Received: 16 September 2016 | Revised: 18 February 2019 | Accepted: 24 June 2019 DOI: 10.1111/geb.12985 RESE ARCH PAPER Different determinants of radiation use efficiency in cold and temperate forests Manuela Balzarolo1,2 | Nadia Valdameri3 | Yongshuo H. Fu3,4 | Lennert Schepers5 | Ivan A. Janssens3 | Matteo Campioli3 1 CSIC, Global Ecology Unit, CREAF‐ CSIC‐UAB, Cerdanyola del Vallès, 08913, Abstract Catalonia, Spain Aim: To verify which vegetation and environmental factors are the most important 2 CREAF, Cerdanyola del Vallès, 08193, in determining the spatial and temporal variability of average and maximum values Catalonia, Spain 3 Centre of Excellence Plants and Ecosystems of radiation use efficiency (RUEann and RUEmax, respectively) of cold and temperate (PLECO), Department of Biology, University forests. of Antwerp, Wilrijk, 2610, Belgium 4 Location: Forty‐eight cold and temperate forests distributed across the Northern College of Water Sciences, Beijing Normal University, Beijing, 100875, China Hemisphere. 5 Ecosystem Management Research Group Major taxa studied: Evergreen and deciduous trees. (ECOBE), Department of Biology, University Time period: 2000–2011. of Antwerp, Wilrijk, 2610, Belgium Methods: We analysed the impact of 17 factors as potential determinants of mean Correspondence RUE (at 8 days interval, annual and interannual level) and RUEmax (at annual and in‐ Manuela Balzarolo, CSIC, Global Ecology Unit, CREAF‐CSIC‐UAB, Cerdanyola del terannual level) in cold and temperate forests by using linear regression and random Vallès, 08193, Catalonia, Spain; CREAF, forests models. Cerdanyola del Vallès, 08193, Catalonia, Spain. Results: Mean annual RUE (RUEann, c. 1.1 gC/MJ) and RUEmax (c. 0.8 gC/MJ) did not Email: m.balzarolo@creaf.uab.cat differ between cold and temperate forests. However, for cold forests, RUEann was Funding information affected by temperature‐related variables, while for temperate forests RUEann was H2020 Marie Skłodowska‐Curie Actions, affected by drought‐related variables. Leaf area index (LAI) was important for both Grant/Award Number: INDRO, grant no. 702717 forest types, while N deposition only for cold forests and cloud cover only for tem‐ perate forest. RUEmax of cold forests was mainly driven by N deposition and LAI, Editor: Andrew Kerkhoff whereas for temperate forests only a weak relationship between RUEmax and CO2 concentration was found. Short‐term variability of RUE was strongly related to the meteorological variables and varied during the season and was stronger in summer than spring or autumn. Interannual variability of RUEann and RUEmax was only weakly related to the interannual variability of the environmental drivers. Main conclusions: Cold and temperate forests show different relationships with the environment and vegetation properties. Among the RUE drivers observed, the least anticipated was N deposition. RUE is strongly related to short‐term and seasonal changes in meteorological variables among seasons and among sites. Our results This is an open access article under the terms of the Creat​ive Commo​ns Attri​bution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2019 The Authors. Global Ecology and Biogeography published by John Wiley & Sons Ltd Global Ecol Biogeogr. 2019;28:1649–1667.  wileyonlinelibrary.com/journal/geb | 1649 1650 | BALZAROLO et al. should be considered in the formulation of climate zone‐specific tools for remote sensing and global models. KEYWORDS meteorological and vegetation influences, forest ecosystems, gross primary production, light use efficiency, meta‐analysis, short‐term variability, spatial and temporal variabilities 1 | I NTRO D U C TI O N factors, was recognized (e.g., Grace et al., 2007). Production effi‐ ciency models define the light conversion factor as the product of Radiation use efficiency (RUE; gC/MJ) has emerged in recent de‐ an optimum RUE value (RUE potential or maximum, RUEmax) and cades as a key parameter to determine photosynthetic carbon other factors that relate to the environmental variables that regulate uptake by vegetation, and thus the carbon exchange between the at‐ photosynthesis and APAR. For example, in the moderate resolution mosphere and biosphere (Monteith, 1972). RUE represents the effi‐ imaging spectroradiometer (MODIS) algorithm, RUE is implicitly cal‐ ciency of vegetation to transform absorbed light energy into organic culated (to determine GPP) based on temperature, light, vapour pres‐ compounds; it is the ratio between gross primary production (GPP; sure deficit and the biome‐specific RUEmax derived from a look‐up gC/m2/year) and the absorbed photosynthetically active radiation table (Zhao, Running, & Nemani, 2006). However, many studies con‐ (APAR; MJ/m2/year), with APAR being the product of the incident sidering different biomes (Garbulsky et al., 2010; Runyon, Waring, photosynthetically active radiation (PAR; MJ/m2/year) and its frac‐ Goward, & Welles, 1994) have shown that RUE might be influenced tion absorbed by vegetation (fAPAR) (Monteith, 1972): not only by environmental factors, but also by a range of biophysi‐ cal and structural factors related to plant properties, sometimes af‐ GPP GPP RUE = = (1) fected by ecosystem management (e.g., irrigation, fertilization). Both APAR fAPAR × PAR environmental and vegetation factors that affect RUE are reported The global ecological modelling and the remote sensing com‐ in Table 1 with the aim of summarizing the current state of the art. munities are particularly interested in the RUE concept (e.g., Cheng, Zhang, Lyapustin, Wang, & Middleton, 2014; Grace et al., 2007; In detail, factors that have been reported to date as determinants King, Turner, & Ritts, 2011; McCallum et al., 2009; Running et al., of RUE are: (a) temperature‐related variables [air and soil tempera‐ 2004; Stocker et al., 2018; Wang, Prentice, & Davis, 2014; Wang et ture, length of the warm period (i.e., the number of months with al., 2017; Yuan et al., 2014;). Nevertheless, to date, no consensus mean temperature above 5 °C), thermal amplitude (i.e., mean maxi‐ has been reached regarding the most suitable algorithm for RUE mum minus mean minimum temperature)]; (b) water status variables (Gitelson & Gamon, 2015) and scientists still need to fully under‐ [vapour pressure deficit (VPD), precipitation, intensity and duration stand if (and how) models should simulate the impact of environ‐ of drought, evaporative fraction, actual and potential evapotranspi‐ mental factors on RUE (Baldocchi, 2018). To do this, first we need to ration, soil water content and deficit, irrigation]; (c) radiation‐related find suitable proxies that capture well RUE variability in space (from variables [diffuse light and leaf area index (LAI)]; (d) variables related local to global) and in time (from daily to annual). RUE changes over to leaf and vegetation characteristics [stand age, leaf habit and type time because GPP is affected by the current environmental condi‐ and biome type] and (e) fertility‐related variables [nitrogen (N) depo‐ tions and because the absorbed radiation is affected by changes in sition, leaf N, management and CO2 concentration]. The relevance of incident PAR and leaf properties (e.g., leaf and chloroplast move‐ these factors is reviewed below based on the literature. ment), which can occur over short time periods. On the other hand, (a) Temperature‐related variables. The impact of air temperature the impact of respiration variability on RUE is expected to be low in on RUE has been tested the most, but contrasting results have the short term (when plant respiration is thought to be independent been found, ranging from significant (Chasmer et al., 2008; Kergoat, of photosynthesis) but it can be greater over longer, annual time‐ Lafont, Arneth, Dantec, & Saugier, 2008; Mäkelä et al., 2008; scales (when plant respiration is thought to be proportional to pho‐ Schwalm et al., 2006) to non‐significant (Jenkins et al., 2007; Turner tosynthesis) (Ryan, Linder, Vose, & Hubbard, 1994). et al., 2003), or with impact limited to periods during the warm sea‐ First, RUE was used as a constant parameter (e.g., Myneni, Los, son (see above definition; Fernández‐Martínez et al., 2014) or only & Asrar, 1995). Subsequently, interannual variability has been rec‐ at cold sites (Garbulsky et al., 2010). The latter evidence is the most ognized. The ratio between the accumulated GPP over the year and anticipated, as the positive effect of temperature on photosynthesis the total solar radiation absorbed by the canopy was then defined as is larger in cold environments and APAR is not expected to depend annual RUE (RUEann). This definition has been widely used in simple significantly on temperature. Furthermore, Fernández‐Martínez crop growth models, based on in situ observations (McCallum et al., et al. (2014) tested the thermal amplitude and the length of the warm 2009), as well as in meta‐analyses seeking general ecological eluci‐ season, finding them not relevant to RUE variability. dations (Fernández‐Martínez et al., 2014). Finally, the intra‐annual (b) Water status variables. Water limitation is generally expected to variability of RUE, due to its dependency on seasonal environmental reduce RUE because of reduced photosynthesis following reduction TA B L E 1 List of the vegetation and environmental drivers tested in the literature as potential drivers of radiation use efficiency (RUE) in forests Calculation of RUE Potential RUE driver Impact GPP fAPAR, APAR Annual or maximum Study period Level Biome type(s) Reference Cold forests BALZAROLO et al. Soil water content * EC RS Annual Growing season Ecosystem Forest: boreal (n = 4, Chasmer et al. Diffuse light * chronosequence) (2008) Vapour pressure deficit * Air temperature * Stand age * Canopy structure * Temperate forests Fertilization 0 ANPP I Annual Annual Ecosystem Forest: temperate Allen et al. (2005) Irrigation * (n = 2) Diffuse light (diffuse : total radiation ratio) * EC I Annual Growing season Ecosystem Forest: temperate Jenkins et al. (2007) Vapour pressure deficit 0 (n = 1) Air temperature 0 Soil temperature 0 Soil water content * ANPP I Annual Annual Ecosystem Forest: temperate Bracho et al. (2012) Vapour pressure deficit * (n = 2) Canopy structure (i.e., LAI), stand age * Fertilization 0 Fertilization * ANPP P Annual Annual Tree Forest: temperate Campoe et al. (2013) Irrigation 0 (n = 1) Fertilization + irrigation * Tree size (effect of dominance) * CO2 concentration * ANPP I Annual Growing season Ecosystem Forest: temperate De Kauwe et al. (n = 2) (2016) Forests in different climate zones Diffuse light * EC I Annual Growing season Ecosystem Forest: boreal (n = 1), Alton, North, and temperate (n = 1), Los (2007) tropical (n = 1) Canopy nitrogen * EC I Maximum RUE Annual Ecosystem Forest: boreal (n = 3), Ollinger et al. (2008) temperate (n = 9) Leaf nitrogen 0 P I Annual Annual Ecosystem Forest: boreal (n = 4), Mäkelä et al. (2008) Air temperature * temperate (n = 3) | Vapour pressure deficit * 1651 Soil water content 0 (Continues) TA B L E 1 (Continued) 1652 | Calculation of RUE Annual or Potential RUE driver Impact GPP fAPAR, APAR maximum Study period Level Biome type(s) Reference Canopy nitrogen * P P Maximum RUE Annual Leaf Forest: boreal (n = 7) Peltoniemi et al. Vapour pressure deficit 0 + temperate (n = 7) (2012) Air temperature 0 Air temperature * EC RS Annual Growing season Ecosystem Forest: boreal (n = 9), Fernández‐Martínez Actual evapotranspiration 0 Annual temperate (n = 21), et al. (2014) Mediterranean Precipitation 0 Annual (n = 9) Water deficit 0 Annual Thermal amplitude 0 Annual Length of warm period 0 Annual Nitrogen deposition 0 Annual Stand age 0 Annual Leaf habit and type 0 Annual Management 0 Annual Forests and other biomes Diffuse light * P P Annual Growing season Ecosystem Forest: boreal (n = 2), Gu et al. (2002) temperate (n = 1) + cropland (n = 1) + grassland (n = 1) Biome type * EC I Annual Growing season Ecosystem Forest: boreal (n = 1), Turner et al. (2003) Diffuse light * temperate (n = 1) + cropland (n = 1) + Vapour pressure deficit 0 grassland (n = 1) Air temperature max 0 Biome type * EC I Annual Annual Ecosystem Forest: boreal Schwalm et al. Leaf nitrogen 0 (n = 11), temper‐ (2006) ate (n = 7), tropical Water cycle‐related variables * (n = 2) + wetland Diffuse light *(n = 6) (n = 3) + grassland Air temperature * (n = 1) LAI * Evaporative fraction * (growing EC RS Annual Growing season Ecosystem Forest: boreal (n = 5), Yuan et al. (2007) (water stress; moisture) season) temperate (n = 18) Air temperature * (begin and + grassland (n = 4) + end of grow‐ savanna (n = 1) ing season) BALZAROLO et al. (Continues) TA B L E 1 (Continued) Calculation of RUE Annual or Potential RUE driver Impact GPP fAPAR, APAR maximum Study period Level Biome type(s) Reference BALZAROLO et al. Leaf nitrogen * EC I Annual Annual Ecosystem Forest: boreal Kergoat et al. (2008) Temperature * (n = 12), tem‐ perate (n = 17), Leaf nitrogen + air temperature * Mediterranean Biome types 0 (n = 3) + tundra (n = 3) + grassland (n = 2) + cropland (n = 5) Long‐term mean annual precipitation * EC RS Annual Annual Ecosystem Forest: boreal (n = 4), Garbulsky et al. Long‐term mean annual temperature 0 temperate (n = 10), (2010) Mediterranean Precipitation 0 (n = 3), tropical Actual evapotranspiration 0 (n = 4) + tundra Air temperature 0 (n = 2) + shrubland (n = 1) + grassland Biome types 0 (n = 7) + savanna (n = 3) + cropland (n = 1) Long‐term mean annual precipitation 0 EC RS Maximum RUE Annual Ecosystem See row above Garbulsky et al. Long‐term mean annual temperature 0 (2010) Precipitation * Air temperature 0 Actual evapotranspiration 0 Biome types 0 Biome types * EC RS Annual Growing season Ecosystem See two rows above Garbulsky et al. Air temperature 0 (2010) Vapour pressure deficit 0 Evaporative fraction * AET/PET 0 Precipitation 0 Symbols: * = significant impact; 0 = no significant impact. Abbreviations: Potential drivers: LAI = leaf area index; AET = actual evapotranspiration; PET = potential evapotranspiration. Variables used to calculate RUE: GPP = gross primary production (measure‐ ment methods of GPP: EC = eddy covariance; ANPP = measurements based on aboveground net primary productivity; P = simulated by processed based models); APAR: absorbed photosynthetically active radiation; fAPAR: fraction of APAR; (measurement methods of fAPAR: I = derived from in situ optical LAI or radiation measurements; P = simulated by processed based models; RS = derived from satellite data). | 1653 1654 | BALZAROLO et al. in stomatal conductance or stomata closure. VPD has frequently but also indirectly (through its impact on leaf and plant structure, been tested as a potential driver of RUE. As can be seen in Table 1, which influence light absorption). Leaf N has been therefore consid‐ both significant (Bracho et al., 2012; Chasmer et al., 2008; Mäkelä ered as a potential driver of RUE in manifold studies on forest eco‐ et al., 2008; Runyon et al., 1994) and non‐significant (Garbulsky et systems (Kergoat et al., 2008; Ollinger et al., 2008; Peltoniemi et al., al., 2010; Jenkins et al., 2007; Turner et al., 2003) impacts of VPD 2012) but has not been found to be consistently significant (Mäkelä on RUE have been found. Also annual precipitation was identified et al., 2008; Schwalm et al., 2006). In a recent study, Fernández‐ both as a significant (Garbulsky et al., 2010) and a non‐significant Martínez et al. (2014) tested the relationship between RUE and driver of RUE (Fernández‐Martínez et al., 2014). In the latter work, N deposition but no significant association was found. Finally, De the water deficit was tested as potential driver of RUE (i.e., an indica‐ Kauwe, Keenan, Medlyn, Prentice, and Terrer (2016) have recently tor of the intensity of water stress that the vegetation must tolerate) shown that CO2 is a key factor controlling RUE variability. These au‐ but with a negative response. Soil water content was reported as a thors found a large increase in RUE due to elevated CO2 for two significant driver of RUE in one study on temperate forests (Bracho long‐term free air carbon enrichment (FACE) forest sites (Oak Ridge: et al., 2012) and one on cold forests (Chasmer et al., 2008), while 1998–2008 and Duke: 1996–2007). it was not relevant in the study of Mäkelä et al. (2008), comparing Overall, even though the underlying mechanisms for the differ‐ cold and temperate forests. The evaporative fraction was found ent potentially influencing factors seem clear, our literature review to be a significant driver by Garbulsky et al. (2010) and Yuan et al. indicates that there is still high uncertainty about the vegetation and (2007). Actual evapotranspiration was identified as an important environmental drivers of RUE, their relative importance and their variable by Garbulsky et al. (2010), but not by Fernández‐Martínez different impacts on forest ecosystems of different climate zones. et al. (2014). Two irrigation experiments on temperate forests (Allen, Moreover, the biological and physiological processes that regulate Will, McGarvey, Coyle, & Coleman, 2005; Campoe et al., 2013) both photosynthesis and radiation absorption vary daily over the grow‐ agreed on the potential influence of irrigation on RUE. ing season and, potentially, among years. Therefore, here, we aim to (c) Radiation‐related variables. The role of diffuse light has been elucidate the spatial (across the Northern Hemisphere) and temporal tested in several studies, focused only on forests or combining for‐ (short‐term, annual and interannual) variability of RUE in temperate ests with other biomes (Table 1). All studies agreed on its potential and cold forests by comparing the impact on RUE and RUEmax of role in defining RUE variability because the reduction that the dif‐ several potential drivers. The short‐term variability refers to the fuse light triggers in APAR is larger than the reduction it triggers in variability of RUE among 8‐day time windows (RUE8days), as 8 days is GPP (Gu et al., 2002). The APAR term in Monteith’s model depends a common time reference for remote sensing. The focus on RUE8days, not only on the amount of PAR, but also on the PAR absorbance RUEann and RUEmax offers the most comprehensive insight into RUE capacity, and thus canopy structure, leaf area index (LAI) (Bracho dynamics and their possible implementation in global monitoring et al., 2012; Schwalm et al., 2006), leaf angle distribution and photo‐ tools. synthetic pigment content. (d) Variables related to leaf and ecosystem characteristics. Stand 2 | M ATE R I A L S A N D M E TH O DS age, leaf habitat and type, and biome types have been tested as potential drivers of RUE. Stand age has been identified both as a 2.1 | Forest categories and sites significant (Bracho et al., 2012; Chasmer et al., 2008) and non‐sig‐ nificant driver of RUE (Fernández‐Martínez et al., 2014), whereas It appears clear that most of the variables (e.g., related to tem‐ leaf habitat and type were found to be non‐significant (Fernández‐ perature, water status, fertility) potentially affecting RUE differ Martínez et al., 2014). It has been reported by several authors that between the two main climate zones where the majority of the RUE varies with vegetation type (Field, Randerson, & Malmstrom, northern forests are situated, that is, the temperate and cold 1995, Garbulsky et al., 2010; Prince & Goward 1995; Schwalm et al., zones. As for most of the previous studies on RUE variability (see 2006; Turner et al., 2003) because of the different ratio of respira‐ Table 1), we therefore separated the two forest types in our analy‐ tion to photosynthesis. RUE was found to be around 2 gC/MJ for sis. This categorization is also important to generalize our findings annual crops, for which the ratio of respiration to photosynthesis is and match the forest land classification used in global vegetation assumed to be low, whereas for woody plants the value of RUE var‐ modelling. ies from 0.2 to 1.5 gC/MJ because the ratio of respiration to photo‐ The categorization of cold and temperate is based on the synthesis is assumed to increase with plant size (Hunt, 1994; Waring Köppen–Geiger classification (Peel, Finlayson, & McMahon, 2007), & Running, 1998). using monthly temperature data from the European Commission— (e) Fertility‐related variables. The impact of fertility on RUE has Joint Research Centre—Monitoring Agricultural ResourceS been tested in two experiments on temperate