Biological and climatic controls on leaf litter decomposition

Introduction Conclusions References


Introduction
In non-fertilized ecosystems, such as typical grasslands and forests, the nitrogen (N) flux in litter is the dominant N input into the soil (Berg and McClaugherty, 2008). The total amount of N mineralized in the soil is the product of the total litter mass production intrinsic characteristics of litter also play an important role in the decomposition rate as leaf litter varies largely in the mineral composition (Berg and McClaugherty, 2008), content of toxic chemicals such as terpenoids and alkaloids that are synthesized to protect against herbivory but also inhibiting soil microbes (Ormeño et al., 2009), anatomical traits like leaf mass per area, and mechanical characteristics like leaf tensile strength 10 (Cornelissen and Thompson, 1997), resistance to fracture (Wright and Illius, 1995), and leaf toughness (Gallardo and Merino, 1993).
Many efforts have been made to model the carbon (C) and N release from decomposing litter across different climates (Bonan et al., 2013;Liski et al., 2005;Zhang et al., 2010). Models are needed to predict future levels of soil N availability and the turnover 15 rate. Moreover, in Europe, climate change is overall expected to increase temperature, and reduce precipitation in southern countries and increase in northern countries (Jacob et al., 2014). These models are urgently needed to predict litter decomposition changes in globally changing climates.
Gaining insight into future climate effects on litter decomposition requires solving 20 biological and climatic controls on the decomposition. Measuring litter decomposition across climatic transects is a technique which allows to quantify the response of litter decomposition traits in relation to the specific climatic variations along the transect (Johansson et al., 1995). On the other hand, to measure the implications of litter substrate characteristics on the decomposition, one can compare the decomposition process 25 throughout litter with different characteristics or origin in a specific climate (Gallardo and Merino, 1993). In this article, we present a combination of both experimental approaches to study the implications of both, the climatic and the litter substrate characteristics, on the decomposition process. We have carried out litter transplantation ex-Introduction periments to study litter decomposition across forest and grassland ecosystems across warm temperate to boreal Europe, with the major aim to separate the biological and climatic controls on liter decomposition. The specific aims of the study were the following: (1) to assess the actual leaf litter decomposition rates and the C and N amounts remaining in litter, (2) to study these as a function of the climatic characteristics and 5 litter quality, and (3) to generate a simple data-based model to predict the litter mass and litter C and N contents remaining after increasing time-steps of decomposition. The European continent includes a large range ecosystems which differ in mean annual temperatures and cumulative annual precipitation. We were especially interested in comparing temperate sites with more northern sites which would present seasons 10 with mean air temperature below 0 • C, because we hypothesize that litter decomposition can be slowed down by freezing temperatures and lack of liquid water. Therefore, one of our goals is to make the model valid for these sites altogether.
2 Material and methods 2.1 Study sites 15 The study was conducted at six sites of the NitroEurope Integrated project (2006-2011 -http://www.nitroeurope.eu/). The sites covered various climates and ecosystems representative of the European continent and were each dominated by a single tree or grass species. The forest sites are Hyytiälä in Finland Portillo-Estrada et al., 2013), Männikjärve in Estonia (Carter et al., 2012;Portsmuth et

