Comparison of net ecosystem carbon exchange estimation in a mixed temperate forest using field eddy covariance and MODIS data
© Wang et al. 2016
Received: 9 November 2015
Accepted: 11 April 2016
Published: 21 April 2016
Quantification of net ecosystem carbon exchange (NEE) between the atmosphere and vegetation is of great importance for regional and global studies of carbon balance. The eddy covariance technique can quantify carbon budgets and the effects of environmental controls for many forest types across the continent but it only provides integrated CO2 flux measurements within tower footprints and need to be scaled up to large areas in combination with remote sensing observations. In this study we compare a multiple-linear regression (MR) model which relates enhanced vegetation index and land surface temperature derived from the moderate resolution imaging spectroradiometer (MODIS), and photosynthetically active radiation with the site-level NEE, for estimating carbon flux exchange between the ecosystem and the environment at the deciduous-dominated Harvard Forest to three other methods proposed in the literature. Six years (2001–2006) of eddy covariance and MODIS data are used and results show that the MR model has the best performance for both training (2001–2004, R 2 = 0.84, RMSE = 1.33 g Cm−2 day−1) and validation (2005–2006, R 2 = 0.76, RMSE = 1.54 g Cm−2 day−1) datasets comparing to the other ones. It provides the potential to estimate carbon flux exchange across different ecosystems at various time intervals for scaling up plot-level NEE of CO2 to large spatial areas.
Quantification of the net carbon exchange between atmosphere and terrestrial ecosystem in global carbon cycle is becoming important with future potential sequestration influenced by increased atmospheric CO2 and changing climate (Nemani et al. 2003). Therefore, accurately estimating the net ecosystem carbon exchange (NEE), which is the difference between photosynthetic uptake and release of CO2 by respiration from autotrophs (vegetation) and heterotrophs (free living fauna in the soil and symbiotic microorganisms), at the regional, continental or global scale, is helpful to improve our understanding of the feedbacks between terrestrial biosphere and atmosphere in the context of global change and facilitate climate policy-making (Canadell et al. 2000; Xiao et al. 2010; Tang et al. 2011; Hu et al. 2014).
Traditionally, inventory studies of biomass and soil carbon were used to quantify an ecosystem NEE over a specific period (Clark et al. 2001). In recent years, the development of eddy covariance technique provides an alternative approach to continuously measure long term carbon exchange at ecosystem scales and evaluating carbon balance as well as its seasonal or annual variations more precisely has become possible (Baldocchi et al. 2001). Carbon budgets and the effects of environmental controls have been quantified with this technique for many forest types across the continent (Powell et al. 2006; Crawford and Christen 2014). However, the EC technique only provide integrated CO2 flux measurements over tower footprints with sizes and shape that vary with tower height, canopy physical characteristics and wind velocity (Osmond et al. 2004). Scaling up beyond the tower footprint to large areas is critically important in the quantification of net CO2 exchange over regions or continents (Gitelson et al. 2006, 2012; Xiao et al. 2010). Satellite remote sensing provides ecosystem observations with temporally and spatially coverage, and is an attractive and powerful tool for up-scaling carbon fluxes. A number of remote sensing based ecosystem carbon exchange models have been proposed recently to extend the role of field plots to capture regional variation and to bridge a major gap between field and satellite observations (Gregory et al. 2010). For example, Gamon et al. (1997) propose the photochemical reflectance index (PRI) that can correlate with light use efficiency (LUE) for carbon exchange estimation at leaf, canopy, stand and landscape levels (Gamon et al. 1997, 2001; Rahman et al. 2001, 2005). Vegetation indices (VI) such as NDVI and the enhanced vegetation index (EVI) are also used to directly estimate carbon fluxes (Xiao et al. 2004; Sims et al. 2006; Wu et al. 2010, 2012). Gitelson et al. (2006) first introduce the greenness and radiation (GR) model utilizing the total chlorophyll vegetation index and photosynthetically active radiation (PAR) to estimate carbon fixation in crops with high accuracy. The temperature and greenness (TG) model developed by Sims et al. (2008) that based on the MODIS EVI and land surface temperature (LST) product is validated in a wide diversity of natural vegetation including both deciduous and evergreen forests across North America. These studies demonstrate that greenness indices like enhanced vegetation indices (EVI) and land surface water index (LSWI), land surface temperature (LST), photosynthetically active radiation (PAR) are reliable proxies indicating plant phenological stages, canopy stresses (air temperature, soil moisture, vapor pressure deficit) and environmental conditions (incoming solar radiation) in estimation of carbon uptake by terrestrial ecosystems referred to as gross ecosystem exchange (GEE), but the ability of these biophysical indices in capturing the net carbon uptake by forest ecosystems namely NEE is less well known. Therefore, the objectives of this study are: (1) to analyze the potential of EVI, LSWI, LST, and PAR in tracking NEE seasonal dynamics, (2) on the basis of previous studies, to compare a newly proposed MR model with other models for NEE estimation in the Harvard deciduous broadleaf forest by selectively incorporating these proxies. This study will explore the implication and ability of eddy covariance and remote sensing observations for quantifying net carbon exchange between the atmosphere and forest ecosystems.
