fluxes. For example, Mastepanov et al. (2013) showed that the differences in growing season CH4 fluxes over 2006–2010 could not be explained by corresponding changes in driving factors like soil temperature or moisture. Instead, they found increases in CH4 fluxes to be related to the date of snowmelt and recommended using the first day of snowmelt as a proxy for the start of the growing season. Raz-Yaseef et al. (2017) linked spring observations of carbon fluxes at a site in Barrow, Alaska (the same site as this study) to the delayed release of biogenic gas production from the previous fall season. Other studies have suggested that the onset and length of the growing season may be shifted by several days in higher latitudes, which can explain some of the variability observed in greenhouse gas fluxes in these regions (Liston et al., 2002; Stow et al., 2004; Tucker et al., 2001).

Temporal variability in GHG fluxes and their relationship to different drivers can be described by simple descriptive statistics (e.g., range, standard deviation, coefficient of variation) or advanced statistical methods (e.g., principal component analysis, K-means clustering) (e.g., Arora and Mohanty, 2017; Dwivedi et al., 2013, Dwivedi et al., 2016). However, simple descriptive statistics have limited use as different environmental factors may demonstrate a number of identical descriptive statistical properties (Matejka and Fitzmaurice, 2017). Moreover, other statistical methods (e.g., correlation analysis, K-means clustering, principal components analysis) typically work under the assumptions of normality or describe linear relationships between variables. Investigating the degree to which environmental factors can impact GHG fluxes in Arctic tundra environments thus requires an integrated approach that can take into account the temporal shifts and complex spatial interactions between predictor and response variables. In this context, entropy methods have proven to be useful in determining the relative contributions of hydrologic interactions, vegetation structure, spatial zonation and other environmental factors to system dynamics (Arora et al., 2016a; Brunsell and Wilson, 2013; Dwivedi and Mohanty, 2016; Ruddell et al., 2013). Moreover, considering the fact that environmental data are naturally stochastic and nonlinear (Reimann and Filzmoser, 2000), we chose to employ trans-information – a nonlinear entropy technique – to extract dependencies between GHG fluxes and environmental variables. Trans-information is defined as a measure of the amount of information that one random variable (e.g., a primary environmental control like soil temperature) contains or explains about another random variable (e.g., GHG fluxes). The main advantage of using trans-information over other techniques is that it is a non-parametric approach that can integrate complex, multivariate datasets without making assumptions regarding the nature of functional dependencies implicit in these datasets (Arora et al., 2016a; Costa et al., 2002). Identifying these dependencies can be particularly useful for developing upscaling strategies, closing the gap with field observations as well as improving the representation of soil carbon stocks and their response to climate change in community land models. In addition, several studies have emphasized the power and strength of trans-information and entropy-based analyses in comparison to other commonly-used statistical approaches such as correlation analysis and classification methods (e.g., Battiti, 1994; Mogheir et al., 2004; Strehl et al., 2000). Considering these advantages, we use a novel classification scheme (described in more detail below) that uses trans-information to disentangle the complex relationships between environmental variables and GHG fluxes under different spatio-temporal settings.

The objectives of this study are to characterize temporal variability in CO2 and CH4 fluxes and investigate possible controls of such variations at a high-Arctic location near Barrow, Alaska using a novel entropy-based classification scheme. To reach these objectives, we chose a set of topographic locations across the site where we have measurements of soil, vegetation and climate parameters as well as greenhouse gas fluxes during three growing seasons (2012–2014). The remainder of this paper is organized as follows. Section 2 describes the Barrow field site, lists a set of potential factors that may impact GHG flux variations based on previous site investigations, and documents field datasets and observations available for the entropy analysis. Details of the entropy-based classification scheme are provided in Section 3. Section 4 presents the entropy analysis results for CO2 and CH4 fluxes and an example for extending the use of the classification scheme to other variables of interest. A summary of the important findings is provided in Section 5.

