Post-print of Forest Ecology and Management 213 (2005) 209–228.

DOI:10.1016/j.foreco.2005.03.042

Sustainability of forest management practices: evaluation through a

simulation model of nutrient cycling

Juan A. Blanco 1

Miguel A. Zavala 2

J. Bosco Imbert 1

Federico J. Castillo 1,*

1
Departamento de Ciencias del Medio Natural, Edificio los Olivos. Universidad Pública

de Navarra, E-31006, Pamplona, Navarra, Spain.

2
Departamento de Ecología, Edificio de Ciencias, Universidad de Alcalá, E-28871,

Alcalá de Henares, Madrid, Spain.

* Contact: Tel: (34) 948 169 115

Fax: (34) 948 168 930

e-mail: [email protected]

Blanco et al. (2005) 1/47

Abstract

Forest harvesting may interfere with long-term ecosystem structure and function and

different harvesting methods will differ in their effects on soil fertility (e.g., whole-tree

harvesting vs. stem removal). In the case of thinning, effects of thinning intensity,

rotation length and site quality must be assessed in order to formulate sustainable

management practices. Assessment of the relative impact of these practices is difficult,

however, given the long-temporal scales involved. In this study, we implement a

process-based model of nutrient cycling to evaluate temporal changes in ecosystem

nutrient dynamics of managed and non-managed forest stands. The model was

specifically designed to asses differences between two contrasting site-quality Pinus

sylvestris L. stands in the western Pyrenees (Navarre, Spain) managed under two

thinning intensities. The model describes the main nutrient fluxes in the stand: litterfall,

decomposition, retranslocation, root uptake and management type, and it was

parameterized and verified with three years of field data. After model verification we

examined the effects of thinning intensity, thinning frequency and harvesting method

(whole-tree vs. stem removal) on potential nutrient losses. The results suggest that in

this heterogeneous region, sustainability of forestry practices is strongly site dependent.

N and P were particularly sensitive to overexploitation and in no case could whole-tree

removal could be recommended as it may have a strong negative effect on nutrient

reserves. In relation to previous nutrient cycling models, our model offers a satisfactory

compromise between simplicity, biological realism and predictability, and it proved to

be a useful tool to predict short-term changes in nutrient reserves as well as to evaluate

possible negative effects of applying current thinning prescriptions on long-term

sustainability of managed forests in the western Pyrenees.

Blanco et al. (2005) 2/47

Key words: Thinning; Pinus sylvestris L.; Scots pine; Navarre; Whole-tree removal;

Forest harvesting; Nitrogen; Phosphorus; Mediterranean pine forests.

1. Introduction

To achieve ecosystem and economic sustainability a forest practice must achieve three

different goals: (1) to be economically profitable and to perpetuate forest cover; (2) to

preserve ecosystem structure (e.g., for biodiversity values); and (3) to preserve

ecosystem function (e.g. nutrient cycling) (Zavala and Oria, 1995; Sverdrup and

Svensson, 2002). Among other effects, thinning reduces stand biomass, nutrient

contents, litterfall (Klemmedson et al., 1990; Harrington and Edwards, 1999), and can

alter decomposition rates (Piene and Van Cleve, 1978; Pérez-Batallón et al., 1998).

Continuous long-term studies of the effect of management practices on nutrient budgets

are expensive and time consuming. As a compliment to long-term experiments,

simulation models are useful tools that allow us to extrapolate observed short-term

changes in nutrient dynamics to longer time scales (Landsberg, 2003). Experimental

studies and modelling are complementary approaches within an adaptive management

context that can identify essential mechanisms controlling short and long-term

processes that are critical for maintaining ecosystem structure and function

(Kirschbaum, 1999; Grigal, 2000). Once identified, a field monitoring program can be

designed that maximize the efficiency of sampling and laboratory analysis. That are

necessary to confirm that management is sustainable.

Blanco et al. (2005) 3/47

Models of nutrient cycling are of particular importance for evaluation of sustainability

of forest practices. Typically these models include feedbacks among litterfall,

retranslocation, tree growth, root uptake and decomposition while above-ground

processes are directly related to tree growth. A number of modelling approaches have

been used to describe decomposition and root uptake processes. These models differ

both in their structure and in their applicability for application in forest management.

