Global Change Biology Semi-empirical modeling of abiotic and biotic factors controlling ecosystem respiration across eddy covariance sites F o Journal: Global Change Biology r Manuscript ID: GCB-10-0015 Wiley - Manuscript type: Primary Research Articles R Date Submitted by the 07-Jane-2010 Author: Complete List of Authors: Migliavaccva, Mirco; University of Milano-Bicocca, Remote Sensing of Environmentail Dynamics; Max Planck Institute for Biogeochemistry, Model Data Integration Group Reichstein, Marekus; Max Planck Institute for Biogeochemistry, Model Data Integration Group Richardson, Andrew; Harvard University, Department of Organismic w and Evolutionary Biology Colombo, Roberto; University of Milano-Bicocca, Remote Sensing of Environmental Dynamics Sutton, Mark A.; Centre for Ecology and Hydrology, Edinburgh O Research Station Lasslop, Gitta; Max Planck Institute for Biogeochemistry, Model n Data Integration Group Wohlfahrt, Georg; University of Innslbruck, Institute of Ecology Tomelleri, Enrico; Max Planck Institute for Biogeochemistry, Model Data Integration Group y Carvalhais, Nuno; Universidade Nova de Lisboa, Faculdade de Ciências e Tecnologia; Max Planck Institute for Biogeochemistry, Model Data Integration Group Cescatti, Alessandro; European Commission, DG-JRC, Institute for Environment and Sustainability Mahecha, Miguel; Max Planck Institute for Biogeochemistry, Model Data Integration Group; Swiss Federal Institute of Technology-, Department of Environmental Sciences Montagnani, Leonardo; Provincia Autonoma di Bolzano, Agenzia per l'Ambiente, Servizi Forestali Papale, Dario; University of Tuscia, DISAFRI Zaehle, Sönke; Max Planck Institute for Biogeochemistry, Department for Biogeochemical System Arain, M Altaf; McMaster University, School of Geography & Earth Sciences Arneth, Almut; Lund University, 13- Department of Physical Geography and Ecosystems Analysis Page 1 of 62 Global Change Biology 1 2 3 4 Black, T Andrew; University of British Columbia, Faculty of Land and Food Systems 5 Dore, Sabina; Northern Arizona University, School of Forestry 6 Gianelle, Damiano; Fondazione Edmund Mach, Centro di Ecologia 7 Alpina 8 Helfter, Carole; Centre for Ecology and Hydrology, Edinburgh 9 Research Station 10 Hollinger, David; USDA Forest Service, NE Research Station 11 Kutsch, Werner; Johann Heinrich von Thünen Institut, Institut für 12 Agrarrelevante Klimaforschung Law, Beverly; Oregon State University, College of Forestry 13 Lafleur, Peter M; Trent University, 20- Department of Geography 14 Nouvellon, Yann; CIRAD, Persyst 15 Rebmann, Corinna; Max-Planck Institute for Biogeochemistry, 16 Biogeochemical Processes; University of Bayreuth, Department of 17 Micrometeorology 18 da Rocha, Humberto; Universidade de São Paulo, Dept. of 19 FAtmospheric Sciences 20 Rodeghiero, Mirco; Fondazione Edmund Mach, Centro di Ecologia oAlpina 21 Olivier, Roupsard; CIRAD, Persyst; Centro Agronómico Tropical de 22 r Investigación y Enseñanza, CATIE 23 Sebastià, Maria-Teresa; University of Lleida, Agronomical 24 Engineering School; Forest Technology Centre of Catalonia, 25 LRaboratory of Plant Ecology and Botany 26 Seufert, Guenther; Institute for Environment and Sustainability, 27 Europeean Commission, DG-JRC Soussana, Jean-Francoise; Institut National de la Recherche 28 Agronomivque 29 van der Moleni, Michiel K; University de Boeleaan, Department of 30 Hydrology and Geo-Environmental Sciences 31 e 32 Ecosystem Respiration, Productivity, FLUXNET, Eddy Covariance, Keywords: 33 Leaf Area Index, Inverse Modeling 34 w In this study we examine d ecosystem respiration (RECO) data from 35 104 sites belonging to FLUXNET, the global network of eddy 36 covariance flux measurements. The main goal was to identify the 37 main factors involved in thOe variability of RECO: temporally and 38 between sites as affected by climate, vegetation structure and plant 39 functional type (PFT) (evergreenn needleleaf, grasslands, etc.). 