Section snippets

Site description

Our study site is located within the Barrow Environmental Observatory (BEO) (71.3°N, 156.61°E) in Arctic Alaska (Fig. 1a). The study site in Barrow, AK has been the subject of intensive investigation of climate change impacts on ecosystem processes as part of the Department of Energy's (DOE) Next Generation Ecosystem Experiments (NGEE-Arctic) project. Although comprehensive descriptions of the NGEE-Arctic “Barrow” site can be found elsewhere (Hubbard et al., 2013; Liljedahl et al., 2012,

Shannon's informational entropy

We used the information theory metrics of Shannon's entropy to examine the temporal variability in carbon fluxes as a function of environmental factors, geomorphic features and other primary controls (Shannon, 1948a, Shannon, 1948b). Previous studies have successfully used these metrics to characterize temporal variability in climatological, geochemical, and other complex data series (e.g., Arora et al., 2016b; Balzter et al., 2015; Kawachi et al., 2001; Rajsekhar et al., 2012). In information

Site trends in CO2 and CH4 fluxes

Fig. 4 shows the mean fluxes of CO2 and CH4 across polygon types for 2012, 2013 and 2014 growing season. While mean CO2 fluxes show minor differences across polygon types (small range of variation), mean CH4 fluxes show clear patterns with highest fluxes being associated with LCPs. The temporal patterns for both CO2 and CH4 fluxes show more variability. In particular, GHG fluxes show an increase from 2012 to 2013 and then a decrease from 2013 to 2014. One exception to this trend is that CO2

Summary and conclusions

We used an entropy-based approach to identify dominant environmental factors associated with significant variability in GHG fluxes in Arctic tundra environments, where climate change appears to be rapidly impacting ecosystem processes. In particular, we classified growing season flux data from 2012 to 2014 using a variety of environmental factors and topographic positions across the Barrow site. CO2 fluxes in 2014 were found to be significantly different than the other two sampling seasons.

Acknowledgments

This material is based upon work supported as part of the Next-Generation Ecosystem Experiments (NGEE-Arctic) at Lawrence Berkeley National Laboratory funded by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research under Award Number DE-AC02-05CH11231.

References (99)

  • H. Balzter et al.

    Multi-scale entropy analysis as a method for time-series analysis of climate data

    Climate

    (2015)
  • D. Bastviken et al.

    Methane emissions from lakes: dependence of lake characteristics, two regional assessments, and a global estimate

    Glob. Biogeochem. Cycles

    (2004)
  • R. Battiti

    Using mutual information for selecting features in supervised neural net learning

    IEEE Trans. Neural Netw.

    (1994)
  • W.D. Billings et al.

    Arctic tundra: a source or sink for atmospheric carbon dioxide in a changing environment?

    Oecologia

    (1982)
  • P. Bousquet et al.

    Contribution of anthropogenic and natural sources to atmospheric methane variability

    Nature

    (2006)
  • N. Brunsell et al.

    Multiscale interactions betweenWater and carbon fluxes and environmental variables in a central U.S. grassland

    Entropy

    (2013)
  • J.L. Bubier et al.

    Spatial and temporal variability in growing-season net ecosystem carbon dioxide exchange at a large peatland in Ontario, Canada

    Ecosystems

    (2003)
  • F.S. Chapin et al.

    Role of land-surface changes in arctic summer warming

    Science

    (2005)
  • M. Costa et al.

    Multiscale entropy analysis of complex physiologic time series

    Phys. Rev. Lett.

    (2002)
  • C.J. Cox et al.

    Drivers and environmental responses to the changing annual snow cycle of Northern Alaska

    Bull. Am. Meteorol. Soc.

    (2017)
  • B. Dafflon et al.

    Geophysical estimation of shallow permafrost distribution and properties in an ice-wedge polygon-dominated Arctic tundra region

    Geophysics

    (2016)
  • B. Dafflon et al.

    Coincident aboveground and belowground autonomous monitoring to quantify covariability in permafrost, soil, and vegetation properties in Arctic tundra

    J. Geophys. Res. Biogeosci.

    (2017)
  • S.J. Davidson et al.

    Vegetation type dominates the spatial variability in CH4 emissions across multiple Arctic tundra landscapes

    Ecosystems

    (2016)
  • D. Dwivedi

    Texas Water Resources: Vulnerability from Contaminants

    (2012)
  • D. Dwivedi et al.

    Hot spots and persistence of nitrate in aquifers across scales

    Entropy

    (2016)
  • D. Dwivedi et al.

    Estimating Escherichia coli loads in streams based on various physical, chemical, and biological factors

    Water Resour. Res.

    (2013)
  • D. Dwivedi et al.

    Impact of the Linked Surface Water-Soil Water-Groundwater System on Transport of E. coli in the Subsurface

    Water. Air. Soil Pollut.

    (2016)
  • R. Engstrom et al.

    Spatial distribution of near surface soil moisture and its relationship to microtopography in the Alaskan Arctic coastal plain

    Hydrol. Res.