Experimental design and litter collection
We used the litterbag method (Bocock and Gilbert, 1957) and carried out reciprocal litter transplantation experiments to study the decomposition process from two perspectives: as an effect of intrinsic litter characteristics and as an effect of environmental and soil microbiological conditions. Foliage litter produced by the dominant species of each 5 site was shipped to all other similar ecosystem sites for decomposition under a different environmental condition to the original. In short, the litter samples were let to decompose at each site and samples were collected at regular intervals over the period of one year, after which the litter mass loss and C and N contents were analyzed. At the forest sites dominated by evergreen conifers (Hyytiälä, Männikjärve and Speul-10 derbos), the senescent litter material was collected in litter traps placed above the forest floor and harvested once every month throughout the year 2008. At the deciduous forest site Sorø, the litter collection was done analogously, but only during the litterfall period between September-November 2008. At the grassland sites, current-year grass litter was harvested in late autumn by clipping the dead leaves at the base. In all cases, 15 the litter was air-dried at room temperature and mixed every second day to avoid the onset of decomposition of non-aerated wet litter. Once the constant mass was reached, the litter was stored in air-tight bags until the start of the experiment. All the litter collected was shipped to the same lab (Estonian University of Life Sciences) where the litter corresponding to leaves and needles was separated from the other litter fractions (e.g. cones, bark, twigs, etc.). All the leaf litter belonging to the same site was mixed together to create a standard mix of litter per site. This was done to avoid a bias in the decomposition rates due to temporal differences in litter C and N contents occurring throughout the year for conifers as showed by Portillo-Estrada et al. (2013)  Flat-shaped litter bags, made out of nylon screen (1 mm mesh size, 15 cm × 15 cm), were filled with 5.5 ± 0.01 g of air dry (48 h oven-drying at 60 • C) leaf material. The mesh size was considered small enough to prevent biomass loss through the mesh (for the conifer Douglas fir (Pseudotsuga menziesii) litter bags, a double layer of tissue was used to minimize the risk of losing leaf needles through the mesh), yet large 5 enough to permit aerobic activity and entry of small soil animals (though excluding earthworms). A color label was inserted in the litter bags to identify their original content (plant species and site origin) during the decomposition.

Litter transplantation and decomposition
Leaf litter from the four forest sites was shipped to all four forest sites for decomposition, 10 and leaf litter from the grassland sites was sent to the two grassland sites. Altogether, a total of 288 tree litter bags was used for the decomposition experiments at forest sites. In all forest decomposition sites, the litter was sampled at five sampling times along the year. At each date, three replicate bags with litter from each site of litter origin were collected at each site of destination (5 × 3 × 4 × 4 = 240 litter bags). Additionally, 15 in Hyytiälä, litter bags were collected at four sampling times during the first month of decomposition, and again, three replicate bags from the four forest sites of origin were collected (4 × 3 × 1 × 4 = 48 litter bags).
For the grass litter, 60 litter bags were prepared, corresponding to: five sampling times along the year at the two decomposition grassland sites, where three replicate 20 bags with litter originated at the two grassland sites were collected (5 × 3 × 2 × 2 = 60 litter bags).
Immediately before installing the litter bags at the sites, the litter bags were moistened by spraying them with deionized water. After moistening, the bags were placed on the topsoil for decomposition. Each bag was fastened to a stainless steel or PVC 25 stick with a nylon thread. The stick was pushed into the soil, in order to keep the bags in place. At each site, the litter bags were installed in the autumn at a representative day at or close to peak litterfall for forest sites and peak leaf die-off for grassland sites 18059 Introduction  Table 1 for dates). Thus, the decomposition period of all the replicate litter bags within a site begun on the same date. The mass remaining after the specified periods of decomposition was measured by randomly harvesting three replicate litter bags of each litter type of the same origin, later, the mass of the three replicates was averaged. The litter bags were collected 5 on days 2, 5, 9 and 16 after the start of the decomposition treatment in Hyytiälä. In all sites, the bags were also collected approximately in 1, 2, 3, 6 and 12 months after the start. The litter bags were transported to the lab, where the remaining litter mass was oven-dried at 60 • C for 48 h and weighed. The dried litter samples were then sent to the laboratory at the Estonian University of Life Sciences where the samples were 10 post-processed and analyzed for C and N content and leaf area.