Eddy covariance data
The direct measurement of long-term carbon fluxes by eddy covariance technique provides the possibility of estimating local carbon sequestration rates of forests and different land use types. EC also improves our understanding of the vulnerability of ecosystem carbon balance to climate changes. Furthermore, it can help to evaluate ecosystem models and provide data for land surface exchange schemes in global models (Valentini et al. 2000). Eddy flux measurements of CO2, H2O and energy at Harvard Forest have been collected since 1991. The 6-year measurements (2001–2006) of NEE of CO2, daily PAR data are provided by researchers at Harvard Forest (http://public.ornl.gov/ameriflux/). Site-specific procedures, including quality control, flux corrections, and data editing are described elsewhere (Barford et al. 2001; Urbanski et al. 2007). The level 4 product consists of NEE data with four different time steps, including half-hourly, daily, weekly (8-day) and monthly. We utilize the weekly NEE data and the sums of PAR calculated over 8-day periods to match the compositing intervals of MODIS. The average of NEE and PAR over such a period was shown to largely eliminate micrometeorological errors caused by variable weather conditions or sampling procedures, with the remaining variability representing variation in ecosystem attributes (Oren et al. 2006).
The MODIS is a key instrument aboard the Terra and Aqua satellites, acquiring data in 36 spectral bands from 450 to 2100 nm. Two collection 5 MODIS products the 8-day land surface reflectance (MOD09A1) and land surface temperature (MOD11A2) from 2001 to 2006 are obtained from the 7 × 7 km subsets of MODIS products available at Oak Ridge National Laboratory’s Distributed Active Archive Center web site (http://www.modis.ornl.gov.modis.index.cfm). Average values of the central 3 × 3 km are extracted within the 7 × 7 km cutouts to better represent the flux tower footprint (Rahman et al. 2005).
The MODIS 8-day land surface temperature and Emissivity products (MOD11A2) in present works are retrieved at 1 km pixels by the generalized split-window algorithm and at 6 km grids by the day/night algorithm. In the split-window algorithm, emissivity in bands 31 and 32 are estimated from land cover types, atmospheric column water vapor and lower boundary air surface temperature are separated into tractable sub-ranges for optimal retrieval. MODIS LST is a measure of soil or canopy leaf temperature at the surface, which agreed with in situ measured LST within 1 K in the 263–322 K (Wan et al. 2002). Several researches have demonstrated that the satellite-derived LST also has a strong correlation with Re (Rahman et al. 2005; Schubert et al. 2010).
Carbon flux models
Our research proposes a predictive model incorporating MODIS and ground level data to estimate NEE. The derived MODIS EVI, LSWI, LST and in situ measured PAR are utilized to develop the new model. All the site level data are split into two sets: the training set (2001–2004) that containing 184 points and the test set (2005–2006) that containing 92 points, respectively. The optimum multi-linear regression (MR) model with the maximum determination coefficient (R 2) and the minimal root mean square error (RMSE) is subsequently generated after analyzing the relationships between these proxies and NEE and the MR model is shown to substantially have the best performance while comparing to the previous ones.