forests but contrasting (EC‐JRC‐MARS, https​://ec.europa.eu/jrc/en/mars) portal. The results were found [significant in Campoe et al. (2013), non‐signifi‐ Köppen–Geiger classification defines a particular site as “cold” when cant in Allen et al. (2005)]. Bracho et al. (2012) identified fertility as the temperature of the hottest month is > 10°C, while the tempera‐ not influential for RUE in drought conditions. Leaf nitrogen content ture of the coldest month is below or equal to 0°C. On the other affects RUE directly (mainly through its impact on photosynthesis) hand, sites are classified as “temperate” when the temperature of BALZAROLO et al. | 1655 the hottest month is > 10°C and the temperature of the coldest about 45% of the total incoming shortwave radiation is in the PAR re‐ month is between 0 and 18°C. gion (Campbell & Norman, 1998). The physical unit of the total short‐ For the analysis on spatial variability of RUEann we used 26 cold wave radiation reported in the EC‐JRC‐MARS database is MJ/m2/day. and 22 temperate forests in the Northern Hemisphere (Figure 1). The physical unit of incoming PAR in the Fluxnet and European Flux These sites were selected because they had both GPP derived from Database Cluster is μmol/photons/m2/s. We converted PAR from eddy covariance flux measurements and the associated satellite μmol/photons/m2/s to J/m2/s using a conversion factor of 4.55 μmol/J value of fAPAR available (see below for fAPAR determination). as proposed by Goudriaan and Van Laar (1994). Finally, we obtained Multiple years were considered when available. Total site–year daily values in MJ/m2/day by multiplying by 0.0864. For sites with combinations were 114 for cold and 102 for temperate forests (see availability of both, preliminary analyses indicated a high correlation Supporting Information Appendix S1). For the analysis of the spatial between PAR data from EC‐JRC‐MARS and the Fluxnet/ European variability of RUEmax, only the sites with seasonal data for GPP and Fluxes Database Cluster (R2 = .8; p < .001; slope = 1.05). fAPAR were selected. A subset of 20 cold forests and 20 temperate fAPAR. 8‐day values of fAPAR (fPAR8days) were derived from the forests was thus used (see Supporting Information Appendix S1). fAPAR/LAI product (MOD15A2, collection 5) from the MODIS/ The temporal analyses (interannual variability of RUEann and RUEmax TERRA satellite sensor/platform as provided by the Oak Ridge and short‐term RUE variability, RUE8days) were conducted by con‐ National Laboratory Distributed Active Archive Center (ORNL sidering 11 sites that had at least 8 years of seasonal data for GPP DAAC). MOD15A2 pixel values represent the optimal fPAR8days at and fAPAR. 1‐km spatial resolution (Myneni et al., 2002). Only the pixels with the highest quality based on the Quality Assurance/Quality Control flags provided by MODIS (e.g., clear conditions without snow) were 2.2 | PAR, fAPAR, GPP and RUE selected and retained for computing fAPAR values. For each year, PAR. The cumulative annual value of PAR (PARann) and 8‐day PAR annual fAPAR (fAPARannW ) was calculated from weighted fAPAR8days (PAR8days) were calculated, for every site–year combination, from data where the weighting was provided by PAR8days data. seasonal daily data for PAR using (a) free fair‐use data files from the APAR. Annual APAR (APARann) was computed as fAPARannW mul‐ Fluxnet and European Fluxes Database Cluster (69% of the sites) and tiplied by the cumulative PAR (PARann). To derive APAR8days, we mul‐ (b) the EC‐JRC‐MARS data set for the remaining sites. (a) In the Fluxnet tiplied fAPAR8days by the PAR8days. and European Fluxes Database Cluster files the measurement of the GPP. GPP data (annual comulative GPP) and 8‐day GPP (GPP8days) incoming radiation in the PAR region (400–700 nm) is reported. (b) were derived from publicly available databases of forest ecosys‐ EC‐JRC‐MARS reports only total shortwave incoming radiation. We tem carbon fluxes (Luyssaert et al., 2007, Fluxnet, European Flux therefore multiplied the radiation by a factor of 0.45, assuming that Database Cluster). F I G U R E 1 Map with distribution of sites: cold forests and temperate forests are represented with red and blue circles, respectively [Colour figure can be viewed at wileyonlinelibrary.com] 1656 | BALZAROLO et al. RUEann, RUEmax and RUE8days. According to Monteith’s equation VPD = es - ea (2) (Equation 1), we calculated RUEann as the ratio between GPPann and APARann. To calculate RUEmax for a year we computed 8‐day RUE ( ) 17.27T (RUE8days) values for the whole growing season from the ratio be‐ es = 0.6108exp (3) T + 237.3 tween GPP8days and APAR8days values. RUEmax was defined as the maximum value of the RUE8days time series. RH ea = e s (4) 100 2.3 | Explanatory drivers where T is the air temperature (°C) and RH is the relative humidity (%). Mean daily values of VPD over an 8‐day time window and the 2.3.1 | Analyses of spatial variability annual mean of daily values were used in our analyses. RH data were available from the Fluxnet and European Fluxes Database We examined the following potential determinants of RUE ann and Cluster for 89% of the sites, while for the other sites we used data RUEmax that were in common with previous studies: leaf type and from the public online Global Weather Data for soil & water as‐ habit, annual potential evapotranspiration, annual precipitation, sessment tool (SWAT) (Dile & Srinivasan, 2014; Fuka et al., 2014) mean annual temperature, mean monthly minimum and maximum that provides interpolated RH from local meteorological stations. air temperature, N deposition, VPD, cloud cover data, LAI, arid‐ Cloud cover was used as a proxy of diffuse light, as the latter is ity index, leaf N and CO2 concentration. Furthermore, we decided a parameter that is not generally measured by meteorological sta‐ to complete the group of determinants with the addition of some tions and flux towers. We extracted cloud cover data (mean annual new variables related to soil fertility, number of days with mean value as percentage) from Climatic Research Unit (CRU) Time‐Series daily temperature below zero (i.e., freezing period) and duration of Version 3.22 (Harris & Jones, 2014). the longest period without rain. These variables were determined for each site and year for which RUE ann was available, except for (b) Soil fertility LAI and leaf N concentration, for which data were not consistently The soil type of each site was derived from the food and agriculture available for all years (averages were made with the available data) organization (FAO) digital Soil Map of the World Version 3.6 (FAO, or sites (leaf N concentration was available only at about half of 2007). Subsequently, soil types were classified into three levels (H: the sites and only those sites were considered in the data analysis high, M: medium, L: low) of nutrient availability (see Supporting for leaf N). Information Appendix S1), based on fertility information reported in Creutzberg (1987). 2.3.2 | Analyses of temporal variability (c) Leaf type and habit For the analysis of short‐term variability of RUE8days the available We collected the description of each site from the database of (meteorological) variables were: mean 8‐day temperature, 8‐day Luyssaert et al. (2007). The two leaf type categories are needle‐ minimum and maximum temperature, 8‐day mean potential evap‐ leaved and broadleaved, while the categories deciduous and ever‐ otranspiration, 8‐day mean VPD and 8‐day mean CO2 concentra‐ green represent the leaf habit of the forest tree species. tion. For the analysis of the interannual variability of both RUE ann and RUEmax , we considered the same meteorological explanatory (d) N deposition variables used for the analysis of short‐term variability of RUE8days . In our analysis, N deposition was considered as the sum of dry and wet deposition. We extracted these data from the global N deposition (a) Meteo‐variables data set simulated with goddard earth observing system‐Chem (GEOS; Mean annual temperature, mean monthly minimum and maximum http://acmg.seas.harvard.edu/geos/index.html) for years from 2004 temperature, annual precipitation and potential evapotranspiration, to 2006, at 2°×2.5° grid resolution (Ackerman, Chen, & Millet, 2018). yearly number of days with mean daily temperature below zero and the duration of the longest period without rain (expressed in num‐ (e) LAI and leaf N concentration ber of days), mean 8‐day temperature, 8‐day minimum and maxi‐ We extracted LAI data from the forest database of Luyssaert et mum temperature, 8‐day mean potential evapotranspiration and al. (2007) and the literature (see Supporting Information Appendix 8‐day mean VPD were calculated using the EC‐JRC‐MARS data set. S2). Leaf N was obtained from the ancillary files of the Fluxnet and The evaporation data that we used are derived from Penman (1948) European Fluxes Database Cluster and the literature (see Supporting and they represented the annual potential evapotranspiration. We Information Appendix S3). For both LAI and leaf N concentration, we calculated the annual aridity index as the ratio of annual precipita‐ used the maximum annual value. tion to annual evapotranspiration. VPD (in kPa) is the difference between the saturation vapour (f) CO2 concentration pressure (es, kPa) at air temperature and actual vapour pressure (ea, For 81% of the sites, we used CO2 concentration data from the kPa) (Allen, Pereira, Raes, & Smith, 1998): Fluxnet database. For the remaining sites, we used values provided BALZAROLO et al. | 1657 in the global data set compiled by the Institute for Atmospheric and 2.5 | Temporal analysis of short‐term RUE Climate Science at the Eidgenössische Technische Hochschule in variability Zürich (Switzerland) for the Northern Hemisphere (https​://www. For this test, we performed univariate analyses considering as de‐ co2.earth/​histo​rical-co2-datasets). pendent variable the time series of RUE8days and as independent variables the 8‐day time series of the predictors (see above). The 2.4 | Statistical analysis analyses were done for each of the 11 selected sites, separately, and averages were done across years. First, the analyses were done for 2.4.1 | Spatial variability of RUEann and