Analysis of carbon and nitrogen content and leaf traits
The content of each litter bag was ground to a fine powder and the total C and N content per dry mass were determined by a Vario MAX CNS elemental analyzer (Elementar Analysensysteme GmbH, Hanau, Germany) to the nearest 0.01 %. The litter N content 15 at different times since the start of decomposition and the litter N content at the end of the first year of decomposition (N f ) were normalized with respect to the initial litter content (N 0 ) to estimate the relative N loss through the decomposition treatment. A subsample of the initial (not subjected to decomposition) dried litter mixture (0.5-1 g) was weighed, then the leaf samples were carefully spread over a A4 flatbed scan-20 ner glass avoiding overlap between leaves, and the leaves were scanned at 300 dpi. The RGB color image of leaves was processed to estimate the projected litter specific leaf area [m 2 kg −1 ] as described by Portillo-Estrada et al. (2015). The total specific leaf area was estimated from the projected specific leaf area by considering different leaf section shapes for different leaf types: the section of Festuca pseudovina was approxi-25 mated to a circle; the section of Pinus sylvestris leaves were considered as hemicircles; Fagus sylvatica and Lolium perenne leaves were considered flat, thus the projected area was multiplied by two; and for Pseudotsuga menziesii, the projected specific leaf 18060 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | area was multiplied by a factor of 2.3 obtained from measurements of needle circumference to width ratio from the leaf cross-sections under a light microscope. The total specific leaf area represented the maximum leaf surface exposed to microorganisms' attack during decomposition. 5 The annual decomposition rate constant, k [year −1 ] (Olson, 1963) was calculated by fitting the fraction of litter mass remaining vs. time of sampling relationships according to the equation:

Litter decomposition rate
where m t is the remaining mass of litter after time t [year], and m 0 is the original mass of litter. Six log-transformed data points (five sampling times plus the initial litter mass) corresponding to the average litter mass of the three replicate bags harvested at each decomposition time step were used in each case. A pairwise comparison (Holm-Sidak at a significance level of P < 0.05) test was used to find differences in k between forest litter types across the decomposition sites. 15 Litter turnover rate [year] was estimated as the inverse of k (Feng, 2009).

Meteorological data
Data on air temperature at 0.5 to 4 m height and soil temperature at 2 cm depth, precipitation, air relative humidity and soil water content at 6 cm depth were retrieved from the NitroEurope database (Owen et al., 2011). The retrieved 30 min average air and soil 20 temperature data were averaged daily and a mean annual air (T a ) and soil (T s ) temperatures were calculated for each site. As the sites spanned over a wide climatic gradient (Table 1), two additional variables for cumulative temperature were created, one for soil (T c, s > 0 ) and other for air temperature (T c, a > 0 ). The characteristics were computed by summing up the • C of days of which daily average temperature was above 0 • C from BGD 12,2015 Biological and climatic controls on leaf litter decomposition M. Portillo-Estrada et al. the start date of the decomposition until the date of the litter collection. These excluded the time periods when water was frozen, and better characterized the control of temperature on the decomposition process. This suggestion was tested in the following analysis. 5 In order to statistically predict the percentage of litter mass remaining from the initial litter mass (M r ) and the C and N contents in litter relative to initial values (C r and N r ) at a certain decomposition time, we generated linear mixed effect models including all meteorological parameters and their two-way interactions that individually presented high explanatory power (Pearson correlation coefficient) with the key dependent vari-10 ables: cumulative air and soil temperature (T c, s > 0 and T c, a > 0 ), cumulative precipitation, air relative humidity, and soil water content. Land use was included as a two-level categorical factor (forests and grasslands) in all models as well, and M r was used as an additional independent variable in the models of C and N. Litter origin was used as a random factor in all models. 15 Minimum value of the Akaike information criterion (AIC-value) was used as the criterion for choosing the best model. AIC is a measure of the relative quality of a statistical model for a given set of data, and models with an AIC value less than different by a value of 5 were considered equivalent. We ended up with relatively simple models for the four studied characteristics (M r , C r , N r and k rate). Model selection was done