40 We demonstrated that a model usingl only climate drivers as 41 predictors of RECO failed to describe part of the temporal variability 42 in the data and that the dependency oyn gross primary production (GPP) needed to be included as an additional driver of RECO. The 43 maximum seasonal leaf area index (LAIMAX) had an additional 44 effect that explained the spatial variability of reference respiration 45 (the respiration at reference temperature Tref=15°C, without Abstract: 46 stimulation introduced by photosynthetic activity and without water 47 limitations), with a statistically significant linear relationship 48 (r2=0.52 p<0.001, n=104) even within each PFT. Besides LAIMAX, 49 we found that the reference respiration may be explained partially by total soil carbon content. For undisturbed temperate and boreal 50 forest a negative control of the total nitrogen deposition on the 51 reference respiration was also identified. 52 We developed a new semi-empirical model incorporating abiotic 53 factors (climate), recent productivity (daily GPP), general site 54 productivity and canopy structure (LAIMAX) which performed well 55 in predicting the spatio-temporal variability of RECO, explaining 56 >70% of the variance for most vegetation types. Exceptions include tropical and Mediterranean broadleaf forests and deciduous 57 58 59 60 Global Change Biology Page 2 of 62 1 2 3 4 broadleaf forests. Part of the variability in respiration that could not be described by our model could be attributed to a range of factors, 5 including phenology in deciduous broadleaf forests and 6 management practices in grasslands and croplands. 7 8 9 10 11 12 13 14 15 16 17 18 F 19 20 o 21 22 r 23 24 25 R 26 27 e 28 v 29 i 30 31 e 32 33 34 w 35 36 37 O 38 39 n 40 l 41 42 y 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 3 of 62 Global Change Biology 1 2 3 1 Semi-empirical modeling of abiotic and biotic factors controlling ecosystem 4 5 2 respiration across eddy covariance sites 6 3 7 4 Mirco Migliavacca1,2, Markus Reichstein2, Andrew D. Richardson3, Roberto Colombo1, Mark A. 8 5 Sutton4, Gitta Lasslop2, Georg Wohlfahrt5, Enrico Tomelleri2, Nuno Carvalhais6,2, Alessandro 9 6 Cescatti7, Miguel D. Mahecha2,8, Leonardo Montagnani9, Dario Papale10 , Sönke Zaehle 11, Altaf 10 11 7 Arain12, Almut Arneth13, T. Andrew Black14, Sabina Dore15, Damiano Gianelle16, Carole Helfter4, 12 8 David Hollinger17, Werner L. Kutsch18, Beverly E. Law19, Peter M. Lafleur20, Yann Nouvellon21, 13 9 Corinna Rebmann22,23, Humberto Ribeiro da Rocha24, Mirco Rodeghiero16, Olivier Roupsard21,25, 14 10 Maria-Teresa Sebastià26,27, Guenther Seufert7, Jean-Francoise Soussana28, Michiel K. van der 15 11 Molen29 16 17 12 18 13 1- Remote Sensing of Environmental Dynamics Laboratory, DISAT, University of Milano- 19 14 Bicocca, Milano, Italy. 20 15 2- Model Data InteFgration Group, Max Planck Institute for Biogeochemistry, Jena, Germany. 21 16 3- Department of Organismic and Evolutionary Biology, Harvard University, Cambridge MA, 22 23 17 USA. o 24 18 4- Centre for Ecology andr Hydrology, Edinburgh Research Station, Bush Estate, Penicuik, 25 19 Midlothian, Scotland, EH 26 0QB, UK. 26 20 5- Institut für Ökologie, Universität Innsbruck, Innsbruck, Austria. 27 21 6- Faculdade de Ciências e TeRcnologia, FCT, Universidade Nova de Lisboa, 2829-516, 28 22 Caparica, Portugal. 29 e 30 23 7- European Commission, DG-JRC, Institute for Environment and Sustainability, Climate 31 24 Change Unit, Via Enrico Fermi 27v49, T.P. 050, 21027 Ispra (VA), Italy. 32 25 8- Department of Environmental Scienceis, Swiss Federal Institute of Technology-ETH Zurich, 33 26 8092 Zurich, Switzerland. e 34 27 9- Agenzia Provinciale per l'Ambiente, Via Amba-Alagi 5, 39100 Bolzano, Italy. 35 36 28 10- DISAFRI, University of Tuscia, via C. de Lwellis, 01100 Viterbo Italy. 37 29 11- Department for Biogeochemical System, Max P lanck Institute for Biogeochemistry, Jena, 38 30 Germany. 39 31 12- McMaster University, School of Geography & Earth Sciences, 1280 Main Street West, O 40 32 Hamilton, ON, L8S 4K1, Canada. 41 42 33 13- Department of Physical Geography and Ecosystems Annalysis, Lund University, Sölvegatan 43 34 12, SE-223 62 Lund, Sweden. l 44 35 14- Faculty of Land and Food Systems, University of British Columbia, Vancouver, BC, y 45 36 Canada. 