RUEmax all 8‐day periods within the growing season. Second, the analyses were run separately for three periods representing the main sea‐ The analyses of RUEann and RUEmax were conducted separately sons: spring (considering 8‐day windows from 15 April to 15 June), for cold forests and temperate forests (see Tables 2 and 3), with summer (from 16 June to 15 August) and autumn (from 16 August the analysis on the entire data set reported only in the Supporting to 15 October). Information (see Appendices S4 and S5). Three main analyses were performed: univariate analysis, random forest and linear models with multiple predictors. Analyses on leaf N were limited to the uni‐ 2.5.1 | Temporal analysis of interannual variate analysis (see below), because we did not have leaf N avail‐ variability of RUEann and RUEmax able for all sites (see above). (a) To describe the correlation between This analysis was done separately on the selected 11 sites (see each predictor and RUEann and RUEmax, we conducted univariate above). We conducted a univariate analysis to evaluate the good‐ analysis for each variable. We used single linear regression for ness of the linear correlation between RUE (RUEann and RUEmax) of continuous variables and one‐way ANOVAs with post hoc Tukey’s each year and the value of each predictor variable (see above for the honestly significant difference (HSD) test for categorical variables. variables list). For RUEann, we used annual values of the predictor Shapiro–Wilk’s normality test and Levene’s test for homoscedas‐ variables. For predictors of RUEmax, we used the values of the 8‐day ticity were always passed. (b) We used random forest analysis window corresponding to the 8‐day period associated with RUEmax. (Breiman, 2001) to estimate the relative importance of the dif‐ All the statistical analyses were done using R (R Core Team, ferent variables. This analysis ranks the factors from the one with 2015). the strongest impact to the one with the lowest impact, while not exclusively considering linearity but also other, nonlinear, types of relationships. The ranking is based on the mean decrease in accu‐ racy of model prediction (%IncMSE) when the variable is randomly 3 | R E S U LT S permuted. We used a standard random forest algorithm (Liaw & Wiener, 2002) with 50,000 trees. For ranking the predictors, we Mean RUEann for cold and temperate forests was very similar: 1.10 preferred random forest to multiple regression analysis because (SD ±0.39) gC/MJ for cold forests and 1.11 (SD ±0.45) gC/MJ for of the ability of random forest analysis to consider also nonlinear temperate forests. The variability of RUEann was larger than the relationships and because the relatively small sample sizes would variability of annual GPP [1,221.46 (SD ±412.60) gC/m2/year and have made the ranking with multiple regression analysis less robust. 1,510.27 (SD ±329.53) gC/m2/year for cold and temperate forests, (c) Finally, we built linear models with the predictors to evaluate respectively]. Cold forests showed the largest RUEmax but also the how much of the variability in RUE could be explained by combina‐ largest variability across sites [0.86 (SD ±0.42) gC/MJ], whereas tem‐ tions of predictors. To build these models, we first detected the perate forests presented slightly lower values [0.79 (SD ±0.26) gC/ variables that were highly correlated by doing a multicollinearity MJ]. test [recording the variance inflation factor (VIF)] and by exploring The results of the univariate analyses are shown in Tables 2 and the relationships among all the variables with a bivariate analysis by 3 for RUEann and RUEmax, respectively, for both cold and temper‐ doing a matrix of Pearson’s correlation coefficients. Practically, we ate forests. For cold forests, temperate‐related variables were sig‐ removed all variables that had a VIF > 5 or had a correlation coef‐ nificantly correlated with RUEann, in particular the number of days ficient > |.8| with another variable. With the variables that passed with mean daily temperature below zero (p = .001, R2 = .35), annual the multicollinearity and correlation test, we performed a stepwise temperature (p = .03, R2 = .14) and, but with a weaker trend, Tmin backwards regression analysis (SBRA). This is a process of building a (p = .06, R2 = .11). Moreover, N deposition (p = .02, R2 = .18) and model considering at first all variables together and successively re‐ LAI (p = .05, R2 = .12) were also significant determinants of spatial moving the least important ones. The original model was compared variability of RUEann (Table 2). On the other hand, RUEmax showed with the new model, with one variable removed, by using the likeli‐ a significant relationship only with N deposition (p = .02, R2 = .20) hood ratio and Akaike information criterion (AIC). The new model and LAI (p = .05, R2 = .16) and only a weak relationship with tem‐ was not accepted if the likelihood ratio was significant (p < .05) or perature (Tmin; p = .07; R2 = .13) (Table 3). For temperate forests, the AIC increased (i.e., we considered as the final model the one a different picture emerged. First, RUEann showed dependencies that respected those assumptions). with variables related to the water status instead of temperature, 1658 | BALZAROLO et al. TA B L E 2 The impact of vegetation and environmental drivers on annual radiation use efficiency (RUEann) for cold (n = 26) and temperate forests (n = 22), from univariate analysis and stepwise backwards regression analysis (SBRA) Cold forest (n = 26) Temperate forest (n = 22) Univariate analysisb Univariate analysisb Adj. R2 Adj. R2 a c Potential predictor Variable type p value (sign) Post hoc SBRA p value (sign) Post hoc SBRAc Water‐related Aridity index con. .73 −.04 (‐) X .003** .34 (+) X Longest period with‐ con. .78 −.04 (‐) X .01* .23 (‐) X out rain Annual precipitation con. .72 −.04 (+) X .02* .19 (+) Evapotranspiration con. .62 −.03 (+) X .56 −.03 (‐) X ○ Vapour pressure con. .34 −.003 (+) .05 .13 (‐) X deficit Temperature‐related Number days under con. .001** .35 (‐) X .27 .01 (+) 0°C Annual temperature con. .03* .14 (+) .16 .05 (‐) X Tmin con. .06○ .11 (+) X .25 .02 (‐) X Tmax con. .99 −.04 (+) X .13 .06 (‐) X Radiation‐related Cloud cover data con. .93 −.04 (+) .002** .34 (+) X LAI con. .05○ .12 (+) .01* .24 (+) X Others N deposition con. .02* .18 (+) .1 .09 (+) d Soil fertility cat. .23 L‐H: .38 M‐H: X −0.30 −0.35 Leaf habite cat. .2 BN‐B: .36 BN‐B: −0.26 0.27 Leaf typef cat. .28 E‐D: .74 E‐D: −0.27 −0.12 Leaf Ng con. .39 −.01 (+) .36 .001 (+) CO2 con. .73 −.04 (‐) .21 .03 (+) Stepwise backwards R2 = .57 R2 = .38 regression model Abbreviations and symbols: LAI = leaf area index; Tmin and Tmax = mean monthly minimum and maximum air temperature, respectively; Number days under 0°C = the number of days in a year with mean daily temperature below zero. ○ = .05 < p < .10; * = .01 < p < .05; ** = .001 < p < .01; *** = p < .001; (+) = positive linear regression; (‐) = negative linear regression. aVariable type: “con.” when continuous; “cat.” when categorical. bFor continuous variables, the significance level of the linear regression (p) and adjusted R2 are reported, with “sign” as the sign of the linear regression. For categorical variables, we report results of one‐way ANOVA (p value) and post hoc Tukey’s honestly significant difference (HSD) test (absolute difference for two significantly different factors and p value of the difference). cFor SBRA, we report the variables of the final model, representing the key predictors of RUE, and (in the last row) the coefficient of determination (R2) of the final model; dSoil fertility was classified as H = high; M = medium; L = low. eLeaf habit was classified as N = needleleaved; B = broadleaved; BN = mixed habit. fLeaf type was classified as D = deciduous; E = evergreen. gLeaf N: variable tested only for univariate analysis as with fewer sites than other variables (cold: n = 11; temperate: n = 12). in particular annual precipitation (p = .02, R2 = .19), aridity index except a weak relationship with CO2 concentration (p = .09, R2 = .10) 2 2 (p = .003, R = .34), longest period without rain (p = .01, R = .23) (Table 3). 2 and VPD (p = .05 R = .13) (Table 2). Second, RUEann of temperate In general, there was a good agreement between the univariate forests showed mainly different patterns from those of RUEann analysis and random forest analysis and the variables that were sig‐ of cold forests such as a significant relationship with cloud cover nificant in the univariate analysis also showed high importance in the (p = .002, R2 = .34) and a non‐significant relationship with N deposi‐ random forest analysis (Figures 2 and 3). However, there were ex‐ 2 tion (p = .62, R = −.04) (Table 2). Finally, RUEmax of temperate forests ceptions. First, for temperate forests, the relation between RUEann did not show significant relationships with any explanatory variable and maximum air temperature was not significant according to the BALZAROLO et al. | 1659 TA B L E 3 The impact of vegetation and environmental drivers on maximum radiation use efficiency (RUEmax) for cold (n = 20) and temperate forests (n = 20), from univariate analysis and stepwise backwards regression analysis (SBRA) Cold forest (n = 20) Temperate forest (n = 20) Univariate analysisb Univariate analysis(b) Adj. R2 Adj. R2 a c Potential predictor Variable type p value (sign) Post hoc SBRA p value (sign) Post hoc SBRAc Water‐related Aridity index con. .23 .03 (+) X .26 .02 (+) X Longest period with‐ con. .36 −.01 (‐) X .11 .09 (‐) X out rain Annual precipitation con. .20 .04 (+) X .56 −.04 (+) X Evapotranspiration con. .79 −.05 (+) X .55 −.03 (‐) X Vapour pressure con. .91 −.05 (+) X .45 −.02 (‐) X deficit Temperature‐related Number days under con. .13 .08 (‐) X .80 −.05 (+) X 0°C Annual temperature con. .14 .07 (+) X .40 −.01 (‐) Tmin con. .07○ .13 (+) X .50 −.03 (‐) X Tmax con. .91 −.05 (‐) X .55 −.03 (‐) X Radiation‐related Cloud cover data con. .93 −.06 (+) X .23 .03 (+) X LAI con. .05* .16 (+) .20 .04 (+) X Others N deposition con. .02* .20 (+) X .62 −.04 (+) X d Soil fertility cat. L‐H: 0.77 X L‐H: 0.36 X Leaf habite cat. BN‐B: BN‐B: X 0.49 0.91 Leaf typef cat. E‐D: 0.49 E‐D: 0.47 X g Leaf N con. .79 −.10 (‐) .29 .02 (+) CO2 concentration con. .80 −.05 (‐) X .09○ .10 (+) 2 Stepwise backwards R = .44 R2 = .88 regression model Abbreviations and symbols: LAI = leaf area index; Tmin and Tmax = mean monthly minimum and maximum air temperature, respectively; Number days under 0°C = the number of days in a year with mean daily temperature below zero. ○ = .05 < p < .10; * = .01 < p < .05; ** = .001 < p < .01; *** = p < .001. (+): positive linear regression; (‐): negative linear regression. aVariable type: “con.” when continuous; “cat.” when categorical. bFor con‐ tinuous variables, the significance level of the linear regression (p) and adjusted R2 are reported, with “sign” as the sign of the linear regression. For categorical variables, we report results of one‐way ANOVA (p value) and post hoc Tukey’s honestly significant difference (HSD) test (absolute differ‐ ence for two significantly different factors and p value of the difference). cFor SBRA, we report the variables of the final model, representing the key predictors of RUE, and (in the last row) the coefficient of determination (R2) of the final model. dSoil fertility was classified as H = high; M = medium; L = low. eLeaf habit was classified as N = needleleaved; B = broadleaved; BN = mixed habit. fLeaf type was classified as D = deciduous; E = evergreen. g Leaf N = variable tested only for univariate analysis as with fewer sites than other variables (cold: n = 11; temperate: n = 12). univariate analysis but was highly ranked in random forest analysis analysis but Tmean was highly ranked in the random forest analy‐ (see Figure 2 and Table 2). This is probably due to the hyperbolic sis, likely because of a few exceptional sites (Supporting Information relationship between maximum temperature and RUEann which was Appendix S7). Third, for RUEann and RUEmax of cold forests, LAI was likely detected by the random forest but not by the univariate analy‐ significant according to the univariate analysis but of very low im‐ sis (Supporting Information Appendix S6). However, we believe this portance in the random forest analysis. This suggests that for RUEann relationship to be spurious as it is driven by only three sites with and RUEmax of cold forests, nonlinear relationships might dominate exceptionally high maximum temperatures (Supporting Information and overshadow the linear dependencies. Appendix S6). Second, similarly to the previous case, RUEmax of cold Tables 2 and 3 also show linear models combining multiple pre‐ forests was not significantly correlated with Tmean in the univariate dictors. The models explained 57 and 38% of the spatial variability of 1660 | BALZAROLO et al. F I G U R E 2 Relative importance of vegetation and environmental drivers for annual radiation use efficiency (RUEann) for cold forests (n = 26) and temperate forests (n = 22). Data are from (a) random forest analysis (cold forests: light grey bars, temperate forests: black bars) with variable importance positively related to accuracy of model prediction (%IncMSE), and negative %IncMSE indicating lack of importance and (b) univariate analysis, with significant (p < .10) drivers marked with “*” symbol (see text and Table 2 for details). Abbreviations: Tmin and Tmax = mean monthly minimum and maximum air temperature, respectively; N. days under 0°C = the number of days in a year with mean daily temperature below 0°C F I G U R E 3 Relative importance of vegetation and environmental drivers for maximum radiation use efficiency (RUEmax) for cold forests (n = 20) and temperate forests (n = 20). Data are from (a) random forest analysis (cold forests: light grey bars, temperate forests: black bars) with variable importance positively related to accuracy of model prediction (%IncMSE), and negative %IncMSE indicating lack of importance and (b) univariate analysis, with significant (p < .10) drivers marked with “*” symbol (see text and Table 3 for details). Abbreviations: Tmin and Tmax = mean monthly minimum and maximum air temperature, respectively; N. days under 0°C = the number of days in a year with mean daily temperature below 0°C RUEann for cold and temperate forests, respectively, and 44 and 88% As mentioned above, we also carried out an analysis on the en‐ of the spatial variability of RUEmax for cold and temperate forests, tire data set, without separating forests into the cold and temperate respectively. However, note that a high number of variables were categories. For RUEann, both temperature‐related variables (import‐ retained in the best models (see Tables 2 and 3). ant for cold forests) and drought‐related variables (important for BALZAROLO et al. | 1661 temperate forests) were significant. Moreover, other variables, sig‐ TA B L E 4 Statistics (mean and SD) for interannual variability in nificant for at least one of the two forest types, retained their impor‐ annual (RUEann) and maximum radiation use efficiency (RUEmax) in cold and temperate forests. Sites’ full names can be found in tance (e.g., cloud cover, LAI, N deposition) (Supporting Information Supporting Information Appendix S1 Appendix S4). SBRA on the whole data set produced for RUEann a model with an R2 (.38, Supporting Information Appendix S4) similar Forest RUEann RUEmax to that for temperate forests (see above). For RUEmax, analyses on the type / Site code N. years Mean SD Mean SD entire data set showed only N deposition and LAI having a significant correlation with it (but aridity index presented p = .08; Supporting Cold forest Information Appendix S5) with a clear similarity to the behaviour of DE‐Hai 8 1.38 ± 0.31 1.06 ± 0.40 cold forests (see Table 3). SBRA on the whole data set produced for DE‐Tha 9 1.39 ± 0.21 0.97 ± 0.18 RUEmax a model with an R2 (.23, Supporting Information Appendix DK‐Sor 9 1.59 ± 0.23 1.18 ± 0.45 S5) lower than those obtained for cold and temperate forests sepa‐ FI‐Hyy 12 0.91 ± 0.11 0.81 ± 0.36 rately (see above). FI‐Sod 8 0.04 ± 0.32 0.75 ± 0.24 RUEmax for cold and temperate forests showed significant IT‐Ren 8 1.02 ± 0.22 0.79 ± 0.16 interannual variability (Table 4, SD ±0.24–0.30 gC/MJ for both Temperate forest forest types) but, within each forest type, still slightly lower than BE‐Bra 10 0.96 ± 0.16 0.82 ± 0.40 the spatial variability. Soroe (DK‐Sor) forest showed the highest BE‐Vie 9 1.59 ± 0.20 0.87 ± 0.27 variability in RUEmax (SD ±0.45 gC/MJ, max. 1.66 gC/MJ, min. FR‐Hes 9 1.37 ± 0.27 1.16 ± 0.27 0.31 gC/MJ) whereas Loobos (NL‐Loo) forest showed the lowest variability (SD ±0.09 gC/MJ, max. 0.95 gC/MJ, min. 0.58 gC/MJ). FR‐Pue 8 0.59 ± 0.07 0.51 ± 0.16 For RUE ann , the interannual variability was lower than the interan‐ NL‐Loo 9 1.21 ± 0.23 0.76 ± 0.09 nual variability of RUEmax (Table 4, SD ±0.19–0.23 gC/MJ for both forest types). In general, the interannual variability of RUE ann and RUEmax was weakly related to the examined variables (Table 5). autumn (mainly evapotranspiration, Tmax and precipitation) and in For RUE ann , for five (out of six) cold forests and three (out of five) any case less relevant (e.g., significant at about half of the sites). temperate forests, interannual variability was not related to any For temperate forests, RUE8days appeared to be related to environ‐ variable. Moreover, the other sites only showed significant rela‐ mental factors mainly in summer and autumn (with dependencies tionships with one or two variables per site (variables involved: for both temperature‐ and drought‐related variables at most sites) mainly evapotranspiration and precipitation but also Tmax; see but to be conservative in spring (with only one variable important Table 5). For RUEmax , significant relationships were found only for at one site). three cold forests and one temperate forest. The interannual vari‐ ability of RUEmax was explained mainly by evapotranspiration and precipitation, with some weak relationships (.05 < p < .10) with 4 | D I S CU S S I O N temperature‐related variables (see Table 5). Table 6 and Supporting Information Appendix S8 report the results One of the main findings of this study is that, on average, cold and of the short‐term variability analysis of RUE8days for the whole season temperate forests exhibit very similar RUEann (c. 1.1 gC/MJ) and and for each season separately (i.e., spring, summer, autumn). The vari‐ RUEmax (c. 0.8 gC/MJ) but their relationships with vegetation and ables selected were generally correlated with RUE8days, particularly for environmental drivers differ significantly. Also, the drivers of RUEann cold forests. In fact, for cold forests, RUE8days was significantly cor‐ differ from the drivers of RUEmax within each forest type. More in related with (a) potential evapotranspiration at five sites (out of six), (b) detail, RUEann of cold forests is influenced by mainly variables related Tmean and Tmin at four sites, and with (c) VPD, annual precipitation to temperature (particularly the number of freezing days), whereas and Tmax at about half of the sites. For temperate forests, only three of RUEann of temperate forests by variables related to water status the five sites presented significant correlations between RUE8days and (particularly aridity index). LAI is important for both forest types, the examined variables (Table 6). SBRA showed in general R2 between whereas cloud cover only for temperate forests and N deposition .21 and .61 (Supporting Information Appendix S9). The low R2 of some only for cold forests. On the other hand, RUEmax of cold forests was of the SBRAs may be related to the fact that important variables were related to LAI and N deposition, whereas RUEmax of temperature for‐ not considered and/or because of nonlinear relationships. ests was only weakly related to CO2 concentration. Concerning the When analysed for the three seasons, cold forests presented temporal variability, our study shows that interannual variability of two main results. First, cold forests showed the strongest rela‐ RUEann and RUEmax is not primarily related to environmental factors. tionships between RUE8days and the environmental factors in On the other hand, we showed that short‐term (8 days) variability of summer, for both temperature‐ and drought‐related variables RUE is strongly related to the environmental conditions, particularly and for all sites. Second, the drivers of RUE8days in spring (mainly in summer. This evidence was valid for both cold‐ and temperate VPD, Tmean and Tmin) were different to the drivers of RUE8days in forests. 