The first decomposition models were correlational (Andersson et al., 2000) and were

based on statistical relationships among the different variables involved (Olson, 1963;

Swift et al., 1979). A second generation of models explicitly considered different

fractions of soil organic matter characterized by chemical composition and

decomposition rates (CENTURY, Parton et al. 1987). These models are currently used

for forest research and management as well as modules within biosphere carbon cycle

simulators (e.g. FORSANA, Grote et al., 1998; FORECAST, Kimmins et al., 1999;

CenW, Kirschbaum, 1999; EFIMOD 2, Komarov et al., 2003; ForNBM, Zhu et al.,

2003). Although they provide adequate biological realism, these models tend to be very

complex. This adds uncertainty to the causes underlying their final predictions and

weakens their heuristic value (Andersson et al., 2000). Finally, analytical models such

as the "Q" model (Agren and Bosatta, 1996) provide useful theoretical insights but they

often omit too much critical detail (Kimmins, 2004) or are based on parameters difficult

to estimate under field conditions. Therefore, the applicability of these simpler models

for forest management is often quite limited (Battaglia and Sands, 1998). From a

management perspective the most adequate model can be defined as the simplest one

among those that meet the users needs both in terms of resolution and precision

(Battaglia and Sands, 1998; Botkin, 2001; Landsberg, 2003). Based on this idea, in this

study we have developed a relatively simple model that can be easily parameterized

Blanco et al. (2005) 4/47

with field data and that allows us to quantify and to compare the sustainability of

different thinning regimes and harvesting methods for P. sylvestris L. stands of

contrasting site quality. The model has been developed in response to current forest

management needs in Navarre (Spain), and to our knowledge is the first biologically-

informed model developed in Spain resulting from the cooperation between the

university system, the public administration and a forestry enterprise. Data and model

validation were performed on two contrasting experimental forests located in the

western Pyrenees which represent the two extremes of a site quality gradient. Aspurz is

a highly productive Mediterranean forest with relatively fast decomposition rates. Garde

in turn exhibits a more continental climate and has a lower production and slower

decomposition rates (Blanco et al., 2003a; 2003b). P. sylvestris L. forests have been

extensively studied in northern Europe (e.g. Berg and Lousier 2000; Agren and Knecht,

2001, among others). Results from these studies, however, may not be directly

applicable to the Iberian P. sylvestris forests (the southern and western distribution

limits for this species) where ecological factors and processes can be qualitatively

different from those operating in northern forests. For example, thinning has been

shown to reduce decomposition rates in the western Pyrenees forests (Blanco et al.,

2003b), while the opposite has been found at higher latitudes (Edmonds, 1990). This,

among other observations, points out the need to develop simple biologically-based

models of nutrient cycling specifically suited for Mediterranean forests that can assist us

in the evaluation of sustainability of current forest management practices (e.g. Harmon,

2001; Verbug and Johnson, 2001). Our modelling approach is based on the

simplification strategy proposed by Tiktak and Van Grinsven (1995). These authors

propose to minimize the dynamic feedback between geochemical processes and stand

growth using independent submodels, to omit or aggregate short-time scale processes

Blanco et al. (2005) 5/47

(daily and seasonal) and to pool compartments, simplifying and aggregating process

descriptions. The main objectives of our work are: (1) to assess whether site quality

influences sustainability of thinning practices in western Pyrenees forests, (2) to forecast

long-term changes of nutrient pools in two contrasting site-quality stands as a function

of harvesting method (whole-tree vs. log removal), thinning intensity and rotation length

and (3), to evaluate how predicted nutrient losses for different management scenarios,

including those currently recommended in P. sylvestris forests in Spain, influence stand

nutrient balance and therefore, long-term sustainability of thinning practices.

2. Materials and methods

2.1. Study sites

The model was tailored to describe nutrient cycling of P. sylvestris stands in the western

Pyrenees (Spain). The model was parameterized with data from two contrasting

experimental sites, representing the two extremes of a site quality gradient throughout

this region. The lower elevation site (Aspurz, 42º 42’ 31’’ N, 1º 8’ 40’’ W) is one of the

most productive P. sylvestris forests in Spain. The stand has developed on a Dystric

Cambisol, with mean A horizon organic C of 7.0 %. Site mean altitude is 642 m asl

with mean annual precipitation and mean annual temperature 912 mm yr -1 and 12 ºC,

respectively. The site is an even-aged P. sylvestris stand resulting from strip-like clear-

cutting carried out in the mid-1960’s. Mean stand age is 37 years and stand density is

3555 trees ha-1, with a dominant height of 14 m and a mean dbh of 24.9 cm. The higher

elevation site (Garde, 42º 48’ 50’’ N, 0º 52’ 30’’ W) is an example of a low production

P. sylvestris forest in Spain. The stand is located on a Haplic Alisol, with mean A

Blanco et al. (2005) 6/47

horizon organic C of 6.3 % at a mean altitude of 1335 m asl. Mean annual precipitation

is 1268 mm year-1, and mean annual temperature is 8.2 ºC. Stand structure is even-aged

resulting from clearcutting during the early sixties. Mean stand age is 32 years, with a

density of 3500 trees ha-1, a dominant height of 15.2 m and a mean dbh of 25.0 cm.

Rubus ulmifolius Schott and Pteridium aquilinum (L.) Kuhn are the two dominant

species in the understory of both Garde and Aspurz but in addition Fagus sylvatica L.

saplings are found in Aspurz. Bedrock for both sites consists of flysch of sandstone and

limestones.