46 37 15- Department of Biological Sciences and Merriam-Powell Center for Environmental 47 38 Research, Northern Arizona University, Flagstaff, Arizona, USA. 48 49 39 16- IASMA Research and Innovation Centre, Fondazione E. Mach, Environment and Natural 50 40 Resources Area, San Michele all’Adige, I-38040 Trento, Italy 51 41 17- USDA Forest Service, NE Research Station, Durham, NH, USA 52 42 18- Johann Heinrich von Thünen Institut (vTI), Institut für Agrarrelevante Klimaforschung, 53 43 Braunschweig, Germany 54 44 19- College of Forestry, Oregon State University, 97331-5752 Corvallis, OR, USA 55 56 45 20- Department of Geography, Trent University, Peterborough, ON K 9J 7B8, Canada. 57 46 21- CIRAD, Persyst, UPR80, TA10/D, 34398 Montpellier Cedex 5, France. 58 47 22- University of Bayreuth, Department of Micrometeorology, Bayreuth, Germany. 59 48 23- Max Planck Institute for Biogeochemistry, Jena, Germany. 60 49 24- Departamento de Ciências Atmosféricas/IAG/Universidade de São Paulo, Rua do Matão, 50 1226 - Cidade Universitária - São Paulo, SP - Brasil. 1 Global Change Biology Page 4 of 62 1 2 3 51 25- CATIE, Centro Agronómico Tropical de Investigación y Enseñanza, Turrialba Costa Rica. 4 52 26- Laboratory of Plant Ecology and Botany. Forest Technology Centre of Catalonia, Solsona, 5 53 Spain. 6 7 54 27- Agronomical Engineering School, University of Lleida, E-25198 Lleida, Spain. 8 55 28- INRA, Institut National de la Recherche Agronomique, Paris, France. 9 56 29- Department of Hydrology and Geo-Environmental Sciences, VU-University, de Boeleaan 10 57 1085, 1081 HV Amsterdam, The Netherlands. 11 58 12 13 59 Corresponding author: 14 60 Mirco Migliavacca 15 61 Remote Sensing of Environmental Dynamics 16 62 Laboratory, DISAT, University of Milano-Bicocca, P.zza della Scienza 1, 20126 17 63 Milan, Italy. Tel.: +39 0264482848; fax: +39 0264482895. 18 19 64 E-mail address: [email protected] 20 65 F 21 22 o 23 24 r 25 26 27 R 28 29 e 30 31 v 32 i 33 e 34 35 36 w 37 38 39 O 40 41 42 n 43 l 44 y 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 2 Page 5 of 62 Global Change Biology 1 2 3 66 4 5 67 Abstract 6 68 7 8 69 In this study we examined ecosystem respiration (R ) data from 104 sites belonging to 9 ECO 10 70 FLUXNET, the global network of eddy covariance flux measurements. The main goal was to 11 12 71 identify the main factors involved in the variability of R : temporally and between sites as ECO 13 14 72 affected by climate, vegetation structure and plant functional type (PFT) (evergreen needleleaf, 15 73 grasslands, etc.). 16 17 74 We demonstrated that a model using only climate drivers as predictors of R failed to ECO 18 19 75 describe part of the temporal variability in the data and that the dependency on gross primary 20 F 21 76 production (GPP) needed to be included as an additional driver of R . The maximum seasonal ECO 22 77 leaf area index (LAI o) had an additional effect that explained the spatial variability of 23 MAX 24 r 78 reference respiration (the respiration at reference temperature Tref=15°C, without stimulation 25 26 79 introduced by photosynthetic activity and without water limitations), with a statistically 27 R 28 80 significant linear relationship (r2=0.52 p<0.001, n=104) even within each PFT. Besides LAI , MAX 29 e 30 81 we found that the reference respiration may be explained partially by total soil carbon content. 31 v 82 For undisturbed temperate and boreal forest a negative control of the total nitrogen deposition on 32 i 33 83 the reference respiration was also identified. e 34 35 84 We developed a new semi-empirical model incorporating abiotic factors (climate), recent 36 w 37 85 productivity (daily GPP), general site productiv ity and canopy structure (LAIMAX) which 38 86 performed well in predicting the spatio-temporal variability of R , explaining >70% of the 39 ECO O 40 87 variance for most vegetation types. Exceptions include tropical and Mediterranean broadleaf 41 42 88 forests and deciduous broadleaf forests. Part of the varianbility in respiration that could not be 43 l 44 89 described by our model could be attributed to a range of factors, including phenology in y 45 90 deciduous broadleaf forests and management practices in grasslands and croplands. 