    (2005)
  • J.C. von Fischer et al.

    Vegetation height and other controls of spatial variability in methane emissions from the Arctic coastal tundra at Barrow, Alaska

    J. Geophys. Res.

    (2010)
  • T. Friborg et al.

    Trace gas exchange in a high-Arctic valley: 2. Landscape CH 4 fluxes measured and modeled using eddy correlation data

    Glob. Biogeochem. Cycles

    (2000)
  • J. a Gamon et al.

    Microtopographic patterns in an arctic baydjarakh field: do fine-grain patterns enforce landscape stability?

    Environ. Res. Lett.

    (2012)
  • C. Gangodagamage et al.

    Extrapolating active layer thickness measurements across Arctic polygonal terrain using LiDAR and NDVI data sets

    Water Resour. Res.

    (2014)
  • N. Gaur et al.

    Evolution of physical controls for soil moisture in humid and subhumid watersheds

    Water Resour. Res.

    (2013)
  • R.F. Grant et al.

    Mathematical modelling of Arctic polygonal tundra with Ecosys: 2. Microtopography determines how CO2 and CH4 exchange responds to changes in temperature and precipitation

    J. Geophys. Res. Biogeosci.

    (2017)
  • P. Grogan et al.

    Arctic soil respiration: effects of climate and vegetation depend on season

    Ecosystems

    (1999)
  • D.R. Harp et al.

    Effect of soil property uncertainties on permafrost thaw projections: a calibration-constrained analysis

    Cryosphere

    (2016)
  • C.W. Harper et al.

    Increased rainfall variability and reduced rainfall amount decreases soil CO2 flux in a grassland ecosystem

    Glob. Chang. Biol.

    (2005)
  • M. Helbig et al.

    Spatial and seasonal variability of polygonal tundra water balance: Lena River Delta, northern Siberia (Russia)

    Hydrogeol. J.

    (2013)
  • E.M. Herndon et al.

    Geochemical drivers of organic matter decomposition in Arctic tundra soils

    Biogeochemistry

    (2015)
  • K.M. Hinkel et al.

    Spatial and temporal patterns of active layer thickness at circumpolar active layer monitoring (CALM) sites in northern Alaska, 1995–2000

    J. Geophys. Res.

    (2003)
  • K.M. Hinkel et al.

    Spatial extent, age, and carbon stocks in drained thaw Lake basins on the barrow Peninsula, Alaska. Arctic

    Antarct. Alp. Res.

    (2003)
  • S.S. Hubbard et al.

    Quantifying and relating land-surface and subsurface variability in permafrost environments using LiDAR and surface geophysical datasets

    Hydrogeol. J.

    (2013)
  • IPCC

    Summary for policymakers

  • I.A. Janssens et al.

    Productivity overshadows temperature in determining soil and ecosystem respiration across European forests

    Glob. Chang. Biol.

    (2001)
  • I. Juszak et al.

    Contrasting radiation and soil heat fluxes in Arctic shrub and wet sedge tundra

    Biogeosciences

    (2016)
  • W.H. Kruskal et al.

    Use of ranks in one-criterion variance analysis

    J. Am. Stat. Assoc.

    (1952)
  • J. Kumar et al.

    Modeling the spatiotemporal variability in subsurface thermal regimes across a low-relief polygonal tundra landscape

    Cryosphere

    (2016)
  • A.H. Lachenbruch et al.

    Changing climate: geothermal evidence from permafrost in the Alaskan Arctic

    Science

    (1986)
  • M.J. Lara et al.

    Polygonal tundra geomorphological change in response to warming alters future CO2 and CH4 flux on the Barrow Peninsula

    Glob. Chang. Biol.

    (2015)
  • Cited by (20)

    • Annual ecosystem respiration of maize was primarily driven by crop growth and soil water conditions

      2019, Agriculture, Ecosystems and Environment
      Citation Excerpt :

      In a study of carbon emissions on the Loess Plateau, wetter and warmer soil was found to stimulate crop growth and biomass accumulation, thereby promoting crop photosynthesis and respiration (Gong et al., 2015). The results in high latitudes indicated that the variability of CO2 flux is mainly related to soil temperature and vegetation dynamics in growing season (Arora et al., 2018). Soil moisture restricted soil respiration by limiting the effective activity of microorganisms and reducing microbial vitality.

    View all citing articles on Scopus
    View full text