2.2. Experimental design

Data were gathered across nine rectangular (30 x 40 m) plots per location. Silvicultural

treatments were implemented by the ‘Departamento del Medio Ambiente - Gobierno de

Navarra’ according to the guidelines of the International Union of Forestry Research

Organizations (IUFRO). The experimental design consisted of three treatments with

three replicates per location: treatment 1 (P0) control with no thinning; treatment 2

(P20) moderate low thinning (20 % of basal area removed according to future tree

selection method; felled trees were mainly canopy suppressed but included dominant or

codominant trees with malformed stems); treatment 3 (P30) heavy low thinning (30 %

of basal area removed using the future tree selection method as in P20). To avoid edge

effects, the silvicultural treatments were applied within a 5 to 10 m strip adjacent to

each plot.

2.3. Model structure, parameterization and data gathering

Blanco et al. (2005) 7/47

Our main objective was to estimate changes in N, P, K, Ca and Mg contents in the

major nutrient pools. The model was designed to reach a satisfactory compromise

between mathematical complexity and biological realism (e.g. Landsberg, 2003), and

the number of parameters were kept as low as possible. Fig. 1 shows the man fluxes and

pools considered in the model and Fig. 2 shows model inputs and outputs. Tree to soil

feedbacks are integrated into the empirical growth function (Tiktak and Van Grinsven,

1995; Agren and Knecht, 2001). Summer drought in these two locations is only

moderate and hydrological processes were not explicitly simulated. Thus, model results

are restricted to non-water limited forests (Verburg and Johnson, 2001). All combined

processes were simulated with one year time-steps with program language STELLA

Research 5.1.1 (High Performance Systems Inc., 1997) which is particularly suited for

modelling ecosystem level processes (Constanza and Gottlieb, 1998).

Stand growth functions were simulated with SILVES (Del Río and Montero, 2001)

which simulates diameter distributions and thinning of P. sylvestris stands in Spain and

calculates stand growth as a function of site index, stand basal area and tree density.

Original parameter values were modified based on growth and yield tables developed

for this species in Navarre (Puertas, 2003) (data provided by "Departamento de Medio

Ambiente - Gobierno de Navarra”). Equations describing stem growth were:

Mass (Mg ha-1) = 495.348 Ln (Age yrs) – 1473.729 (R2 = 0.99, P<0.001) for Aspurz,

and Mass (Mg ha-1) = 308.595 Ln (Age yrs) – 902.013 (R2 = 0.99, P<0.001) for Garde.

Stem, needles and branches mass were predicted from stem mass through the following

allometric equations determined from field data: Needles (Kg) = 0.0258 · Stem (Kg) +

0.9428 (R2 = 0.66, P<0.001); Branches (Kg) = 0.0992 · Stem (Kg) + 1.2482 (R2 = 0.74,

P<0.001). To estimate needle, branch and stem biomass before thinning, trees were

Blanco et al. (2005) 8/47

classified into five dbh classes. Then, the modal tree per dbh class was felled for each

plot, for a total of five trees per plot. After determining the component biomass for

felled trees, total stand above-ground biomass was calculated by multiplying the

biomass of each dbh class by the number of trees in that class. To determine nutrient

amounts in needles, branches and stems, we multiplied tissue concentrations by the

appropriate component. Root biomass was not estimated in the field, so we assumed

that root nutrient contents represent a constant fraction of the above-ground tree

nutrients contents as Margolis et al. (1995). Using above-below ground ratios (Table 1)

reported by Malkönen (1974) and Litton et al. (2003). A similar assumption have been

made in more complex forest models such as HYBRID (Friend et al., 1993) and

C_CHANGE (Beets et al., 1999).

Litterfall in litter traps was collected monthly from April 2000 to October 2002 in both

forests (9 litter traps of 0.29 m2 per plot x 3 treatments x 3 replicates x 2 sites x 31

months = 5022 samples). Samples were separated in the laboratory into six components

(needles, branches, fruits, bark, other pine organs and other litter) dried at 72 ºC and

weighed. For comparison with the simulation only two litterfall fractions were used: 1)

needles and 2), a combined branches, bark and fruit (referred to below as the woody

litter fraction). Litterfall inputs from tree species other than P. sylvestris and from

understory plant species were small important, particularly in Garde, and therefore they

have not been included in the model. Leaf litterfall and wood biomass has been assumed

to be a fraction of total above-ground leaf and stemwood biomass, respectively

(Kimmins et al., 1999; Kirschbaum, 1999; Komarov et al., 2003). Litterfall values for

our two experimental sites (Table 1) were similar to mean litterfall values of pine forests

reported by Agren (1983), and to those used in models developed by Agren and Knecht