46 47 91 48 49 92 Keywords: Ecosystem Respiration, Productivity, FLUXNET, Eddy Covariance, Leaf Area 50 51 93 Index, Inverse Modeling 52 53 94 54 55 95 Introduction 56 96 57 58 97 Respiration of terrestrial ecosystems (R ) is one of the major fluxes in the global carbon cycle 59 ECO 60 98 and its responses to environmental change is important for understanding climate-carbon cycle 99 interactions (e.g. Cox et al., 2000, Houghton et al., 1998). It has been hypothesized that relatively 3 Global Change Biology Page 6 of 62 1 2 3 100 small climatic changes may impact respiration with the effect of rivalling the annual fossil fuel 4 5 101 loading of atmospheric CO (Jenkinson et al., 1991, Raich & Schlesinger, 1992). 2 6 7 102 Recently, efforts have been made to mechanistically understand how temperature and other 8 103 environmental factors affect ecosystem and soil respiration, and various modeling approaches have 9 10 104 been proposed (e.g. Davidson et al., 2006a, Lloyd & Taylor, 1994, Reichstein & Beer, 2008, 11 12105 Reichstein et al., 2003a). Nevertheless, the description of the conceptual processes and the complex 13 14106 interactions controlling R are still under intense research and this uncertainty is still hampering ECO 15 107 bottom-up scaling to larger spatial scales (e.g. regional and continental) which is one of the major 16 17 108 challenges for biogeochemists and climatologists. 18 19109 Heterotrophic and autotrophic respiration in both data-oriented and process-based 20 F 21110 biogeochemical models are usually described as a function of air or soil temperature and 22 23111 occasionally soil water conotent (e.g. Lloyd & Taylor, 1994, Reichstein et al., 2005, Thornton et al., 24 r 112 2002), although the functional form of these relationships varies from model to model. These 25 26 113 functions represent the dominant role of reaction kinetics, possibly modulated or confounded by 27 R 28114 other environmental factors such as soil water content or precipitation, which some model 29 e 30115 formulations include as a secondary effect (e.g. Carlyle & Ba Than, 1988, Reichstein et al., 2003a, 31 v 32116 Richardson et al., 2006). i 33 117 A large number of statistical, climate-driven models of ecosystem and soil respiration have been e 34 35118 tested and compared using data from individual sites (Del Grosso et al., 2005, Janssens & 36 w 37119 Pilegaard, 2003, Richardson & Hollinger, 2005, Sava ge et al., 2009), multiple sites (Falge et al., 38 120 2001, Rodeghiero & Cescatti, 2005), and from a wide range of models compared across different 39 O 40 121 ecosystem types and measurement techniques (Richardson et al., 2006). 41 42 n 122 Over the course of the last decades, the scientific community has debated the role of productivity 43 l 44123 in determining ecosystem and soil respiration. Several authors (Bahn et al., 2008, Curiel Yuste et y 45 46124 al., 2004, Davidson et al., 2006a, Janssens et al., 2001, Reichstein et al., 2003a, Valentini et al., 47 125 2000) have discussed and clarified the role of photosynthetic activity, vegetation productivity and 48 49 126 their relationship with respiration. 50 51127 Linking photosynthesis and respiration might be of particular relevance when modelling R ECO 52 53128 across biomes or at the global scale. Empirical evidence for the link between GPP and R is ECO 54 129 reported for most, if not all, ecosystems: grassland (e.g. Bahn et al., 2008, Bahn et al., 2009, Craine 55 56 130 et al., 1999, Hungate et al., 2002), crops (e.g. Kuzyakov & Cheng, 2001, Moyano et al., 2007), 57 58131 boreal forests (Gaumont-Guay et al., 2008, Hogberg et al., 2001) and temperate forests, both 59 60132 deciduous (e.g. Curiel-Yuste et al., 2004, Liu et al., 2006) and evergreen (e.g. Irvine et al., 2005). 4 Page 7 of 62 Global Change Biology 1 2 3 133 Moreover, several authors have found a time lag between productivity and respiration response. 4 5 134 This time lag depends to the vegetation structure it is related to the translocation time of assimilates 6 7 135 from aboveground to belowground organs through the phloem. Although the existence of a time lag 8 136 is still under debate, it has been found to be a few hours in grasslands, and croplands and a few 9 10 137 days in forests (Baldocchi et al., 2006, Knohl & Buchmann, 2005, Moyano et al., 2008, Savage et 11 12138 al., 2009). 