Litter mass loss during decomposition
The decomposition during the first 100 days was faster than that in the remaining period in all sites and for all types of litter ( Fig. 1). Analysis of the early-stage decomposition at Hyytiälä further indicated that the mass loss rate was most pronounced during the 5 first days of decomposition, 6 to 12 % mass loss during the first two days (Fig. 1g). This rapid loss was followed by slowing down and a certain increase in the litter mass in the subsequent days such that in one month, 87-92 % of initial mass was remaining ( Fig. 1g). After three months of decomposition, we identified a general decrease in the mass loss rate in all the forest sites ( Fig. 1a-d) and grassland sites ( Fig. 1e and f).
During the first days, there was a drop followed by a constant mass over the period of 100-200 days corresponding to the winter period and more pronounced where a snow cover was present ( Fig. 1c and d).
The tree litter remaining mass after one year of decomposition differed between the litters of different origin when decomposing in the same site. Nevertheless, a trend of higher remaining mass in the Northern sites Hyytiälä (mean±SE value was 71.0 ± 2.7 %) and Männikjärve (69.1 ± 1.8 %) was observed if compared to the more Southern sites Sorø (61.5 ± 2.4 %) and Speulderbos (56.9 ± 4.6 %) ( Fig. 1a-d).
In more detail, the decomposition of conifer litter followed a similar trend in all forest sites, characterized by a greater mass loss than that for the broadleaved beech litter.

20
This difference was more pronounced at the sites with a higher mean annual air temperature (Table 1), with Speulderbos being the site where the remaining mass differed most between conifer and broadleaved litter types. Regarding the conifer litter, Douglas fir (Pseudotsuga menziesii) litter decomposed faster in the first months than Pinus sylvestris litter in Sorø, Männikjärve and Hyytiälä ( Fig. 1b-d). The remaining mass of 25 Pinus sylvestris over the decomposition period did not differ (P = 0.392; paired t test) between the litter originated at Hyytiälä and Männikjärve. Also there was no difference (P = 0.669; paired t test) in the remaining mass at the end of the decomposition. The temporal dynamics of the remaining mass of grass litter was very similar for both types of litter ( Fig. 1e and f). After one year of decomposition, the remaining mass of litter was substantially smaller at Easter Bush (mean value 19.8 %) than at Bugac (46.6 %).
The average standard error for the three replicate litter bags of the litter mass re-5 maining after one year was 0.7 % across all sites. Within the given site, the litter mass remaining after one year since the start of the decomposition did not show differences among leaf litters with different origin (One way repeated measures ANOVA test, P > 0.05; Fig. 1). Thus, the decomposition showed a marked dependency on the decomposition site characteristics, rather than on litter type. As an exception, mass 10 loss of Fagus sylvatica litter from Sorø decomposed in Speulderbos was slower than that of conifer litters (Holm-Sidak pairwise multiple comparison test, P < 0.05; Fig. 1a).
The values of the decomposition rate constant, k, were calculated as the slope of a linear fit (N = 6) using log-transformed data of remaining litter mass (Eq. 1). The Pearson correlation (r) coefficient across all species and sites was very high (average 15 ±SE of 0.940 ± 0.010, P < 0.05 in all cases). Decomposition k rate was negatively correlated to the total specific leaf area in tree species (r 2 = 0.38; P = 0.011), but did not correlate to the initial N content (r 2 = 0.021, P = 0.59).

Relationships between litter decomposition rates and site climatic characteristics
20 T a and P for different decomposition sites were positively correlated (r 2 = 0.57), although marginally significant (P = 0.08 and N = 6). Cumulative air temperature (T c, a > 0 ) and cumulative precipitation (P c ) measured at each sampling time along the year were positively correlated across the sites (r 2 = 0.91; P < 0.0001).
Collectively, the remaining litter mass at different stages of decomposition was nega-25 tively correlated with T a > 0 and P in forest ( Fig. 2a and b) and grassland ( Fig. 2c and  The decomposition rate constants, k, of all forest decomposition experiments together correlated positively with the mean annual temperature (T a ) of the decomposition site (r 2 = 0.45, P = 0.0043; Fig. 3a). The correlation was high for each individual litter type: r 2 = 0.99 (P = 0.0065) for Pinus sylvestris (Hyytiälä), r 2 = 0.80 (P = 0.10) for Pinus sylvestris (Männikjärve), r 2 = 0.91 (P = 0.045) for Fagus sylvatica (Sorø), and 5 r 2 = 0.94 (P = 0.029) for Pseudotsuga menziesii. The value of k was also positively correlated with the site mean annual precipitation (P ) (Fig. 3b). Analogously, the higher T a and P , the more litter mass was lost after one year of decomposition ( Fig. 3c and d).
As a consequence of the correlations of mass loss and k with site climatic variables, the estimated litter turnover time was negatively correlated with T a (Fig. 3e) and P (Fig. 3f).