Blanco et al. (2005) 9/47

(2001). Green and senescent needles were collected bimonthly from December 2001

until October 2002. Three branches per plot at 5 m above forest floor were cut every

sampling date at , and green needles were separated into three cohorts. A total of 972

green samples were collected during this study (3 branches per plot x 3 cohorts x 3

treatments x 3 replicates x 2 sites x 6 dates). Mean retranslocation efficiency for all

cohorts together was calculated following Aerts et al. (1999) as: % retranslocation =

100 x (Mean concentration in green needles – Mean concentration in senescent needles)

/ Mean concentration in green needles. Change in nutrient content of the green needles

after retranslocation was calculated by multiplying retranslocation percentage by total

needle nutrient content. We assumed that retranslocation from roots and woody litter

fractions was negligible (Verbug and Johnson, 2001). To calculate the uptake of

nutrients by the roots we assumed that the uptake equalled nutrients requirements by

trees (Cole and Rapp, 1981; Waring and Running, 2001), which were provided in part

by nutrient retranslocation from senescing needles. To estimate tree nutrient

requirements we added up the amount of nutrients used for needle, branch, stem and

root growth, plus losses through litterfall of branches and needles, plus below-ground

transfers of dead roots and root exudates to the soil.

Decomposition rates for needles were obtained from 50 litterbags (2 mm mesh size, 10

g of needles) per plot placed in the stands in November 2000. Three bags were collected

for mass loss measurement, monthly for the three first months and then bimonthly (from

December 2000 to October 2002). A total of 702 samples were collected in this study (3

bags per plot x 3 treatments x 3 replicates x 2 sites x 13 sampling dates). Decomposition

of needle and woody litter were simulated assuming that both fractions consisted of a

single type of material and exhibited negative exponential weight losses over time

Blanco et al. (2005) 10/47

(Olson, 1963). We considered different needles cohorts which fell down in different

years and did not change neither their chemical composition nor their decomposition

rate through the decomposition process. Nutrients released from decomposing litter

were estimated as

Mt = Mo e-k t (Olson, 1963)

where Mt is mass remnant at time t, Mo is initial mass and k is the decomposition rate.

To estimate the amount of organic matter decomposition at a given time t (OMDt), we

estimated variation in soil organic matter as the balance between inputs from litterfall

and losses from decomposition according to:

OMDt = OMlitterfallt - OMdecompositiont

This is

OMdecomposedt = OMlitterfallt - OMDt = OMlitterfallt – (OMt – OMt-1)

where OMt is the organic matter content in decomposing litter. Litter is composed of a

number of cohorts generated at different times. At a given time organic matter is the

sum of the remnant fraction in each of these cohorts which can be estimated as a sum of

negative exponential curves (see a CENTURY-based simplification by Bolker et al.,

1998). Thus, the amount of organic matter that is lost at any given time from

decomposing litter was estimated as

 x=t 
   
x=t
OMdecompt  OMlitterfall t    OMlitterfall x  e- k  x   OMlitterfall x -1  e- k (x -1) 
 x =0 x =0 

To calculate the decomposition of woody materials we used the rates reported by Agren

and Bosatta (1996) (Table 1), following the same procedure as for needles. To simplify

Blanco et al. (2005) 11/47

the model we assumed that nutrients and organic matter were lost at the same rate over

the decomposition process (Andersson et al., 2000; Chertov et al., 2001; Potter et al.,

2001). By expressing organic matter changes through time as a sum of first order

exponential curves, we can quantify the sensitivity of the model to organic matter

changes and make a better use of the data available (Bolker et al., 1998). As for roots,

and given that they were not directly studied, we kept the decomposition process as

simple as possible. Therefore, dead roots and roots exudates were pooled together and

decomposition rate was estimated as a fixed percentage of total root biomass for all the

soil profile following Malkönen (1974) (Table 1)

Soil B horizon samples were collected once in November 1999. To better characterize

the upper A horizon we collected soil samples 5 cm deep bimonthly in all plots from

April 2001 to April 2003. Soil and leaves (green and senescent) samples were ground

with an electric mill. N (soil and leaves) was analyzed using the Kjeldhal method

(Harwitte, 1980). Soil and foliar P were analyzed following Bray-Kurz (1945) and the

phosphomolibdo-vanadate method (MAFF, 1986), respectively. For both soil and leaves

samples Ca and Mg were analyzed by atomic absorption spectrophotometry, and K by

flame photometry. In the model all soil horizons were pooled together excluding

branches and needles in the forests floor. To calculate initial forest floor total nutrient

content from which to initiate all simulations, we ran the model for 30 years starting

from a soil with no organic layer. Values for different litter fractions at the end of the

simulation were used as the initial values on posterior runs of the model (Mäkipää et al.,

1998; Kimmins et al., 1999).