13 14139 While the link between productivity and respiration appears to be clear, to our knowledge, few 15 140 model formulations include the effect of productivity or photosynthesis as a biotic driver of 16 17 141 respiration and these models are mainly developed for the simulation of soil respiration using a 18 19142 relatively small data set of soil respiration measurements (e.g. Hibbard et al., 2005, Reichstein et 20 F 21143 al., 2003a). 22 23144 In this context, the inocreasing availability of ecosystem carbon, water and energy flux 24 r 145 measurements collected by means of the eddy covariance technique (e.g. Baldocchi, 2008) over 25 26 146 different plant functional types (PFTs) at more than 400 research sites, represents an useful tool for 27 R 28147 understanding processes and interactions behind carbon fluxes and ecosystem respiration. These 29 e 30148 data serve as a backbone for bottom-up estimates of continental carbon balance components (e.g. 31 v 32149 Ciais et al., 2005, Papale & Valentini, 200i3, Reichstein et al., 2007) and for ecosystem model 33 150 development, calibration and validation (e.g. Baldocchi, 1997, Hanson et al., 2004, Law et al., e 34 35151 2000, Owen et al., 2007, Reichstein et al., 2003b, Reichstein et al., 2002, Verbeeck et al., 2006). 36 w 37152 The database includes a number of added products such as gap-filled net ecosystem exchange 38 153 (NEE), gross primary productivity (GPP), ecosystem respiration (R ) and meteorological drivers 39 ECO O 40 154 (air temperature, radiation, precipitation etc) aggregated at different time-scale (e.g. half-hourly, 41 42 n 155 daily, annual) and consistent for data treatment (Papale et al., 2006, Reichstei et al., 2005) 43 l 44156 In this paper we analyze with a semi-empirical modeling approach the R at 104 different sites yECO 45 46157 belonging to the FLUXNET database with the primary objective of synthesizing and identifying the 47 158 main factors controlling i) the temporal variability of R , ii) the between-site (spatial) variability 48 ECO 49 159 and iii) to provide a model which can be used for diagnostic up-scaling of R from eddy 50 ECO 51160 covariance flux sites to large spatial scales. 52 53161 Specifically, the analysis and the model development followed these two steps: 54 162 1. we developed a semi-empirical R model site by site (site-by-site analysis) with the aim of 55 ECO 56 163 clarifying if and how GPP should be included into a model for improving the description of 57 58164 R and which factors are best suited for describing the spatial variability of reference ECO 59 60165 respiration (i.e. the daily R at the reference temperature without moisture limitations). ECO 166 We follow these three steps: 5 Global Change Biology Page 8 of 62 1 2 3 167 o the analysis of R data was conducted by using a purely climate driven model: ‘TP ECO 4 5 168 Model’ (Raich et al., 2002). The accuracy of the model and the main bias were 6 7 169 analyzed and discussed; 8 170 o we evaluated the inclusion of biotic factors (i.e. GPP) as drivers of R . A range of 9 ECO 10 171 different model formulations, which differ mainly in regard to the functional 11 12172 responses of R to photosynthesis, were tested in order to identify the best model ECO 13 14173 formulation for the daily description of R at each site; ECO 15 174 o we analyzed variability of the reference respiration estimated at each site with the 16 17 175 aim of identifying, among the different site characteristics, one or more predictors of 18 19176 the spatial variability of this crucial parameter. This can be extremely useful for the 20 F 21177 application of the model at large spatial scale; 22 23178 2. we optimized the deoveloped model for each PFT (PFT analysis) with the aim of generalizing 24 r 179 the model parameters in a way that can be useful for diagnostic, PFT-based, up-scaling of 25 26 180 R . The accuracy of the model was assessed by a cross-validation technique and the main ECO 27 R 28181 weak points of model were critically evaluated and discussed. 