Camera derived vegetation greenness index as proxy for gross primary production in a low Arctic wetland area

https://doi.org/10.1016/j.isprsjprs.2013.09.006Get rights and content

Abstract

The Arctic is experiencing disproportionate warming relative to the global average, and the Arctic ecosystems are as a result undergoing considerable changes. Continued monitoring of ecosystem productivity and phenology across temporal and spatial scales is a central part of assessing the magnitude of these changes. This study investigates the ability to use automatic digital camera images (DCIs) as proxy data for gross primary production (GPP) in a complex low Arctic wetland site. Vegetation greenness computed from DCIs was found to correlate significantly (R2 = 0.62, p < 0.001) with a normalized difference vegetation index (NDVI) product derived from the WorldView-2 satellite. An object-based classification based on a bi-temporal image composite was used to classify the study area into heath, copse, fen, and bedrock. Temporal evolution of vegetation greenness was evaluated and modeled with double sigmoid functions for each plant community. GPP at light saturation modeled from eddy covariance (EC) flux measurements were found to correlate significantly with vegetation greenness for all plant communities in the studied year (i.e., 2010), and the highest correlation was found between modeled fen greenness and GPP (R2 = 0.85, p < 0.001). Finally, greenness computed within modeled EC footprints were used to evaluate the influence of individual plant communities on the flux measurements. The study concludes that digital cameras may be used as a cost-effective proxy for potential GPP in remote Arctic regions.

Introduction

Recent estimates of terrestrial carbon (C) stored in Arctic soils suggests a pool of 1400–1850 Pg of C (McGuire et al., 2009, Tarnocai et al., 2009), totaling more than twice the estimated atmospheric C pool (Lal, 2003). In the Arctic, the near-surface warming trend has been almost twice the global average throughout recent decades (ACIA, 2005, Graversen et al., 2008). This warming affects the physical and ecological processes in the region (McGuire et al., 2009); indeed, evidence indicates that this warming has accelerated the exchange of C in the terrestrial ecosystem (Loranty and Goetz, 2012, Myers-Smith et al., 2011), advanced spring snowmelt by approximately 18 days, and increased vegetation productivity by 18% over the last two to three decades (Kimball et al., 2007, Tedesco et al., 2009). Nevertheless, the nature of climatically induced changes to ecosystem production is complex and occurs over different temporal scales. Regularly repeated measurements are thus essential to monitor and evaluate changes in productivity and ecosystem integrity (ABA, 2013, Lindenmayer et al., 2010).

Estimations of ecosystem productivity based on remote sensing have shown that data from currently orbiting satellites can be used to monitor and estimate both gross primary productivity (GPP) (Hilker et al., 2008, Rossini et al., 2012, Turner et al., 2006) and net ecosystem exchange (NEE) (Loranty et al., 2011). This is accomplished by combining spectral vegetation indices (VIs) with ecosystem productivity models. Yet, the effectiveness of optical multi-temporal satellite data is constrained in Arctic environments due to (1) a short growing season of 45–75 days (Bliss, 1971), (2) persistent cloud cover (Eastman and Warren, 2010, Stow et al., 2004), (3) solar geometry, (4) standing water, (5) snow and ice cover (Beck et al., 2006, Hope and Stow, 1996), and (6) sparse vegetation that can grow in significantly different vegetation communities and in patches down to a few m2 (Fletcher et al., 2011, Rosswall and Heal, 1978, Shaver et al., 2007). Hence, the parameterization of productivity models using representative biophysical data is challenging. For instance, GPP products with high temporal resolution, such as the MOD17A2 (i.e., an 8-day composite at 1 km spatial resolution) from Terra MODIS, has been shown to overestimate GPP in sparsely vegetated biomes (Turner et al., 2006) partly due to difficulties in finding representative input values for the large area pixels used in the GPP model (Chasmer et al., 2009). Variations in vegetation, barren ground, and snow cover in the meter size range also result in pronounced subpixel differences when monitoring vegetation with medium-resolution sensors, such as Terra MODIS or NOAA AVHRR. In contrast, higher spatial resolution data (e.g., from Landsat TM and Terra ASTER) cannot fully capture intra-seasonal variations in productivity during the short growing season due to the limited overpass frequency of 16 days. Furthermore, the spatial resolution is insufficient for the observation of vegetation at the plant community level. Finally, cloud cover can lead to incomplete time series for all space-borne optical sensors and even for polar orbiting satellites with several daily overpasses, cloud-free observations may not occur for weeks at a time (Stow et al., 2004).