Blanco et al. (2005) 12/47

Simulation of leaching losses of N, Ca and Mg was based on simulated available

quantity of these nutrients less the amounts taken up by the trees. This approach is also

used in BIOME-BGC (Hunt and Running, 1992), Q (Rolff and Agren, 1999), CenW

(Kirschbaum, 1999), FORECAST (Kimmins et al., 1999) and NuCSS (Verbug and

Johnson, 2001). In the case of P and K, it was assumed that nutrients not used by trees

were inmobilized in the soil (Cooke, 1981). Atmospheric inputs data (wet and dry

deposition) used in the model were taken from a nearby site (<40 km distant) by the

Spanish Forest Protection Service (Ministerio de Medio Ambiente, Madrid, 2003); the

values were similar to mean deposition rates in non-polluted areas in Europe. Given that

both experimental sites have similar parent material and that distance between them is

less than 40 km, we assumed similar atmospheric inputs and geological weathering for

both locations. Forest soil weathering rates for this type of bedrock have been reported

in Global data base sets elaborated by Kimmins (2004) and Fisher and Binkley (2000).

2.4. Model analysis: verification and sensitivity analysis

We verified our model in several ways. Predicted versus observed values for short-term

responses to thinning were evaluated for the three years following thinning. We

simulated changes in nutrient contents in the control plots after thinning, and then

compared model estimates with data from the thinned plots at both locations. A more

detailed deviation analysis between observed and simulated data was carried out for N

as this nutrient is the most commonly limiting nutrient in forests (e.g. Parton et al.,

1996; Gilmanov et al., 1997). The mean absolute deviation (Dabs) was calculated as:

Blanco et al. (2005) 13/47

 1 n
D abs    | X est ( t i )  X obs (t i ) |
n i =1

where Xest and Xobs are estimated and observed values for a given variable, and t i (i=1,

..., n) represents time in years. Maximum N was defined as the maximun observed value

of a particular flux or pool of N in the experimental plots. We compared for each type of

N flux or pool the percentage of the maximum N value accounted by the Dabs with the

percentage of the maximum N value accounted by the standard error of the observed

data.

To determine the sensitivity of the output variables to variation in input parameter

values, we varied each parameter input ±10% when the trees reached 100 years, and

then assessed the changes shown by above-ground nutrient content of tree biomass

(stems + branches + needles), decomposing plant debris (woody litter + needles) and the

totals for all soil horizons (soil + woody litter + needles). The percentage change in the

output variables was partitioned into percentage variation due to each parameter,

obtaining a sensitivity percentage. Thus, a sensitivity value higher than 100% indicates

that the model amplifies parameter changes while a value lower than 100 % indicates

that model internal dynamics reduce them (Kirschbaum, 1999).

2.5. Management scenarios

To simulate different thinning treatments we imposed different combinations of

thinning intensity (percentage removal of basal area), thinning frequency (number of

years between two successive thinnings) and removal types (stems or whole-trees). In

addition, we have analyzed current management prescriptions for these forests in Spain

Blanco et al. (2005) 14/47

(Díez and Fernández-Golfin,1997; Del Río and Montero, 2001) and specifically those

adapted to Navarre (Puertas, 2003). According to these prescriptions the exploitation

cycle may be reduced to 80 years in high quality stands. Thinning initiates when stands

are 20 or 25 years old, and continues every 10 years removing 30% of the total basal

area when timber extracted exceeds 50 m3 ha-1 (economic criteria). This prescription

was simulated for Aspurz. A low intensity thinning in low quality stands, however, is

not profitable and production cycles can be as long as 120 years, with a thinning

program that removes 20% of the total basal area every 15 years. This was the thinning

program simulated for Garde. We established a 100 year-production cycle for both

localities.

3. Results

3.1. Model analysis: verification and sensitivity analysis

There was a strong correlation between predicted and observed values for all nutrients

inspected (Table 2). Dabs of N removed within the stems was 3.9 kg N ha-1, representing

5% of the maximum observed value in this pool (89.1 kg N ha-1). This percentage was

similar to the results of the maximum observed N pool accounted for by the standard

error of the observed values (6 %). For N in branches, the Dabs percentage (6 %) was

slightly lower than the percentage for the observations error (7%). For N in needles, the

Dabs percentage (12%) was similar to the observational error percentage (9%). Dabs for

N in needle litterfall (1.3 kg N ha-1), was within the precision range of the

measurements and its percentage (5%) was almost equal to the observational error

percentage (4%). Finally, for woody litterfall (10.1 kg N ha-1), the Dabs percentage

Blanco et al. (2005) 15/47

(11%) was bigger than the observational error percentage (6%). Sensitivity analysis

showed that the model behaved similarly in both localities, diminishing the changes for

all the parameters (less than 100% of variation for all variables, see Materials and

Methods). The most sensitive variable was N content in decomposing leaf litter (needles

+ woody litter), as it varied 55% in Aspurz when the litterfall fraction parameter was

changed. For other parameters sensitivity values were lower than those for the N content

in decomposing litter.