29 e 30182 31 v 32183 Material and Methods i 33 184 e 34 35 185 Data set 36 w 37186 38 39187 The data used in this analysis is based on the dataset from the FLUXNET (www.fluxdata.org) O 40 41188 eddy covariance network (Baldocchi, 2008, Baldocchi et al., 2001). The analysis was restricted to 42 n 43189 104 sites (cf. Table in Appendix I and II) on the basis of the alncillary data availability (i.e. only 44 190 sites containing at least both leaf area index (LAI) of understorey and overstorey were selected) and y 45 46191 of the time series length (all sites containing at least one year of carbon fluxes and meteorological 47 48192 data of good quality data were used). Further, we only analyzed those sites for which the relative 49 193 standard error of the estimates of the model parameters E (activation energy) and reference 50 0 51 194 respiration (R ) (please see further sections for more details on the meaning of parameters) were 52 0 53195 less than 50% and where E estimates were within an acceptable range (0–450 K). 0 54 55 196 The latitude spans from 71.32° at the Alaska Barrow site (US-Brw) to -21.62° at the Sao Paulo 56 57 197 Cerrado (BR-Sp1). The climatic regions include tropical to arctic. 58 59 198 All the main PFTs as defined by the IGBP (International Geosphere-Biosphere Programme) 60 199 were included in this study: the selected sites included 28 evergreen needleleaf forests (ENF), 17 6 Page 9 of 62 Global Change Biology 1 2 3 200 deciduous broadleaf forests (DBF), 16 grasslands (GRA), 11 croplands (CRO), 8 mixed forests 4 5 201 (MF), 5 savannas (SAV), 9 shrublands (SHB), 7 evergreen broadleaved forests (EBF) and 3 6 7 202 wetlands (WET). Due to limited number of sites and their similarity, the class SAV included both 8 203 the sites classified as savanna (SAV) and woody savannas (WSA), while the class SHB included 9 10 204 both the open (OSH) and closed (CSH) shrubland sites. For abbreviations and symbols refer to 11 12205 Appendix III. 13 14 206 Daily R , GPP and the associated uncertainties of NEE data, together with daily 15 ECO 16207 meteorological data such as mean air temperature (T ) and 30-day precipitation running average A 17 18208 (P), were downloaded from the FLUXNET database. 19 209 At each site data are storage corrected, spike filtered, u -filtered according to Papale et al. (2006) 20 * F 21 210 and subsequently gap-filled and partitioned as described by Reichstein et al. (2005). Only days 22 o 23 211 containing both meteorological and daily flux data with a percentage of gap-filled half hours below 24 r 25212 15% were used for this analysis. The median of the u threshold applied in the FLUXNET database * 26 27213 for the site-years used in the analRysis are listed in the Appendix II. The average of the median u* 28 214 values are lower for short canopies (e.g. for grasslands 0.075±0.047 ms-1) and higher for tall 29 e 30215 canopies (e.g. for evergreen needleleaf forests 0.221 ±0.115 ms-1). 31 v 32216 Along with fluxes and meteorological daita, main ancillary data such as maximum ecosystem 33 34217 LAI (overstory and understory for forest sitese) (LAI ), LAI of overstory (LAI ), stand age MAX MAX,o 35 218 for forests (StandAge), total soil carbon stock (SoilC) and the main information about disturbance 36 w 37219 (date of cuts, harvesting) were also downloaded fro m the database. Total atmospheric nitrogen 38 39220 deposition (N ) is based on the atmospheric chemistry transport model TM3 (Rodhe et al., 2002) depo O 40 41221 and calculated at 1°x1° resolution. These data are grid-average downward deposition velocities and 42 n 43222 do not account for vegetation effects. The data used for the seleclted sites are shown in the Appendix 44 223 II. y 45 46 224 47 48225 Development of the ecosystem respiration model 49 50226 51 227 Site-by-site analysis – TP Model description 52 53 228 54 55229 For the analysis of R we started from a widely used climate-driven model: ‘TP Model’ (Eq. 1) ECO 56 57230 proposed by Raich et al. (2002) and further modified by (Reichstein et al., 2003a). Here we used the 58 59231 ‘TP Model’ for the simulation of RECO at the daily time-step using as abiotic drivers daily TA and P: 60 232 233 R = R ⋅ f (T )⋅ f (P) (1) ECO ref A 7