Ultimately, researchers need to develop techniques to detect ecological surprises in the rapidly changing Arctic, i.e., unexpected findings or outcomes that had not been originally hypothesized (Lindenmayer et al., 2010). Since optical satellite data may not detect such surprises at a local scale, and field measurements in the Arctic are constrained by harsh weather conditions, logistics (Newberry and Southwell, 2009), and an inherent limit on the spatial extent (Wiens, 1989), the study of techniques to support satellite data with repeated ground-based spectral measurements is highly relevant. Indeed, improved spatial and temporal scale can be achieved by monitoring ecosystems with automated digital cameras. Cameras can provide sufficient resolution to capture seasonal changes at the plant community level while also linking field measurements to satellite imagery. Indices based on the 3-dimensional color space of red, green, and blue (RGB), i.e., what is available from digital cameras, have been found to correlate with the normalized difference vegetation index (NDVI) over both cropland and boreal forests (Adamsen et al., 1999, Lebourgeois et al., 2008, Richardson et al., 2007). Ahrends et al. (2009) found that camera derived vegetation greenness correlated with GPP in a boreal forest canopy, and Migliavacca (2011) found a similar relationship in a homogenous Alpine grassland. However, among the few existing studies on GPP and greenness, none investigated performance in Arctic environments that had high variability in terms of species composition. Therefore, in this study, we investigated the potential use of digital cameras as a non-disturbing and cost-effective proxy for GPP within a complex low Arctic wetland site in Greenland. The method allows for the monitoring of vegetation dynamics at a high resolution (i.e., daily images at a 0.3–2 m spatial resolution), thereby allowing the detection of seasonal vegetation patterns and unexpected responses in vegetation greenness. Greenness was computed from the RGB color space, and the results are compared to ground truth classifications of fen, copse, and heath. Modeled GPP data from eddy covariance (EC) measurements and an NDVI product from the sun-synchronous WorldView-2 satellite were subsequently used for a comparison with the greenness results from the vegetation classes. Finally, modeled greenness data from the camera images were used to equalize the greenness signal and analyze spatial differences between the three plant communities.

Section snippets

Study site

This study was carried out between May 30th and September 7th of 2010 in Kobbefjord, southwestern Greenland (64°07′N; 51°21′W). The site is located 20 km from the Greenlandic capital of Nuuk. Kobbefjord is the subject of an extensive cross-disciplinary ecological monitoring program (NERO, http://www.nuuk-basic.dk) by Greenland Ecosystem Monitoring (GEM). The monitored area is located approximately 550 m from the head of a fjord, approximately 40 m above sea level (Fig. 1). From 1961 to 1990, the

Comparison with satellite data

Greenness indices and NDVI from the WorldView-2 image were stacked using ArcMap 10 (ESRI, 2012) and analyzed for spatial correlation using a band statistics local function (Fig. 5). The local functions work on a cell by cell basis; i.e., each output cell value from the computation is a function of input values at the same spatial location in different data layers.

The spatial correlations between the WorldView-2 NDVI, the 2G_RBi, and the Channel%g were significant (p < 0.001). The correlation

Discussion

The spatial comparison of RGB-based VIs and satellite NDVI showed significant correlation in our study, supporting the usefulness of cameras as a source of spatial data (Table 3). The correlation was sensitive to changes in lens distortion correction, especially the geometric distortion. This underlines the importance of precision for all parameters when orthorectifying high resolution image data. Channel%g was resampled to 2 m to match the resolution of the WorldView-2 image; additionally,

Conclusion

GPP varies for different Arctic vegetation classes (Kade et al., 2012), which underlines the complexity of interpreting the signal from EC measurements in heterogeneous landscapes. The use of visual tools allowed us to spatially separate GPP signals in a complex, low Arctic wetland area. The different slope of the linear relationship between GPP at light saturation and the measured Channel%g for the three vegetation classes shows that we need to address the variations in vegetation cover when

Acknowledgments

This work was supported by Greenland Ecosystem Monitoring (GEM), The Danish Energy Agency, The Environmental Protection Agency, and participating institutions. We wish to thank Bo Holm Rasmussen (University of Copenhagen) and Anders Scheel Nielsen (Technical University of Denmark) for their support in computer programming. Moreover, we wish to acknowledge the staff at the Center for Permafrost (CENPERM) for valuable inputs.

References (83)

  • T. Key et al.

    A comparison of multispectral and multitemporal information in high spatial resolution imagery for classification of individual tree species in a temperate hardwood forest

    Remote Sensing of Environment

    (2001)
  • Y. Kim et al.