3.2. Changes in nutrient pools over time

Woody litter was the main pool for N in the control stands as its low decomposition

rates favoured accumulation, particularly in Garde (Olson’s k was 42% lower than in

Aspurz, Blanco et al., 2003b) (Fig. 3 and 4). On the contrary, stems were the pool with

the lowest N contents in the control stands. A total of 1819 kg N ha-1 had accumulated

in Aspurz at the end of the simulation, which represented 55 % more than that

accumulated in Garde (1177 kg N ha-1). The simulated thinning cycle caused a 60 and a

27 % total N reduction in Aspurz and Garde, respectively (Fig. 3 and 4). The most and

the least sensitive pools were needles and woody litter, respectively. These differences

caused changes in the relative importance of each pool, particularly for needles which

ranked first before thinning, and second after thinning in both sites. For P, the branches

were the most important pool in the control stands with 39 kg P ha-1 in Aspurz and 44

kg P ha-1 in Garde at the end of the simulation. The relative importance of woody litter

increased over time, and became the second most important pool in both sites at the end

of the simulation. Total above-ground content of P was 145 kg ha-1 and 131 kg ha-1 in

Aspurz and Garde, respectively. The distribution of P among pools changed

Blanco et al. (2005) 16/47

dramatically after applying the recommended management practices (Fig. 3 and 4).

Thus, total P content in leaf litter increased 69 % and 37% in Aspurz and Garde,

respectively, while total aboveground content of P decreased 19% and 21% in Aspurz

and Garde, respectively.

Unlike P and N, K content in the thinning plots was smallest in the decomposing pools

(leaf litter and woody litter), while in the control plots the pools with the highest K

content were stems and branches. At the end of the simulation, the stems had 322 kg K

ha-1 in Aspurz and 296 kg K ha-1 in Garde. Total K content was similar in both forests

(913 kg kg K ha-1 in Aspurz and 834 kg K ha-1 in Garde, Fig. 3 and 4). The

decomposing woody litter was the most important pool for K in Aspurz after thinning,

but in Garde the stems and the branches were the most important pool before and after

thinning (Fig. 3 and 4). In addition, only 39% and 73% of K remained in Aspurz and in

Garde, respectively, at the end of the simulation. The response of Ca reflected its

structural role in stems, being the pool with the highest contents. In the reference plots

Ca accumulation in the stems was 780 kg ha -1 in Aspurz and 468 kg ha-1 in Garde. After

thinning, total Ca showed a 52% reduction, but in Garde it only diminished 2%. Finally,

differences between sites were higher for Mg than for Ca (Fig. 3 and 4). Decomposing

leaf litter was the most important pool (170 kg Mg ha-1 at the end of the simulation) in

Aspurz baseline plots. In Garde, however, stems were the most important pool with 124

kg Mg ha-1. Thinning caused little changes in the relative importance of Mg among

pools in both sites, although the total content of Mg decreased 68% in Aspurz and 27%

in Garde.

3.3. Potential nutrient losses due to the thinning and tree removal method

Blanco et al. (2005) 17/47

Potential nutrient losses are defined as the sum of nutrients extracted due to tree

removal and nutrients lixiviation when mineralization rates exceed root absorption.

Accumulated potential losses resulting from simulations in Aspurz averaged 345 kg N

ha-1 when only tree stems were removed, and 635 kg N ha-1 for whole-tree removal.

These values were lower in Garde: 269 kg N ha-1 and 523 kg N ha-1 respectively. An

increase in thinning intensity up to 20% resulted in higher potential losses particularly

for whole-tree removal (Fig. 5). The rate of increase diminished after a thinning

intensity approximately 20%. Phosphorus mean potential losses for stem removal

equalled 18 kg P ha-1 in Aspurz and 10 kg P ha-1 in Garde, and considerably increased if

the whole tree was removed (41 kg P ha-1 in Aspurz and 42 kg P ha-1 in Garde).

Maximum losses in Garde occurred for a range of thinning intensities between 20 and

30%, while in Aspurz, nutrient losses increased and reached a plateau after 10% of the

basal area was removed (Fig. 6). Mean potassium losses in Aspurz equalled 163 kg K

ha-1 for stem-only extraction and 400 kg K ha-1 for whole-tree removal and 161 kg K ha-
1
and 370 kg K ha-1 in Garde, respectively. Maximum potassium losses in Garde were

found for thinning intensities around 30% (whole-tree removal). These trends were not

as pronounced in the Aspurz plots (Fig. 7). For calcium, differences among the two

types of tree removal were not as dramatic 423 kg Ca ha-1 and 510 kg Ca ha-1 in Aspurz,

and 276 kg Ca ha-1 and 340 kg Ca ha-1, in Garde for stem and whole-tree removal,

respectively. Loss rates increased dramatically with thinning intensity up to 20% and

were less pronounced after 20% to decrease afterwards (Fig. 8). Similar trends were

described for magnesium with average potential losses of 87 kg Mg ha -1 and 138 kg Mg

ha-1 (Aspurz) and 83 kg Mg ha-1 and 114 kg Mg ha-1 (Garde), respectively for stem and

whole-tree removal (Fig. 9).