Results and discussion
Correlation and seasonal variation of NEE with PAR, VI and LST
Correlations of NEE and ecological proxies PAR, VIs, LST
It could also be seen from Fig. 2 that the MODIS derived EVI had a strong seasonal variation and reflected well the growth status of the deciduous broadleaf forest. EVI successfully captured the beginning and ending of the vegetation growing phase in 2001–2006. It firstly had a significant increase on WOY 15, reached the maximum value during 25–28 and then subsequently gradually declined to remain at a low level after 39. For all phenological indicators, the spring phenology is thought to exert a major effect on carbon balance. Earlier spring onset could consistently although not always significantly lead to higher R e and GEE for seasonal and annual flux integrals. In response to this, the less increased R e comparing to GEE would cause an increase NEE of springtime and this phenomenon was more obvious in 2004 than the other years by a magnitude of 2–4 g Cm−2 day−1.
Estimating NEE from incorporated MODIS and eddy covariance data
Although the proposed model has a better performance than the others in tracking seasonal dynamics of ecosystem carbon exchange for the deciduous-dominated Harvard Forest, it still contains significant uncertainties in NEE estimation.
To match the compositing intervals of MODIS data, weekly NEE data are adopted in our research because seasonal fluctuations of CO2 exchange could be well reflected over such a period. However, the magnitudes of carbon sources and sinks fluctuate remarkably on longer timescales such as annual due to geographical location, climate variation, land use change, disturbance by fire and pests, and age distribution as well as species composition of the ecosystem (Johnson et al. 2007). Therefore, the reliability of the MR model needs to be further validated across multiple biomes and over various time scales.
The proxies we selected here to estimate NEE in this study only includes PAR, EVI, and LST, it provides the opportunity to assess CO2 exchange directly from remote sensing observations since MODIS already produces EVI and LST products and remote estimation of PAR from MODIS aerosol type and atmospheric conditions would further make the model attractive for operational applications based on entirely remote sensing data (Liang et al. 2006). However, other factors that also have significant influences on CO2 exchange are not fully evaluated. NEE is the sum of canopy photosynthesis (GEE) and ecosystem respiration (R e ), and GEE is related with sunlight, temperature, ambient humidity, canopy growth and nutrient status while R e has been identified to have connections with soil moisture, air temperature, nutrient availability, stocks of living and dead biomass, seasonal carbon allocation as well as ecosystem productivity (Boone et al. 1998). Therefore, further research may focus on the underlying mechanism of ecophysiological interactions between the ecosystem and the environment variables.
Accurate estimation of terrestrial carbon exchange is of great importance for regional and global studies of carbon balance. We compare a multiple-linear regression model which relates vegetation indices and several environmental factors with the site-level NEE representing carbon flux exchange, with other models proposed in the literature in this mixed temperate forest. Results show that MR model could track seasonal fluctuations of ecosystem carbon exchange better than the others at the deciduous-dominated Harvard Forest site level. The discrepancies between model simulated NEE and measured NEE may be attributed with different dynamical ranges in EVI and the relatively importance of various environmental factors. Given the best performance in the accuracy of carbon flux exchange estimation by the model developed here, it is worthwhile to evaluate the efficacy of this method across different ecosystems at various time intervals for scaling up plot-level NEE of CO2 to large spatial areas.
This work was carried out in collaboration with all authors. YW designed the study and wrote the protocol, interpreted the data. XT proposed suggestions and relevant method analysis for the paper. JWM provides the valuable data and explanations. This word was supported by XH. All authors read and approved the final manuscript.
This work uses the data at the Harvard Forest site from the Ameriflux and we are grateful of PI for this site in providing the valuable data and the explanations. This study is supported by the Science and Technology Planning Project of Jiangxi Provincial Education Department No. GJJ151016 and the CAS strategic Priority Research Program Grant-Climate Change: Carbon Budget and Relevant Issues No. XDA05130703 and the Knowledge Innovation of the Chinese Academy of Sciences No. KZCX2-YW-224.