1662 | TA B L E 5 Results of univariate analysis for interannual variability in annual (RUEann) and maximum radiation use efficiency (RUEmax). The significance level of the linear regression (p) and adjusted R2 are reported, as well as the sign of the regression. Data are reported for six cold and five temperate forest sites. Sites’ full names can be found in Supporting Information Appendix S1 Cold forest Temperate forest DE‐Hai DE‐Tha DK‐Sor FI‐Hyy FI‐Sod IT‐Ren BE‐Bra BE‐Vie FR‐Hes FR‐Pue NL‐Loo (n = 8) (n = 9) (n = 9) (n = 12) (n = 8) (n = 8) (n = 10) (n = 9) (n = 9) (n = 8) (n = 9) 2 2 2 2 2 2 2 2 2 2 RUE Potential predictor p Adj R p Adj. R p Adj. R p Adj. R p Adj. R p Adj. R p Adj. R p Adj. R p Adj. R p Adj. R p Adj. R2 RUEann Precipitation .66 −.13 (‐) .21 .10 (+) .61 −.10 (+) .00** .85 (+) .46 −.06 (‐) .97 −.17 (‐) .23 .07 (‐) .01* .57 (+) .00** .77 (+) .26 .07 (+) .91 −.14 (+) ** Evapotranspiration .57 −.10 (‐) .23 .09 (‐) .47 −.06 (+) .01* .53 (‐) .38 −.02 (‐) .28 .06 (‐) .65 −.09 (‐) .00 .82 (‐) .03* .42 (‐) .34 .01 (‐) .59 −.09 (+) Vapour pressure .68 −.13 (‐) .38 −.02 (‐) .38 −.01 (+) .20 .10 (+) .80 −.15 (‐) .33 .02 (‐) .14 .16 (‐) .56 −.09 (+) .67 −.11 (‐) .50 −.07 (+) .47 −.05 (‐) deficit Tmean .86 −.16 (+) .10 .25 (‐) .29 .04 (+) .68 −.10 (‐) .71 −.14 (‐) .23 .10 (‐) .13 .17 (‐) .28 .04 (‐) .15 .17 (‐) .94 −.17 (+) .78 −.13 (‐) ○ Tmin .88 −.16 (‐) .10 .25 (‐) .32 .02 (+) .92 −.12 .66 −.13 (+) .33 .02 (‐) .09 .24 (‐) .98 −.14 (‐) .35 .00 (‐) .85 −.16 (+) .47 −.06 (‐) Tmax .47 −.06 (+) .35 .00 (‐) .24 .07 (+) .23 .07 (‐) .55 −.10 (‐) .28 .06 (‐) .84 −.12 (‐) .02* .48 (‐) .31 .02 (‐) .57 −.10 (+) .68 −.11 (+) CO2 concentration .33 .02 (‐) .91 −.14 (‐) .93 −.14 (‐) .45 −.04 (+) .52 −.08 (+) .81 −.15 (‐) .53 −.07 (+) .27 .05 (+) .88 −.14 (+) .61 −.11 (+) .95 −.14 (+) RUEmax Precipitation .78 −.15 (‐) .68 −.11 (+) .45 −.05 (+) .01* .56 (+) .80 −.18 (‐) .85 −.16 (+) .81 −.12 (‐) .03* .46 (+) .27 .05 (+) .87 −.16 (+) .82 −.13 (+) ○ ○ Evapotranspiration .03* .48 (‐) .89 −.14 (‐) .59 −.09 (+) .04* .37 (‐) .61 −.13 (+) .07 .35 (‐) .46 −.05 (‐) .08 .29 (‐) .46 −.05 (‐) .27 .06 (‐) .66 −.11 (‐) Vapour pressure .41 −.03 (+) .55 −.08 (‐) .96 −.14 (+) .69 −.10 (+) .76 −.18 (‐) .14 .21(‐) .07○ .27 (‐) .25 .06 (+) .84 −.14 (‐) .95 −.17 (‐) .15 .16 (‐) deficit Tmean .96 −.17 (‐) .61 −.10 (‐) .77 −.13 (+) .39 −.02 (‐) .97 −.20 (+) .07○ .36 (‐) .06○ .30 (‐) .79 −.13 (‐) .28 .04 (‐) .61 −.11 (‐) .21 .10 (‐) ○ Tmin .32 .02 (+) .02* .51 (‐) .73 −.12 (+) .89 −.12 (‐) .91 −.20 (‐) .10 .28 (‐) .06 .29 (‐) .35 .00 (+) .25 .07 (‐) .70 −.14 (‐) .30 .03 (‐) Tmax .70 −.13 (‐) .64 −.11 (+) .76 −.13 (+) .08○ .24 (‐) .64 −.14 (+) .16 .18 (‐) .24 .07 (‐) .17 .15 (‐) .55 −.08 (‐) .87 −.16 (+) .58 −.09 (‐) CO2 concentration .69 −.13 (‐) .76 −.13 (+) .68 −.11 (‐) .50 −.06 (+) .03* .56 (‐) .70 −.14 (+) .62 −.09 (+) .25 .07 (+) .68 −.11 (‐) .19 .14 (+) .69 −.11 (+) ○ Abbreviations and symbols: Tmean = mean temperature; Tmin and Tmax = minimum and maximum air temperature, respectively. = .05 < p < .10; * = .01 < p < .05; ** = .001 < p < .001. (+) = positive linear regression; (‐) = negative linear regression. BALZAROLO et al. BALZAROLO et al. | 1663 TA B L E 6 Results of univariate analysis for short‐term variability in radiation use efficiency (RUE8days) of cold and temperate forests, for the whole growing season and spring, summer and autumn separately. Cells in orange are for correlations with p ≤ .05 and in yellow for correlations with .05 < p ≤ .10. All detailed data (p and adjusted R2) are reported in Supporting Information Appendix S8. Data are reported for six cold and five temperate forest sites. Sites’ full names and descriptions can be found in Supporting Information Appendix S1 Cold forest Temperate forest Potential predictor DE‐Hai DE‐Tha DK‐Sor FI‐Hyy FI‐Sod IT‐Ren BE‐Bra BE‐Vie FR‐Hes FR‐Pue NL‐Loo All seasons Tmean Tmin Tmax Precipitation Vapour pressure deficit Evapotranspiration CO2 concentration SPRING: from mid‐April to mid‐June Tmean Tmin Tmax Precipitation Vapour pressure deficit Evapotranspiration CO2 concentration SUMMER: from mid‐June to mid‐August Tmean Tmin Tmax Precipitation Vapour pressure deficit Evapotranspiration CO2 concentration AUTUMN: from mid‐August to mid‐October Tmean Tmin Tmax Precipitation Vapour pressure deficit Evapotranspiration CO2 concentration Abbreviations: Tmean = mean temperature; Tmin and Tmax = minimum and maximum air temperature, respectively. The fact that (low) temperature plays an important role in the the hydraulic system that supplies leaves with water and nutrients eco‐physiology of trees in the cold zone is not surprising. Subfreezing (e.g., freezing‐induced xylem cavitation) (Jackson, Sperry, & Dawson, temperatures stop photosynthesis, because leaf stomata are forced 2000; Sperry, Nichols, Sullivan, & Eastlack, 1994; Sperry & Sullivan, to close (Waring & Running, 1998). Moreover, negative effects 1992). Our temporal analysis showed that drought‐related variables of freezing on RUE could be related to direct frost damage to the (e.g., VPD) were also important to determine RUE of cold forests but leaves (Marchand, 1996) or to indirect, freezing‐induced damage to mainly in summer. Thus, they should not be neglected for seasonal 1664 | BALZAROLO et al. modelling of RUE but it is also clear that they did not emerge in the account factors found to be important by Wang et al. (2017), such annual analyses of either RUEann or RUEmax. as the ratio of internal leaf CO2 to external CO2, or elevation. On Our results support previous findings that identified drought‐ the other hand, both studies found that temperature and a drought‐ related variables (e.g., annual precipitation, aridity index, longest related variable [Wang et al. (2017) considered VPD, we potential period without rain, VPD) as highly significant determinants of RUE evapotranspiration; Table 6] are crucial determinants of the short‐ for temperate forests. The negative relationship between drought term variability of RUE. and RUE in temperate forests is related to the fact that drought Two general remarks should also be made concerning our study. impacts on stomatal conductance and restricts the photosynthe‐ First, the short‐term temporal analysis clearly revealed differences sis process (e.g., Bracho et al. (2012)). Moreover, in typical drought among sites, with some sites showing stronger relationships be‐ conditions, a canopy receives more direct light and this has a neg‐ tween RUE and the environment than others. This might be the ative impact on RUE (De Boeck & Verbeeck, 2011). Our temporal reason why so many contrasting results have been found about the analysis for temperate forests showed that temperature‐related environmental drivers of RUE dynamics in the past (Table 1). Second, variables are not to be neglected for short‐term modelling of RUE one general difficulty in understanding the effects of environmental in summer, but as for the case of drought‐related variables for cold variables on RUE is finding a good surrogate of each environmen‐ forests, these variables are not important in the annual analyses of tal factor at the global scale and their availability at daily to annual either RUE ann or RUEmax. time‐scales. The accuracy of available meteorological data is of par‐ Both forest types showed relationships with radiation‐related ticular importance for modelling the RUE dynamics of ecosystems variables. The important impact of radiation type (i.e., diffuse ver‐ under stress conditions such as drought. Moreover, note that under sus direct radiation) and radiation interception capacity (i.e., LAI) on drought conditions other specific factors might also play a role as photosynthetic efficiency and RUE of forests is well known. Diffuse determinants of RUE (Garbulsky et al., 2010; Goerner et al., 2009; radiation penetrates deeper in the canopy and does not saturate the Stocker et al., 2018). photosynthetic capacity at leaf level (Gu et al., 2002). Therefore, In summary, our study synthesized existing knowledge on the greater cloud cover improves the photosynthetic efficiency by de‐ determinants of the variability of RUE in cold and temperate forests creasing the denominator of Equation 1 at similar values of GPP. and tested their potential roles as predictor variables considering a Furthermore, more leaves (i.e., higher LAI) correspond to a higher high number of sites in the Northern Hemisphere. We found that, fAPAR. on average, RUEann and RUEmax do not differ markedly between The key role of nutrient availability in affecting plant and cold and temperate forests, but the influence of different veg‐ ecosystem processes is well known. In particular, high N avail‐ etation and environment drivers on RUE does differ significantly ability allows the maintenance of high Rubisco concentration in between the two climatic zones. These findings primarily indicate the leaves, which contributes to a high photosynthetic capacity that global tools using RUE should differentiate their algorithms be‐ (Field, Merino, & Mooney, 1983). However, the role of nutrient tween climate zones. For instance, MODIS GPP might improve if availability in determining RUE was unclear (Table 1). Our study current RUEmax modulators (temperature, light and VPD) become shows no impact of leaf N or site fertility but a positive impact of region‐dependent. Also, our study suggests that the use of envi‐ N deposition. This indicates the relevance of a fertilization effect, ronmental variables only does not suffice to describe the variability which might be more important than actual N pools (leaf and soil of RUE, and therefore that ecological models based on Monteith’s N) and more important in colder (typically nutrient limited) high approach should include new parameters that describe plant char‐ latitude areas. The past uncertainty in understanding the effect acteristics such as LAI. N deposition should also be accounted for of N deposition on RUE might also have been related to the qual‐ in modelling RUE of cold forests. Moreover, our results show that, ity and