J Indian Soc Remote Sens DOI 10.1007/s12524-016-0564-7
RESEARCH ARTICLE
Estimating Gross Primary Production of a Forest Plantation Area Using Eddy Covariance Data and Satellite Imagery Joyson Ahongshangbam 1 & N. R. Patel 1 & S. P. S. Kushwaha 1 & Taibanganba Watham 1 & V. K. Dadhwal 2
Received: 1 May 2015 / Accepted: 27 January 2016 # Indian Society of Remote Sensing 2016
Abstract Gross primary production (GPP) is the basic biophysical parameter of an ecosystem. The quantification of GPP has been a major challenge in understanding the global carbon cycle. Eddy covariance (EC) measurements at flux tower provide valuable direct information on seasonal dynamics of GPP and allow model optimization. In this paper, the GPP of forest plantation was estimated using light use efficiency (LUE-based) model and validated with flux tower GPP observations in Terai Central Forest Division, Nainital, India. The LUE model is mainly based upon the photosynthetically active radiation (PAR), satellite-derived normalized difference vegetation index (NDVI), land surface wetness index (LSWI), and the air temperature. The simulation of the model was carried out using vegetation indices generated from Landsat imagery and the meteorological data from flux tower. The predicted GPP showed distinct significance of spatiotemporal dynamics of GPP. The environmental variables,
* Joyson Ahongshangbam
[email protected] N. R. Patel
[email protected] S. P. S. Kushwaha
[email protected] Taibanganba Watham
[email protected] V. K. Dadhwal
[email protected]
1
Indian Institute of Remote Sensing, ISRO, 4-Kalidas Road, Dehradun 248001, Uttarakhand, India
2
National Remote Sensing Centre (ISRO), Balanagar, Hyderabad 500037, Andhra Pradesh, India
viz., PAR and NDVI showed distinct effect on the GPP prediction. Comparison between predicted and the measured GPP on flux tower site showed good agreement (R2 = 0.626, RMSE = 2.08 and MAPE = 18.46). The study demonstrated the potential of LUE model for estimating GPP and scaling up of GPP over large areas, which is a major parameter in the study of the carbon cycle on regional to global scales. Keywords Eddy covariance . GPP . LUE . NDVI . PAR
Introduction The carbon sequestration by terrestrial biomes plays a significant role in global carbon cycle under increased atmospheric CO2 and changing climate (Wofsy et al. 1993; Nemani et al. 2003, 2002). Gross primary production (GPP) is an important biophysical parameter of any ecosystem that plays a key role in the spatio-temporal dynamics of CO2. The prediction of the magnitude of GPP has been a major challenge in quantifying global carbon cycle (Canadell et al. 2000). It is also one of the important questions faced by ecologists and global climate change experts in quantifying the net terrestrial carbon uptake and its spatio-temporal variation (Barford et al. 2001; IPCC 2001). The uncertainty in understanding net carbon uptake at local, regional and global scales can be overcome by monitoring the earth surface processes on high spatial and temporal resolutions (Zhao et al. 2005). Continuous and direct CO2 flux measurements using eddy covariance (EC) technique and tall towers facilitate a detailed study of the global carbon cycle (Wofsy et al. 1993; Falge et al. 2002a, b). A study conducted in Harvard forest shows that the seasonal and inter-annual variations of net carbon exchange was regulated by light, temperature and the moisture (Barford et al. 2001). The CO2 flux data can also be used to
J Indian Soc Remote Sens
optimize and improve the process-based ecosystem models (Law et al. 2000). Eddy covariance flux towers provide integrated flux measurements over the footprint area ranging from few metres to hectares depending upon the tower height, canopy physical characteristics, and the wind velocity. It is an important yet challenging task to scale-up the CO2 fluxes from footprint area to large areas due to high spatial heterogeneity and temporal dynamics of the forest ecosystem. Many approaches to up-scale and extrapolate the flux tower site measurements have been developed. One common approach is to use process-based biogeochemical models driven by climate, soil and vegetation layers (Law et al. 2000; Churkina et al. 2003). Use of satellite observations and climate data has been widely adopted for upscaling (Turner et al. 2003; Xiao et al. 2004). Satellite imagery provides synoptic and consistent observations of vegetation and thus, helps in characterization of the vegetation structure (Potter et al. 1993; Prince and Goward 1995). Normalized difference vegetation index (NDVI) is one of the common indices to decipher the vegetation characteristics with strong linear relationship with the fraction of absorbed PAR (fAPAR) (Myneni and Williams 1994). A number of satellite-based modelling techniques have been used to estimate GPP or NPP at large spatial scales in India (Nayak et al. 2010, 2013) and abroad (Ruimy and Saugier 1994, 1999; Running et al. 1999, 2000). Various remote sensing driven modeling approaches viz., temperature-greenness, light-use efficiency (LUE) and regional scale CASA (Carnegie–Ames– Stanford Approach) models have been tested to quantify primary production over terrestrial ecosystems in India (Nayak et al. 2010; Patel et al. 2010, 2011). Although remote sensing-based LUE approach has been extensively used worldwide, the accuracy (and the performance) of these models largely depends on parameterization of ecosystem-specific parameters such as maximum LUE (ε*) and the temperature optima (Topt) of gross photosynthesis. EC technique is now becoming popular for deriving these ecosystem parameters as an input to LUE models. EC towers, established in India under National Carbon Project (NCP) of ISRO Geosphere Biosphere Programme, have opened new avenues for monitoring carbon fluxes over terrestrial ecosystems. The LUE model is based upon two fundamental assumptions: (i) the ecosystem GPP is directly related to absorbed photosynthetically active radiation (APAR) and (ii) the down-regulated maximum LUE of particular vegetation type by environmental stressors such as non-optimal temperature or water stress (Running et al. 2004; Landsberg 1986). In view of this, a present study was undertaken to estimate high resolution monthly gross primary productivity of a forest plantation using LUE model being parameterized and driven by flux-tower measurements and monthly Landsat TM satellite data.
Materials and Methods Study Area The study was conducted in a 45.48 km2 area of Terai Central Forest Division, located in Nainital district of the Uttarakhand state in India. The study area consists mainly of mixed forest plantation comprising Tectona grandis, Eucalyptus hybrid and Populus deltoides tree species and managed by State Forest Department (Fig. 1). The terrain is flat with