The authors declare that they have no competing interests.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
- Baldocchi D, Falge E, Gu L (2001) FLUXNET: a new tool to study the temporal and spatial variability of ecosystem scale carbon dioxide, water vapor, and energy flux densities. Bull Am Meteorol Soc 82:2415–2434View ArticleGoogle Scholar
- Barford CC, Wofsy SC, Goulden ML, Munger JW (2001) Factors controlling long and short term sequestration of atmospheric CO2 in a mid-latitude forest. Science 294:1688–1691View ArticleGoogle Scholar
- Boone RD, Nadelhoffer KJ, Canary JD, Kaye JP (1998) Roots exert a strong influence on the temperature sensitivity of soil respiration. Nature 396:570–572View ArticleGoogle Scholar
- Canadell JG, Mooney HA, Baldocchi DD, Berry JA, Ehleringer B, Field CB (2000) Carbon metabolism of the terrestrial biosphere: a multi-technique approach for improved understanding. Ecosystems 3:115–130View ArticleGoogle Scholar
- Clark DA, Brown S, Kicklighter DW, Chambers JQ, Thomlinson JR, Ni J (2001) Measuring net primary production in forests: concepts and field methods. Ecol Appl 11:356–370View ArticleGoogle Scholar
- Crawford B, Christen A (2014) Spatial variability of carbon dioxide in the urban canopy layer and implications ofr flux measurements. Atmos Environ 98:308–322View ArticleGoogle Scholar
- Fensholt R, Sandholt I (2003) Derivation of a shortwave infrared water stress index from MODIS near and shortwave infrared data in a semiarid environment. Remote Sens Environ 87:111–121View ArticleGoogle Scholar
- Gamon JA, Serrano L, Surfus JS (1997) The photochemical reflectance index: an optical indicator of photosynthesis radiation use efficiency across species, functional types, and nutrient levels. Oecologia 112:492–501View ArticleGoogle Scholar
- Gamon JA, Field CB, Fredeen AL, Thayer S (2001) Assessing photosynthetic down regulation in sunflower stands with an optically based model. Photosynth Res 67:113–125View ArticleGoogle Scholar
- Gitelson AA, Viña A, Verma SB, Rundquist DC, Arkebauer TJ, Keydan G (2006) Relationship between gross primary production and chlorophyll content in crops: Implications for the synoptic monitoring of vegetation productivity. J Geophys Res 111:D08S11. doi:10.1029/2005JD006017 Google Scholar
- Gitelson AA, Peng Y, Masek JG, Rundquist DC, Verma SV, Suyker A (2012) Remote estimation of crop gross primary production with Landsat data. Remote Sens Environ 121:404–414View ArticleGoogle Scholar
- Gregory PA, George VNP, Joseph M, David EK, John KC, James J (2010) High resolution forest carbon stocks and emissions in the Amazon. PNAS 107:16738–16742View ArticleGoogle Scholar
- Hu QW, Wu Q, Yao B, Xu XL (2014) Ecosystem respiration and its components from a Carex meadow of Poyang Lake during the drawdown period. Atmos Environ 100:124–132View ArticleGoogle Scholar
- Huete A, Didan K, Miura T, Rodriguez EP, Gao X, Ferreira LG (2002) Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens Environ 83:195–213View ArticleGoogle Scholar
- Jackson TJ, Chen D, Cosh M, Li F, Anderson M, Walthall C, Doriaswamy P, Hunt ER (2004) Vegetation water content mapping using Landsat data derived normalized difference water index for corn and soybeans. Remote Sens Environ 92:475–482View ArticleGoogle Scholar
- Johnson D, Murphy JD, Walker RF, Glass DW, Miller WW (2007) Wildfire effects on forest carbon and nutrient budgets. Ecol Eng 31:183–192View ArticleGoogle Scholar
- Liang S, Zheng T, Liu R, Fang H, Tsay SC, Running S (2006) Estimation of incident photosynthetically active radiation from Moderate Resolution Imaging Spectrometer data. J Geophys Res 111:D15208. doi:10.1029/2005JD006730 View ArticleGoogle Scholar
- Nemani RR, Keeling CD, Hashimoto H, Jolly WM, Piper SC, Tucker CJ (2003) Climate-driven increases in global terrestrial net primary production from 1982 to 1999. Science 300:1560–1563View ArticleGoogle Scholar