representativeness of N deposition data. In this work, we within each climatic zone, RUEann and RUEmax have different rela‐ used very recent global simulations for N deposition for the period tionships with environmental and vegetation variables. Therefore, 2004–2006 (Ackerman et al., 2018). equations parameterized for one index cannot be used for the other The effect of CO2 concentration was found to be of some signifi‐ index. Our analysis showed also that RUEmax is less related to envi‐ cance but not a key driver of RUE as found by De Kauwe et al. (2016). ronmental conditions than RUEann. This helps modelling and remote This might be due to the fact that we analysed natural variability sensing applications. In detail, while RUEmax for cold forests de‐ of CO2 concentration (overall 380 ± 8 ppm) whereas De Kauwe et pends on two variables (LAI and N deposition), RUEmax of temperate al. (2016) analysed long‐term FACE forest sites with CO2 concentra‐ forests is not related to any variables. So, the latter might be used as tions ranging from 370 to 550 ppm. A high CO2 concentration may a constant. Finally, concerning the temporal variability, our analysis increase the effect of CO2 on RUE. at a short‐term scale (i.e. 8‐day), indicated that the relationships be‐ Our study is difficult to compare to the one of Wang et al. (2017) tween RUE and predictors differ when the whole growing season or on a universal mechanism driving RUE across biomes, as the stud‐ spring, summer and autumn are considered separately. On the other ies are methodologically very different. For instance, our seasonal hand, interannual variability of RUEann and RUEmax was less related short‐term analysis of RUE variability could have not taken into to the environment. BALZAROLO et al. | 1665 AC K N OW L E D G M E N T S AU T H O R C O N T R I B U T I O N IAJ acknowledges the Belgian Science Policy Office (BELSPO) MB, NV, IAJ and MC conceived the paper; MB performed the analy‐ STEREO III program ECOPROPHET (Improved Ecosystem ses; MB, NV and MC wrote the text; YHF gathered cloud cover data; Productivity Modelling by Innovative Algorithms and Remotely LS gathered soil‐type description; all authors contributed substan‐ Sensed Phenology Indicators) project (SR/00/334). NV and IAJ ac‐ tially to discussions. knowledge the European Research Council Synergy grant ERC‐2013‐ SyG‐610028 IMBALANCE‐P. MC was a Postdoctoral Fellow of the DATA AVA I L A B I L I T Y S TAT E M E N T Research Foundation – Flanders (FWO). YHF was supported by the University of Beijing Normal University, the General Program of A summarized version of the data set is available in Supporting National Nature Science Foundation of China (grant no. 31770516). Information Appendix S1. The full data set can be made available by This work used eddy covariance data acquired by the FLUXNET the leading author upon request. community and in particular by the following networks: AmeriFlux [U.S. Department of Energy, Biological and Environmental Research, ORCID Terrestrial Carbon Program (DE‐FG02‐04ER63917 and DE‐ FG02‐04ER63911)], AfriFlux, AsiaFlux, CarboAfrica, CarboEuropeIP, Manuela Balzarolo https://orcid.org/0000-0002-7888-1501 CarboItaly, CarboMont, ChinaFlux, Fluxnet‐Canada (supported Yongshuo H. Fu https://orcid.org/0000-0002-9761-5292 by CFCAS, NSERC, BIOCAP, Environment Canada and NRCan), GreenGrass, KoFlux, LBA, NECC, OzFlux, TCOS‐Siberia, USCCC. We acknowledge the financial support to the eddy covariance data REFERENCES harmonization provided by CarboEuropeIP, FAO‐GTOS‐TCO, iL‐ Ackerman, D. E., Chen, X., & Millet, D. B. (2018). Global nitrogen depo‐ EAPS, Max Planck Institute for Biogeochemistry, National Science sition (2°×2.5° grid resolution) simulated with GEOS‐Chem for 1984– Foundation, University of Tuscia, Université Laval and Environment 1986, 1994–1996, 2004–2006, and 2014–2016. Retrieved from Canada and U.S. Department of Energy and the database develop‐ the Data Repository for the University of Minnesota. https​://doi. ment and technical support from Berkeley Water Center, Lawrence org/10.13020/​D6KX2R Allen, C. B., Will, R. E., McGarvey, R. C., Coyle, D. R., & Coleman, M. Berkeley National Laboratory, Microsoft Research Science, Oak D. (2005). Radiation‐use efficiency and gas exchange responses to Ridge National Laboratory, University of California – Berkeley, water and nutrient availability in irrigated and fertilized stands of University of Virginia. We gratefully thank all scientists and techni‐ sweetgum and sycamore. Tree Physiology, 25, 191–200. https​://doi. cians who helped collect these data and make them available for anal‐ org/10.1093/treep​hys/25.2.191 Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. (1998). Crop evapotrans‐ ysis. We also acknowledge the authors and repository institutions of: piration‐guidelines for computing crop water requirements. FAO irri‐ Global Forest Database; European Fluxes Database Cluster; ‘Global gation and drainage paper 56. Rome, Italy: FAO. nitrogen deposition (2°×2.5° grid resolution) simulated with GEOS‐ Alton, P., North, P., & Los, S. (2007). The impact of diffuse sunlight on Chem for 2004‐2006' (University of Minnesota); EC‐JRC‐MARS canopy light‐use efficiency, gross photosynthetic product and net data set (© European Union, 2011–2014) created by MeteoConsult ecosystem exchange in three forest biomes. Global Change Biology, 13, 776–787. https​://doi.org/10.1111/j.1365-2486.2007.01316.x based on European Centre for Medium‐Range Weather Forecasts Baldocchi, D. D. (2018). Must we incorporate soil moisture information (ECMWF) model outputs and a reanalysis of ERA‐Interim; Global when applying light use efficiency models with satellite remote Weather Data for SWAT (http://globa​lweat​her.tamu.edu); Land and sensing information? New Phytologist, 218, 1293–1294. https​://doi. Water Development Division, FAO, Rome, for the Digital Soil Map org/10.1111/nph.15176​ Bracho, R., Starr, G., Gholz, H. L., Martin, T. A., Cropper, W. P., & Loescher, of the World (http://www.fao.org/geone​t work​); University of East H. W. (2012). Controls on carbon dynamics by ecosystem structure Anglia Climatic Research Unit (CRU) Time‐Series (TS) Version 3.22 for and climate for southeastern US slash pine plantations. Ecological cloud cover data (http://dx.doi.org/10.5285/18BE2​3F8-D252-482D- Monographs, 82, 101–128. https​://doi.org/10.1890/11-0587.1 8AF9-5D6A2​D40990C); Land Processes Distributed Active Archive Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32. Campbell, G. S., & Norman, J. M. (1998). An introduction to environmental Center (LP DAAC), US Geological Survey (USGS) Earth Resources biophysics. New York, USA: Springer Science & Business Media. Observation and Science (EROS) Center (lpdaac.usgs.gov) for provid‐ Campoe, O. C., Stape, J. L., Albaugh, T. J., Allen, H. L., Fox, T. R., ing MODIS data. We thank Erik Fransen (StatUa Center for Statistics, Rubilar, R., & Binkley, D. (2013). Fertilization and irrigation effects University of Antwerp) for statistic consultancy and Sara Vicca and on tree level aboveground net primary production, light intercep‐ James Weedon for support on data analysis. This project has re‐ tion and light use efficiency in a loblolly pine plantation. Forest Ecology and Management, 288, 43–48. https​://doi.org/10.1016/j. ceived funding from the EU Horizon 2020 Research and Innovation foreco.2012.05.026 programme under a Marie Skłodowska‐Curie grant (INDRO (Remote Chasmer, L., McCaughey, H., Barr, A., Black, A., Shashkov, A., Treitz, P., sensing INdicators for DROught monitoring), grant no. 702717). & Zha, T. (2008). Investigating light‐use efficiency across a jack pine chronosequence during dry and wet years. Tree Physiology, 28, 1395– 1406. https​://doi.org/10.1093/treep​hys/28.9.1395 C O N FL I C T O F I N T E R E S T Cheng, Y.‐B., Zhang, Q., Lyapustin, A. I., Wang, Y., & Middleton, E. M. (2014). Impacts of light use efficiency and fPAR parameter‐ The authors declare no competing financial interests. ization on gross primary production modeling. Agricultural and 1666 | BALZAROLO et al. Forest Meteorology, 189, 187–197. https​://doi.org/10.1016/j.agrfo​ Harris, I., & Jones, P. D. (2014). CRU TS3.22: Climatic Research Unit rmet.2014.01.006 (CRU) Time‐Series (TS) version 3.22 of high resolution gridded Creutzberg, D. (1987) Description of units of the FAO‐UNESCO soil map data of month‐by‐month variation in climate (Jan. 1901‐ Dec. of the world legend (ISRIC Report 1987/01). Wageningen, NL. 2013). NCAS British Atmospheric Data Centre. Retrieved from Retrieved from http://www.isric.org/sites/​all/modul​es/pubdl​cnt/ ht tps​: //catal​o gue.ceda.ac.uk /uuid/4a6d 0​7 1383​9 76a5​f b24b​ pubdl​c nt.php?file=/isric/​w ebdo​c s/docs/ISRIC_Report_1987_01. 