Blanco et al. (2005) 18/47

Overall, a strong interaction between silvicultural treatments and sites was observed.

For each nutrient, thinning resulted in a broader range of variation of potential nutrient

losses (difference between maximum and minimum losses) in Garde when compared to

Aspurz. This variation was also affected differentially by tree removal procedure in

each site, being more pronounced in Garde than in Aspurz. In particular whole-tree

removal amplified this variation in Garde but ameliorated it in Aspurz. With respect to

thinning frequency, an increment from 5 to 25 years resulted in losses of around 50%

for each nutrient in both sites (Figs. 5 to 9).

4. Discussion

Despite its simplicity, our model was in a good agreement with the observations over a

short time frame (Table 2). Also low model sensitivity suggests that model structure is

robust with respect to error propagation of initial estimates (Chertov et al., 2001).

4.1. Factors influencing sustainability of forestry practices

Site quality at each location were the most important factor influencing sustainability of

forest management. Potential N losses in Garde, a site with poorer site productivity and

N stem content, were significantly lower than those found in Aspurz (Fig. 5). Similar

patterns were observed for Ca and Mg. For P and K, however, lower outputs were

compensated by larger nutrient concentrations found in Garde trees. Thus, potential

nutrient P and K losses were similar in both sites. Comparable results have been found

by Morris et al. (1997) who pointed out that sustainability of forestry practices may

Blanco et al. (2005) 19/47

critically depend on site ecological features such as decomposition rate (in our study

higher in Aspurz than in Garde, see Blanco et al., 2003b). Potential lixiviation rates for

N, Ca and Mg were estimated as the difference between mineralization release and root

uptake. Thus, higher mineralization rates may result in higher potential losses. Also an

increment in lixiviated nutrients after thinning has been evidenced in experimental

studies by Baeumler and Zech (1998).

In addition to the site, tree removal procedure (stem vs. whole-tree) had the second

impact on sustainability. For all nutrients and sites, whole-tree removal resulted in

significantly higher potential losses with respect to traditional on-site tree processing,

allowing nutrients within leaves, twigs and branches to recycle. Increments in potential

nutrient losses due to whole-tree removal were particularly critical for P and K and are

mostly located in branches and needles. Nutrient losses were less for Ca and Mg which

are primarily found in the stems. According to the model results, these increments turn

out to be critical for the long-term sustainability of current thinning practices. For

example, in Aspurz, N inputs compensate for N losses for a range of thinning regimes

but this balance reverses under whole-tree removal operations. This leads to important

reductions in nutrient reserves that may compromise future forest productivity. Similar

results have been found by Morris et al. (1997) and Rolff and Agren (1999). Our model

is conservative with respect to other factors such as competition of understory plants

that may amplify this reduction (Jacobson et al., 2000).

The third determinant of sustainability was thinning intensity. Potential nutrient losses

increased dramatically between unmanaged and managed stands for all sites and

nutrients. Thinning effects on nutrient losses increased up to a maximum for a basal

Blanco et al. (2005) 20/47

area removal of approximately 20%, and decreased afterwards. The effect of thinning

intensity interacted with site. In Garde, maximum P and K losses were found under

thinning intensities around 30% and tended to decrease at higher intensities. These

variations, however, were significantly smoother in Aspurz (Figs. 6 and 7). At

intermediate thinning intensities the increment in standing crop due to increased growth

of released trees, compensates for the decrease in production derived from lower tree

density. Hence, it may be possible to maintain or even to increase wood extraction

relative to that of lower intensity thinning regimes. At higher thinning intensities,

however, a point is reached in which reduction in tree density is so severe than growth

increments of remnant trees do not compensate for the density reduction and, thus wood

extracted in each cycle tends to decrease. On the other hand, decrease in decomposition

rates in response to thinning in Aspurz (Blanco et al., 2003b) results in a decrease in

mineralization rates and likely a reduction in lixiviation. Nutrient losses due to

extraction were relatively higher than those caused by an increase in lixiviation and

were therefore in agreement with previous studies (Grigal, 2000; Zhu et al., 2003;

Kimmins, 2004). Johnson and Todd (1987) also showed that this difference had

significant effects for N and P concentrations in forests in Tennessee (U.S.A.). As these

authors suggest it is unlikely that lixiviation alone results in a decrease in productivity,

even during late successional stages when losses due to lixiviation equal ecosystem

inputs (Vitousek and Reiners, 1975).