- Oren R, Hsieh CL, Stoy P, Albertson J, Rmccarthy H, Harrell P, Katul G (2006) Estimating the uncertainty in annual net ecosystem carbon exchange: spatial variation in turbulent fluxes and sampling errors in eddy-covariance measurements. Glob Change Biol 12:883–896View ArticleGoogle Scholar
- Osmond B, Ananyev G, Berry J, Langdon C, Kolber Z, Lin GH (2004) Changing the way we think about global change research: scaling up in experimental ecosystem science. Glob Change Biol 10:393–407View ArticleGoogle Scholar
- Powell TL, Bracho R, Li JH, Dore S, Hinkle CR, Drake BG (2006) Environmental controls over net ecosystem carbon exchange of scrub oak in central Florida. Agric For Meteorol 141:19–34View ArticleGoogle Scholar
- Rahman AF, Gamon JA, Fuentes DA (2001) Modeling spatially distributed ecosystem flux of boreal forest using hyperspectral indices from AVIRIS imagery. J Geophys Res 106:33579–33591View ArticleGoogle Scholar
- Rahman AF, Sims DA, Cordova VD, El-masri BZ (2005) Potential of MODIS EVI and surface temperature for directly estimating per-pixel ecosystem C fluxes. Geophys Res Lett 32:L19404.1–L19404.6View ArticleGoogle Scholar
- Schubert P, Eklundeh L, Lund M, Nilsson M (2010) Estimating northern peatland CO2 exchange from MODIS time series data. Remote Sens Environ 114:1178–1189View ArticleGoogle Scholar
- Sims DA, Rahman AF, Cordova VD, El-Masri BZ, Baldocchi DD, Flanagan LB (2006) On the use of MODIS EVI to assess gross primary productivity of North American ecosystems. J Geophys Res 111:G04015. doi:10.1029/2006JG000162 Google Scholar
- Sims DA, Rahman AF, Vicente DC (2008) A new model of gross primary productivity for North American ecosystems based solely on the enhanced vegetation index and land surface temperature from MODIS. Remote Sens Environ 112:1633–1646View ArticleGoogle Scholar
- Tang XG, Liu DW, Song KS, Munger JW, Zhang B, Wang ZM (2011) A new model of net ecosystem carbon exchange for the deciduous-dominated forest by integrating MODIS and flux data. Ecol Eng 37:1567–1571View ArticleGoogle Scholar
- Urbanski S, Barford C, Wofsy S, Kucharik C, Pyle E, Budney J, Mckain K (2007) Factors controlling CO2 exchange on time scales from hourly to decadal at Harvard Forest. J Geophys Res 112:1–25Google Scholar
- Valentini R, Matteucci G, Dolman AJ (2000) Respiration as the main determinant of carbon balance in European forests. Nature 404:861–865View ArticleGoogle Scholar
- Wan ZM, Zhang YL, Zhang QC, Li ZL (2002) Validation of the land surface temperature products retrieved from Terra Moderate Resolution Imaging Spectroradiometer data. Remote Sens Environ 83:163–180View ArticleGoogle Scholar
- Wu C, Munger JW, Niu Z, Kuang D (2010) Comparison of multiple models for estimating gross primary production using MODIS and eddy covariance data in Harvard Forest. Remote Sens Environ 114:2925–2939View ArticleGoogle Scholar
- Wu C, Chen JM, Desai AR, Hollinger DY, Arain MA, Margolis HA (2012) Remote sensing of canopy light use efficiency in temperate and boreal forests of North America using MODIS imagery. Remote Sens Environ 118:60–72View ArticleGoogle Scholar
- Xiao X, Boles S, Frolking S, Salas W, Moore B, Li C (2002) Observation of flooding and rice transplanting of paddy rice fields at the site to landscape scales in China using VEGETATION sensor data. Int J Remote Sens 23:3009–3022View ArticleGoogle Scholar
- Xiao X, Braswell B, Zhang Q, Boles S, Frolking S, Moore B (2003) Sensitivity of vegetation indices to atmospheric aerosols: continental-scale observations in Northern Asia. Remote Sens Environ 84:385–392View ArticleGoogle Scholar
- Xiao X, Zhang Q, Braswell B, Urbanski S, Boles S, Wofsy S (2004) Modeling gross primary production of temperate deciduous broadleaf forest using satellite images and climate data. Remote Sens Environ 91:256–270View ArticleGoogle Scholar
- Xiao J, Zhuang Q, Law BE, Chen J, Baldocchi DD, Cook DR (2010) A continuous measure of gross primary production for the conterminous U.S. derived from MODIS and AmeriFlux data. Remote Sens Environ 114:576–591View ArticleGoogle Scholar