    Satellite detection of increasing Northern Hemisphere non-frozen seasons from 1979 to 2008: implications for regional vegetation growth

    Remote Sensing of Environment

    (2012)
  • R. Lal

    Soil erosion and the global carbon budget

    Environment International

    (2003)
  • M. Migliavacca et al.

    Using digital repeat photography and eddy covariance data to model grassland phenology and photosynthetic CO2 uptake

    Agricultural and Forest Meteorology

    (2011)
  • A.M. Moffat et al.

    Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes

    Agricultural and Forest Meteorology

    (2007)
  • J.B. Moncrieff et al.

    A system to measure surface fluxes of momentum, sensible heat, water vapour and carbon dioxide

    Journal of Hydrology

    (1997)
  • J.I. Peltoniemi et al.

    BRDF measurement of understory vegetation in pine forests: dwarf shrubs, lichen, and moss

    Remote Sensing of Environment

    (2005)
  • J.M. Peña-Barragán et al.

    Object-based crop identification using multiple vegetation indices, textural features and crop phenology

    Remote Sensing of Environment

    (2011)
  • H. Soegaard et al.

    Carbon dioxide exchange over agricultural landscape using eddy correlation and footprint modelling

    Agricultural and Forest Meteorology

    (2003)
  • D.A. Stow et al.

    Remote sensing of vegetation and land-cover change in Arctic tundra ecosystems

    Remote Sensing of Environment

    (2004)
  • D.P. Turner et al.

    Evaluation of MODIS NPP and GPP products across multiple biomes

    Remote Sensing of Environment

    (2006)
  • T. Vesala et al.

    Flux and concentration footprint modelling: State of the art

    Environmental Pollution

    (2008)
  • L.A. Vierling et al.

    Differences in arctic tundra vegetation type and phenology as seen using bidirectional radiometry in the early growing season

    Remote Sensing of Environment

    (1997)
  • X.Y. Zhang et al.

    Monitoring vegetation phenology using MODIS

    International Journal of Environment

    (2003)
  • ABA, 2013. Arctic Biodiversity Assessment. Status and trends in Arctic biodiversity. In: Meltofte, H. (Ed.)....
  • ACIA, 2005. Arctic Climate Impact Assessment – Scientific Report. Cambridge University...
  • F.G. Adamsen et al.

    Measuring Wheat Senescence with a Digital Camera

    Crop Science

    (1999)
  • H.E. Ahrends et al.

    Tree phenology and carbon dioxide fluxes: use of digital photography for process-based interpretation at the ecosystem scale

    Climate Research

    (2009)
  • M. Arndal et al.

    Seasonal variation in gross ecosystem production, plant biomass, and carbon and nitrogen pools in five high arctic vegetation types

    Arctic, Antarctic, and Alpine Research

    (2009)
  • P.M. Atkinson et al.

    Spatial scale problems and geostatistical solutions: a review

    The Professional Geographer

    (2000)
  • Aubinet, M., Grelle, A., Ibrom, A., Rannik, Ü., Moncrieff, J., Foken, T., Kowalski, A.S., Martin, P.H., Berbigier, P.,...
  • C. Bay et al.

    The NERO line. A vegetation transect in Kobbefjord, West Greenland

    (2008)
  • L.C. Bliss

    Arctic and alpine plant life cycles

    Annual Review of Ecology and Systematics

    (1971)
  • Callaghan, T.V., Johansson, M., Brown, R.D., Groisman, P.Y., Labba, N., Radionov, V., 2011. Chapter 4. Changing snow...
  • J.G. Corripio

    Snow surface albedo estimation using terrestrial photography

    International Journal of Remote Sensing

    (2004)
  • K.M. de Beurs et al.

    A statistical framework for the analysis of long image time series

    International Journal of Remote Sensing

    (2005)
  • L.S. Defoliart et al.

    Seasonal patterns of photosynthesis and nutrient storage in Eriophorum vaginatum L., an arctic sedge

    Functional Ecology

    (1988)
  • R. Eastman et al.

    Arctic cloud changes from surface and satellite observations

    Journal of Climate

    (2010)
  • B.J. Fletcher et al.

    Photosynthesis and productivity in heterogeneous arctic tundra: consequences for ecosystem function of mixing vegetation types at stand edges

    Journal of Ecology

    (2011)
  • S.E. Frolking et al.

    Relationship between ecosystem productivity and photosynthetically active radiation for northern peatlands

    Global Biogeochemical Cycles

    (1998)
  • J.H.C. Gash

    A note on estimating the effect of a limited fetch on micrometeorological evaporation measurements

    Boundary-Layer Meteorology

    (1986)
  • Cited by (0)

    View full text