10
The values of k at each site were lower for the broad-leaved Fagus sylvatica (Sorø) litter compared to the conifer Pinus sylvestris (Hyytiälä) (P = 0.001) and Pinus sylvestris (Männikjärve) (P = 0.002) litter types, and marginally significantly lower from the k values for Pseudotsuga menziesii (P = 0.060). Similarly, the pairwise tests showed that the litter mass loss after one year of decomposition ( Fig. 3c and d) and 15 estimated turnover time ( Fig. 3e and f) calculated for each sited depended on the litter type, showing statistical differences between the broadleaved Fagus sylvatica (Sorø) litter and the other three conifer litter types, with no differences between conifer litters.

Litter carbon and nitrogen contents through decomposition
The final to initial N content ratio (N f : N 0 ) in the forest litter showed a positive trend if 20 plotted against site's T a and P ( Fig. 3g and h) such that at warmer and more humid decomposition sites, litter N content increased more than at colder and drier sites.
Analogously to litter mass, C and N contents were expressed relative to the initial level at the beginning of the decomposition in order to compare the dynamics along different litter types (Fig. 4). Pooling all the decomposition data together, forest litter 25 C content non-linearly increased during the decomposition process (Fig. 4a), in contrast with grassland litter, that remained constant through the decomposition period (Fig. 4b) forest and grassland litters ( Fig. 4c and d), although during the first 10 days of decomposition it remained constant for the four forest litter types at Hyytiälä (Fig. 4c inset). Therefore, the C : N ratio steadily decreased during the first year for both types of litter ( Fig. 4e and f). The litter N content after different decomposition periods relative to the initial value 5 was positively correlated with the cumulative litter mass loss across all decomposition experiments (Fig. 5). However, there was a slight decrease of litter N content during the first period of decomposition (at around 10 % of cumulative mass loss) (Fig. 5).

Results of the statistical modeling analysis
The best model for M r contained two independent cumulative meteorological variables 10 for air temperature and precipitation and the site type factor (grassland or forest), while the models for both C and N were strongest with only remaining litter mass, air temperature and land-use. The percentage of remaining litter mass relative to the initial value (M r ) at forest sites was calculated as and for grassland sites as Where P c is the cumulative precipitation [mm] and T c, a > 0 the cumulative air temperature [ • C] on days where daily average temperature was above 0 • C.