5b42e​28cf28f pdf&nxml:id=332 Jackson, R. B., Sperry, J. S., & Dawson, T. E. (2000). Root water up‐ De Boeck, H., & Verbeeck, H. (2011). Drought‐associated changes take and transport: Using physiological processes in global predic‐ in climate and their relevance for ecosystem experiments and tions. Trends in Plant Science, 5, 482–488. https​://doi.org/10.1016/ models. Biogeosciences, 8, 1121–1130. https​://doi.org/10.5194/ S1360-1385(00)01766-0 bg-8-1121-2011 Jenkins, J. P., Richardson, A. D., Braswell, B. H., Ollinger, S. V., Hollinger, De Kauwe, M. G., Keenan, T. F., Medlyn, B. E., Prentice, I. C., & Terrer, D. Y., & Smith, M. L. (2007). Refining light‐use efficiency calcu‐ C. (2016). Satellite based estimates underestimate the effect of CO2 lations for a deciduous forest canopy using simultaneous tower‐ fertilization on net primary productivity. Nature Climate Change, 6, based carbon flux and radiometric measurements. Agricultural and 892–893. Forest Meteorology, 143, 64–79. https​://doi.org/10.1016/j.agrfo​ Dile, Y. T., & Srinivasan, R. (2014). Evaluation of CFSR climate data for rmet.2006.11.008 hydrologic prediction in data‐scarce watersheds: An application in Kergoat, L., Lafont, S., Arneth, A., Le Dantec, V., & Saugier, B. (2008). the Blue Nile River Basin. Journal of the American Water Resources Nitrogen controls plant canopy light‐use efficiency in temperate and Association, 50, 1226–1241. https​://doi.org/10.1111/jawr.12182​ boreal ecosystems. Journal of Geophysical Research: Biogeosciences, FAO. (2007). Digital soil map of the world version 3.6. Land and water de‐ 113, G04017. https​://doi.org/10.1029/2007J​G 000676 velopment division. Rome, Italy: FAO. Retrieved from http://www. King, D. A., Turner, D. P., & Ritts, W. D. (2011). Parameterization of a diag‐ fao.org/geone​t work​?uuxml​:id=446ed​4 30-8383-11db-b9b2-000d9​ nostic carbon cycle model for continental scale application. Remote 39bc5d8 Sensing of Environment, 115, 1653–1664. https​://doi.org/10.1016/j. Fernández‐Martínez, M., Vicca, S., Janssens, I. A., Luyssaert, S., Campioli, rse.2011.02.024 M., Sardans, J., … Peñuelas, J. (2014). Spatial variability and con‐ Liaw, A., & Wiener, M. (2002). Classification and regression by random‐ trols over biomass stocks, carbon fluxes, and resource‐use effi‐ Forest. R News, 2, 18–22. ciencies across forest ecosystems. Trees, 28, 597–611. https​://doi. Luyssaert, S., Inglima, I., Jung, M., Richardson, A. D., Reichstein, M., org/10.1007/s00468-013-0975-9 Papale, D., … Janssens, I. A. (2007). CO2 balance of boreal, temperate, Field, C., Merino, J., & Mooney, H. A. (1983). Compromises between and tropical forests derived from a global database. Global Change water‐use efficiency and nitrogen‐use efficiency in five species Biology, 13, 2509–2537. of California evergreens. Oecologia, 60, 384–389. https​://doi. Mäkelä, A., Pulkkinen, M., Kolari, P., Lagergren, F., Berbigier, P., Lindroth, org/10.1007/BF003​76856​ A., … Hari, P. (2008). Developing an empirical model of stand GPP Field, C. B., Randerson, J. T., & Malmstrom, C. M. (1995). Global net pri‐ with the LUE approach: Analysis of eddy covariance data at five con‐ mary production: Combining ecology and remote sensing. Remote trasting conifer sites in Europe. Global Change Biology, 14, 92–108. Sensing of Environment, 51, 74–88. Marchand, P. (1996). Life in the cold: An introduction to winter ecology. Fuka, D. R., Walter, M. T., MacAlister, C., Degaetano, A. T., Steenhuis, Lebanon, NH: University Press of New England. T. S., & Easton, Z. M. (2014). Using the Climate Forecast System McCallum, I., Wagner, W., Schmullius, C., Shvidenko, A., Obersteiner, M., Reanalysis as weather input data for watershed models. Hydrological Fritz, S., & Nilsson, S. (2009). Satellite‐based terrestrial production Processes, 28, 5613–5623. https​://doi.org/10.1002/hyp.10073​ efficiency modeling. Carbon Balance and Management, 4. https​://doi. Garbulsky, M. F., Penuelas, J., Papale, D., Ardo, J., Goulden, M. L., Kiely, org/10.1186/1750-0680-4-8 G., … Filella, I. (2010). Patterns and controls of the variability of radi‐ Monteith, J. (1972). Solar radiation and productivity in tropical eco‐ ation use efficiency and primary productivity across terrestrial eco‐ systems. Journal of Applied Ecology, 9, 747–766. https​://doi. systems. Global Ecology and Biogeography, 19, 253–267. https​://doi. org/10.2307/2401901 org/10.1111/j.1466-8238.2009.00504.x Myneni, R. B., Hoffman, S., Knyazikhin, Y., Privette, J. L., Glassy, J., Tian, Gitelson, A. A., & Gamon, J. A. (2015). The need for a common basis Y., … Running, S. W. (2002). Global products of vegetation leaf area for defining light‐use efficiency: Implications for productivity esti‐ and fraction absorbed PAR from year one of MODIS data. Remote mation. Remote Sensing of Environment, 156, 196–201. https​://doi. Sensing of Environment, 83, 214–231. https​://doi.org/10.1016/ org/10.1016/j.rse.2014.09.017 S0034-4257(02)00074-3 Goerner, A., Reichstein, M., & Rambal, S. (2009). Tracking seasonal Myneni, R. B., Los, S. O., & Asrar, G. (1995). Potential gross primary drought effects on ecosystemlight use efficiency with satellitebased productivity of terrestrial vegetation from 1982–1990. Geophysical PRI in a Mediterranean forest. Remote Sensing of Environment, 113, Research Letters, 22, 2617–2620. https​://doi.org/10.1029/95GL0​ 1101–1111. https​://doi.org/10.1016/j.rse.2009.02.001 2562 Goudriaan, J., & Van Laar, H. H. (1994). Modelling potential crop growth Ollinger, S. V., Richardson, A. D., Martin, M. E., Hollinger, D. Y., Frolking, processes. Textbook with exercises. Dordrecht, The Netherlands: S. E., Reich, P. B., … Schmid, H. P. (2008). Canopy nitrogen, carbon Kluwer Academic Publishers. assimilation, and albedo in temperate and boreal forests: Functional Grace, J., Nichol, C., Disney, M., Lewis, P., Quaife, T., & Bowyer, P. (2007). relations and potential climate feedbacks. Proceedings of the Can we measure terrestrial photosynthesis from space directly, using National Academy of Sciences USA, 105, 19336–19341. https​://doi. spectral reflectance and fluorescence? Global Change Biology, 13, org/10.1073/pnas.08100​21105​ 1484–1497. https​://doi.org/10.1111/j.1365-2486.2007.01352.x Peel, M. C., Finlayson, B. L., & McMahon, T. A. (2007). Updated world Gu, L., Baldocchi, D., Verma, S. B., Black, T., Vesala, T., Falge, E. M., map of the Koppen‐Geiger climate classification. Hydrology and Earth & Dowty, P. R. (2002). Advantages of diffuse radiation for ter‐ System Sciences, 11, 1633–1644. restrial ecosystem productivity. Journal of Geophysical Research: Peltoniemi, M., Pulkkinen, M., Kolari, P., Duursma, R. A., Montagnani, Atmospheres, 107. L., Wharton, S., … Makela, A. (2012). Does canopy mean nitrogen Hunt, E. R. Jr (1994). Relationship between woody biomass and PAR con‐ concentration explain variation in canopy light use efficiency across version efficiency for estimating net primary production from NDVI. 14 contrasting forest sites? Tree Physiology, 32, 200–218. https​://doi. International Journal of Remote Sensing, 15, 1725–1730. org/10.1093/treep​hys/tpr140 BALZAROLO et al. | 1667 Penman, H. L. (1948). Natural evaporation from open water, bare soil and dioxide uptake by plants. Nature Plants, 3(9), 734–741. https​://doi. grass. Proceedings of the Royal Society of London Series A, Mathematical org/10.1038/s41477-017-0006-8 and Physical Sciences, 193, 120–145. Waring, R. H., … Running, S. W. (1998). Forest ecosystems, analysis at mul‐ Prince, S. D., & Goward, S. N. (1995). Global primary production: A re‐ tiple scales (2nd ed.). USA: Academic Press. mote sensing approach. Journal of Biogeography, 22, 815–835. Yuan, W., Cai, W., Xia, J., Chen, J., Liu, S., Dong, W., … Wohlfahrt, G. R Core Team. (2015). R: A language and environment for statistical com‐ (2014). Global comparison of light use efficiency models for simu‐ puting. Vienna, Austria: R Foundation for Statistical Computing. lating terrestrial vegetation gross primary production based on the Retrieved from http://www.r-proje​c t.org/ LaThuile database. Agricultural and Forest Meteorology, 192–193, 108– Running, S. W., Nemani, R. R., Heinsch, F. A., Zhao, M., Reeves, M., & 120. https​://doi.org/10.1016/j.agrfo​rmet.2014.03.007 Hashimoto, H. (2004). A continuous satellite‐derived measure of Yuan, W., Liu, S., Zhou, G., Zhou, G., Tieszen, L. L., Baldocchi, D., … Wofsy, global terrestrial primary production. BioScience, 54, 547–560. S. C. (2007). Deriving a light use efficiency model from eddy covari‐ https​: //doi.org/10.1641/0 0 06-3568(20 04) 054[0547:ACSMO​ ance flux data for predicting daily gross primary production across G]2.0.CO;2 biomes. Agricultural and Forest Meteorology, 143, 189–207. https​:// Runyon, J., Waring, R. H., Goward, S. N., & Welles, J. M. (1994). doi.org/10.1016/j.agrfo​rmet.2006.12.001 Environmental limits on net primary production and light‐use effi‐ Zhao, M., Running, S. W., & Nemani, R. R. (2006). Sensitivity of moderate ciency across the Oregon transect. Ecological Applications, 4, 226–237. resolution imaging spectroradiometer (MODIS) terrestrial primary Ryan, M. G., Linder, S., Vose, J. M., & Hubbard, R. M. (1994). Dark production to the accuracy of meteorological reanalyses. Journal Respiration of Pines. Ecological Bulletins, 43, 50–53. of Geophysical Research‐Biogeosciences, 111, G01002. https​://doi. Schwalm, C. R., Black, T. A., Arniro, B. D., Arain, M. A., Barr, A. G., org/10.1029/2004J​G 000004 Bourque, C. P. A., … Wofsy, S. C. (2006). Photosynthetic light use ef‐ ficiency of three biomes across an east‐west continental‐scale tran‐ sect in Canada. Agricultural and Forest Meteorology, 140, 269–286. B I O S K E TC H https​://doi.org/10.1016/j.agrfo​rmet.2006.06.010 Sperry, J. S., Nichols, K. L., Sullivan, J. E., & Eastlack, S. E. (1994). Xylem Manuela Balzarolo is currently a Marie Curie Postdoctoral embolism in ring‐porous, diffuse‐porous, and coniferous trees of Research Fellow at Ecological and Forestry Applications Research northern Utah and interior Alaska. Ecology, 75, 1736–1752. https​:// doi.org/10.2307/1939633 Centre (CREAF) in Spain. Her research focuses on the analysis of Sperry, J. S., & Sullivan, J. E. (1992). Xylem embolism in response to the impact of climate on ecosystem productivity by using remote freeze‐thaw cycles and water stress in ring‐porous, diffuse‐porous, sensing data. and conifer species. Plant Physiology, 100, 605–613. https​://doi. org/10.1104/pp.100.2.605 Stocker, B. D., Zscheischler, J., Keenan, T. F., Prentice, I. C., Peñuelas, S U P P O R T I N G I N FO R M AT I O N J., & Seneviratne, S. I. (2018). Quantifying soil moisture impacts on Additional supporting information may be found online in the light use efficiency across biomes. New Phytologist, 218, 1430–1449. https​://doi.org/10.1111/nph.15123​ Supporting Information section at the end of the article. Turner, D. P., Urbanski, S., Bremer, D., Wofsy, S. C., Meyers, T., Gower, S. T., & Gregory, M. (2003). A cross‐biome comparison of daily light use efficiency for gross primary production. Global Change Biology, 9, How to cite this article: Balzarolo M, Valdameri N, Fu YH, 383–395. https​://doi.org/10.1046/j.1365-2486.2003.00573.x Schepers L, Janssens IA, Campioli M. Different determinants Wang, H., Prentice, I., & Davis, T. (2014). Biophsyical constraints on gross of radiation use efficiency in cold and temperate primary production by the terrestrial biosphere. Biogeosciences, 11, forests. Global Ecol Biogeogr. 2019;28:1649–1667. 5987–6001. https​://doi.org/10.5194/bg-11-5987-2014 Wang, H., Prentice, I. C., Keenan, T. F., Davis, T. W., Wright, I. J., Cornwell, https​://doi.org/10.1111/geb.12985​ W. K., … Peng, C. (2017). Towards a universal model for carbon