20
The percentage of carbon content in litter relative to the initial value (C r ) at forest sites was calculated as and for grassland sites as The percentage of nitrogen content in litter relative to the initial value (N r ) at forest sites was calculated as and for grassland sites as In addition, the decomposition k rate was calculated by a linear model (r 2 = 0.96; P < 0.0001) as a function of site's mean annual air temperature accounting days with daily average above 0 • C (T a > 0 ), and leaf litter total specific leaf area (S LA ) as and for grassland sites (r 2 = 0.94; P = 0.029) as, The P values as well as individual standard errors of the modeled parameters for each equation can be seen in Table 2. 15 Decomposition experiments usually do not focus on the very first days of decomposition (e.g. Vestgarden, 2001)  loss rate occurs. Our study confirms with experimental data that the mass loss measured at Hyytiälä after one month of decomposition is mainly due to the high mass loss occurring in the very first days, being the absolute maximum rate of the decomposition during the whole period. This few-days period, known as the leaching phase, is driven by the loose of water extractable compounds that physically leak the sample 5 (Berg and Laskowski, 2005). Additionally, we observed variations in the remaining litter mass including mass increases relative to the previous sampling during the first month of decomposition (Fig. 1g). This could be related to the invasion of microorganisms: mainly fungal mycelia and microbes (Dighton, 2007). In other studies these variations in the remaining litter mass were shown in later periods than the first month (Liu et al.,10 2015; Gallardo and Merino, 1993), but in few they occurred during the first days; this is yet another motivation to measure the decomposition process during the first days in experiments, focusing on the dynamics of microbial activity and colonization of the litter substrate. After three months of decomposition, the litter mass loss rate decreased in many 15 cases generating a plateau shape at around 100 days of decomposition as observed in Fig. 1. This was also found by other authors (Liu et al., 2015;Zhang et al., 2014). This was not noticeable in the litter decomposed in Männikjärve (Fig. 1c) because the third sampling was not done during the snow cover period but after it. We speculate that the dynamic of the litter remaining mass could have been similar to the one ob-20 served at Hyytiälä (Fig. 1d), revealing also a noticeable decrease in the decomposition rate during the winter after three months of decomposition. We theorize that the slower decomposition rate phase was generated by the combination of the following factors: firstly, this period coincided with the winter period, where lower air and soil temperature and the presence of a snow cover or the lack of liquid water in some sites reduced 25 the decomposition rate. To corroborate this hypothesis, we found that the remaining biomass and the input of energy to the system estimated by the cumulative air temperature were correlated (Fig. 2a and c). In addition, the decomposition rate increased after the winter period. Secondly, the decomposition usually begins by the more di-  , 2005), generating a faster decomposition rate during the first months. Consequently, after the initial leaching phase, when the substrate is less decomposable, the leaf litter mass loss rate slows down, and collectively with the winter effect creating a plateau. 5 In most of the cases, the remaining litter mass did not depend on the litter type, being statistically similar during the decomposition for each site. The exception was Fagus sylvatica litter from Sorø when decomposed in Speulderbos. In this case, after the pairwise comparison, two clearly distinct groups were identified, coinciding with the different nature of the leaf litter: the remaining mass of conifer litter differed with 10 the broad-leaved deciduous leaf litter. The decomposition of grass litter types showed a strong influence by the decomposition sites' climatic characteristics and not between litter types. This was noticeable in the high similarity of the remaining litter mass dynamic of the grass litter when decomposing in the same site, as well as by the similar values achieved of remaining litter mass after one year of decomposition.

Litter carbon and nitrogen content during decomposition
Nitrogen is released from leaf litter during decomposition firstly due to leaching and secondly because it is consumed as a substrate by decomposing organisms. Berg and Laskowski (2005) showed that the content of N in the litter sample increases as decomposition goes on. They argued that litter is colonized by decomposing organisms 20 and since N is usually a limiting nutrient to these, it may actively been brought into the decomposing leaf through ingrowing fungal mycelia. As a result, the N content in the whole sample (including the litter substrate and the decomposers) increased. Our study corroborates the positive trend of N content over decomposition time (Fig. 4c and d) and cumulative mass loss (Fig. 5). These results make the current year litter layer an 25 important sink of N during the first year of decomposition, being richer in N as climate is warmer and wetter ( Fig. 3g and h). This can also have implications in the N turnover in future climate scenarios in the frame of a global change, since T a and P is predicted to 18069 Introduction increase during the present century for the Atlantic to boreal European climates, where our forest study sites are found (Jacob et al., 2014). The litter C content during the decomposition followed a different dynamic between tree and grass litter types: the decomposition had almost no effect (r 2 = 0.008) on the C content in grass litter whilst it resulted in a rise in the C content in tree litter. We 5 hypothesize that fungal hyphae and mycorrhizae growing on the litter substrate could have brought considerable amounts of C and N onto the litter samples, and the overall C content in grass litter be biased by that increase, consequently keeping the carbon loss and carbon gain in equilibrium. Subsequently, the difference between tree and grass litter was taken into account as a random effect in the equations (Eqs. 4 and 5), 10 and satisfactorily generated highly significance models for both land use types.

Leaf litter decomposition traits across different climates and litter types
The relationships studied with empirical data allowed us to generate a few models including most of the interesting traits when studying litter decomposition: litter mass loss, and C and N contents during the decomposition process. Remarkably, the models 15 for forest litter worked for different species and origins, including conifer and deciduous litter, with a high range of initial N and C contents and total specific leaf area. Seen the highly significant relationships between the litter remaining mass with the climatic characteristics (cumulative air temperature and precipitation), and the relationships between the decomposition rate, mass loss and litter turnover with T a and P , we inferred 20 that the climatic characteristics could be sufficient predictors for estimating the speed of the decomposition process. Additionally, we found that the models beneficiated from including the land use (forest/grassland) as a fixed factor. We observed that in the grassland sites, the remaining litter mass was highly influenced by the decomposition site characteristics and not by the litter species. Similarly, in the forest sites, there was 25 a certain trend of lower remaining mass at a certain decomposition time as the site presented a warmer and wetter climate, resulting in a higher turnover rate, as studied by Kirschbaum (2000) in the American continent. With this experiment we found two key points in the relationships of litter decomposition traits with the climatic variables. Firstly, the correlations of the litter traits (remaining mass, and C and N contents) with climatic variables (air temperature and precipitation) were better when transforming the climatic variables to cumulative instead of using annual climatic averages for the given sites. This allowed to study the decomposition 5 process as a function of the input of energy and availability of moisture at each decomposition step along the year. Secondly, using T c, a > 0 allowed to overcome the fact that certain periods of the year are not favorable for decomposition; that is when the air temperature is ≤ 0, water is frozen and microbial activity is essentially stopped. As showed previously, we detected that the decomposition slowed down during winter time 10 as well as snow cover periods, and therefore, discarding the days with mean temperature below 0 • C increased the significance of our models. This variable has certain resemblance to the degree days used to describe and predict plant growth, which usually uses the lower temperature limit at around 10 • C. As addressed in the introduction, using T c, a > 0 and T > 0 is especially important for Northern sites, which present long pe-15 riods of freezing temperatures and litter decomposition is virtually stopped. We believe these variables should be taken in account for future modelling analysis and predictions.
Equations (2) and (3) performed a highly significance prediction for the litter mass remaining in the decomposition sites with only knowing P c and T c, a > 0 , which is of 20 paramount importance to extrapolate the litter turnover speed in these regions and in a climate change scenario. The importance of using cumulative variables in this particular case is that these can incorporate seasonal variations in the precipitation and temperature regimes, as is predicted to happen in Europe (Jacob et al., 2014).
We found high correlations between the cumulative precipitation (P c ) and air tem-25 perature (T c, a > 0 ) along the decomposition period, and individually with C r and N r , and consequently the C r , N r and k models rejected one of both variables (Eqs. 4-9). This happens because using T c, a > 0 as a predictor for C r and N r was sufficient for explaining much of the variation, and adding P c would not increase the explaining power of Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the models. Therefore, when including both climatic variables, the model AIC-value increased, P was not significant and thus was discarded. Similarly this happened with air and soil temperature variables (T c, a > 0 and T c, s > 0 ) when the model rejected one of both, choosing T c, a > 0 as the most significant. In conclusion, because climatic variables were highly correlated with each other in our sites, the models rejected predictors 5 which explained similarly the variation of the independent variables, and finally only few predictors were needed for the models, which was one of the aims of this paper. Similarly, Liski et al. (2003) used few climatic parameters (air temperature, precipitation and evapotranspiration) to predict the litter first-year mass loss. In the same way, our study was performed in a range of European climates which kept certain relationship 10 between P and T a , therefore we cannot predict goodness of our models in more extreme climates where this relationship would not be kept (e.g. semi-arid climate in SE Spain or subarctic climate in Lapland). We found a strong effect of the climatic characteristics on the decomposition of different litter types, corroborated by similar trends in different litter types when decom-15 posing in the same site. We found that the broad-leaved litter performed lower k rates than the conifer litter, and this could be a consequence of significantly higher specific leaf area in the broad-leaved litter. Therefore, the prediction models of tree litter k rates improved when including the initial total specific leaf area as a characteristic (Eqs. 8 and 9). Contrarily, the initial chemical composition (C and N contents) were excluded 20 from the equations.
Overall, despite having several climatic variables as inputs into the models, the AIC results were more favorable when the number of variables were less and the equations got simpler. In conclusion, having reached a simple model is in fact highly relevant when estimating these decomposition traits with few meteorological data available. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Conclusions
We found strong climatic influence driven by air temperature and precipitation on the litter mass remaining during the first year of decomposition in different types of litter. Models with few climatic parameters were enough to predict the remaining litter mass, decomposition k rate, C r , and N r content with high certainty. Litter nitrogen content in-5 creased during the first year of decomposition as the litter remaining mass decreased and the climate was wetter and warmer. The models generated better predictions when accounting for daily average air temperatures above 0 • C. Our models could be valid for extrapolation to other European climates where annual air temperature and precipitation are correlated, as it was in our case.  15 This work has been funded by the EU 7th Framework Programme NitroEurope (8-2/T6018PKPK06), the Technical University of Denmark, the Academy of Finland, the Maj Tor Nessling Foundation, the Estonian Science Foundation through SA Archimedes (DoRa), the Estonian Ministry of Science and Education (institutional grant IUT-8-3), the European Commission through the European Regional Fund (the Center of Excellence in Environmental Adapta-20 tion), and the Flemish fund for scientific research (FWO).
We thank Endla Nature Reserve (Estonia). We thank Poul T. Sørensen and Tiina Tosens for their technical assistance.  Berg, B. andLaskowski, R.: Nitrogen dynamics in decomposing litter, Adv. Ecol. Res., 38, 157-183, 2005. Berg, B. andMcClaugherty, C.: Initial litter chemical composition, in: Plant Litter -Decomposition, Humus Formation, Carbon Sequestration, 2nd Edn., edited by: Berg, B. andMcClaugherty, C., Springer-Verlag, Berlin, Heidelberg, 53-84, 2008. 5 Berg, B., Berg, M. P., Bottner, P., Box, E., Breymeyer, A., Deanta, R. C., Couteaux, M., Escudero, A., Gallardo, A., Kratz, W., Madeira, M., Malkonen, E., Mcclaugherty, C., Meentemeyer, V., Muñoz, F., Piussi, P., Remacle, J., and Desanto, A. V.: Litter mass-loss rates in pine forests of Europe and eastern United-States -some relationships with climate and litter quality, Biogeochemistry, 20, 127-159, doi:10.1007/Bf00000785, 1993 BGD BGD 12,2015 Biological and climatic controls on leaf litter decomposition M. Portillo-Estrada et al.  Zhang, C. F., Trofymow, J. A., Jamieson, R. C., Meng, F. R., Gordon, R., and Bourque, C. P. A.: Litter decomposition and nitrogen mineralization from an annual to a monthly model, Ecol. Model., 221, 1944-1953, doi:10.1016/j.ecolmodel.2010.04.015, 2010 Table 2. Individual estimates (with t value and P values) of the parameters modeled in Equations 2 to 9. The predictors which contain "site g " apply for models corresponding to grassland sites. In these cases, the resulting value is the result of the addition of the original predictor lacking of "site g " and the predictor containing "site g ". P c is cumulative precipitation, T c, a > 0 is the cumulative of daily average air temperatures higher than 0 • C, T a > 0 is mean annual temperature accounting for days with positive temperatures, and S LA is the total specific leaf area.   sites (a, b) and grass litter from two species in two grasslands (c, d) across Europe (see Table 1 for details). The cumulative air temperature is the sum of daily average temperatures above 0 • C from the beginning of the decomposition period until the date of litter collection for every litter bag. The cumulative precipitation is the sum of daily precipitation (in mm) from the beginning the decomposition period until the date of litter collection. The solid lines represent the Pearson's linear regression best fit (n = 336 for forest sites and n = 72 for grassland sites). Introduction

Conclusions References
Tables  . Litter decomposition traits of four different tree litter species with different origin decomposed in four European sites with different mean annual temperature (T a ) and mean annual precipitation (P ). Symbols as in Fig. 1. Data corresponding to the same origin of litter are connected with a solid line to visually evaluate the evolution of the trait across the T a and P range. The dashed line represents the Pearson's linear regression best fit of all the data. Traits are first-year decomposition k rate (a, b), percentage of mass loss after one year of decomposition relative to initial mass (c, d), estimated litter turnover time (e, f), and ratio between final litter N content (N f ) after one year of decomposition and the initial nitrogen content (N 0 ). Introduction