Finally, according to model results, sustainability of the thinning regime is influenced

by its frequency or rotation cycle which determines the average number of years

between two consecutive forest operations. The shorter the cycle, the larger the number

of interventions and thus, the larger potential losses of nutrients. This trend can be

Blanco et al. (2005) 21/47

observed for all the nutrients considered (Figs. 5 to 9), and interacts with thinning

intensity, so its effect is relatively larger for low thinning intensities. This reflects forest

capability to recover initial biomass after moderate interventions. The remnant trees can

restore or even surpass the previous standing crop density, providing that operational

lag is long enough (Thornley and Cannell, 2000). However, if thinning intensity is

greater than 30% stand ability to restore its initial biomass decreases (Montero et al.,

1999). In this case, a longer interval between interventions does not allow for stand

recovery and the relative impact of thinning frequency decreases. Short rotations also

may increase the time of soil exposure and low forest cover (increasing the period over

which mineralization rates are greater than tree demand) therefore, increasing

lixiviation. Rolff and Agren (1999) predicted a decrease in productivity for shorter time

intervals between interventions, while Seely et al. (2002) observed that shorter rotations

can result in a substantial decrease in forest nutrient reserves in the long term. Our

modeling results suggest that thinning frequency effects on nutrient reserves are not as

critical as those of the tree removal procedure and thinning intensity. This does not

imply, however, that this aspect of forest management is not critical for other

components of sustainability.

4.2. Silvicultural implications

All situations that compromise ecosystem nutrient recharge capability should be

avoided as they will result in a gradual decrease of nutrient reserves and productivity

(Kimmins, 2004) . Nevertheless, it is possible, to attain a level of exploitation that

allows to sustain long-term forest productivity or even to induce a recovery of nutrient

reserves (Morris et al., 1997). Our modeling study suggests that P reserves can be

Blanco et al. (2005) 22/47

particularly sensitive to overexploitation (Fig. 6). This is in agreement with previous

experimental studies conducted at our site which suggest P limitations (Blanco et al.,

unpublished data) and indicate the need to pay more attention to phosphorus cycling

under different site and thinning regimes, as well as its effects on site productivity and

community composition. The strong variability across sites regarding their response to

thinning suggests that results should not be extrapolated from one site to another.

Chiefly, site dependency of tree growth rates result in quantitative and qualitative

differential responses in nutrient dynamics (Figs. 3 and 4). Thus, forestry prescriptions

for P. sylvestris in the Central and Iberian ranges in Spain (Del Río and Montero, 2001)

may result in nutrient reserve dynamics close to recharge capability. In particular, for P

reserves application of these prescriptions may be below the sustainability threshold in

the least productive site, and on the verge of this threshold in the most productive site.

On the high quality site these practices may result in N losses below inputs, but for P

reserves the balance results in significant net losses suggesting the need to reduce

thinning intensity or rotation cycles. Whole-tree extraction is not advised in any of our

sites as it resulted into unsustainable P losses for all thinning regimes considered. This

practice would only be justified if applied along with N-P fertilization as proposed by

Rolff and Agren (1999), although its costs are likely not viable in this region.

Finally, other forest interventions such as fine wood extraction to decrease fire risk,

slash crushing to accelerate decomposition or the use of litter for gathering game or

domestic range could have very different impact on natural vs. managed forests. This

may be especially true when decomposing material is the most important forest nutrient

reserve, which among other factors, depends on the nutrient considered, locality and

management practice. In this context, in order to ameliorate CO2 emissions, using

Blanco et al. (2005) 23/47

biomass such as leaves and branches as a substitute for fossil fuels has been claimed to

be more beneficial from social and economic perspectives than sequestering the carbon

in forests (Hall, 1997). However, the results from this study suggest caution to favor this

practice without proper evaluation as forest sustainability may be negatively affected by

removing the whole tree from the forest.

4.3. Model limitations and further work

Firstly, we have only considered nutrient cycling associated with P. sylvestris while

understory effects that can have an influence on nutrient retention and reduce lixiviation

have been ignored. Secondly, we have oversimplified root description and despite low

model sensitivity, the decomposition of dead roots and exudates should be adequately

described as they can constitute an important soil biomass input (Beets et al., 1999).

Thirdly, our experimental data do not indicate effects of thinning on decomposition

rates at Garde after two years of study but show a lowering effect at Aspurz, possibly

due to alterations of the decomposer community (Blanco et al., 2003b). We have

tentatively explored how these latter results may affect nutrient losses but we need

further empirical studies to understand in which direction these changes will take place.

For example, increments in incident radiation and temperature, due to a reduction in tree

density, may stimulate mineralization and result in losses from lixiviation higher than

the ones described in our model. Fourthly, if nutrient concentration in trees declines

with age (Kimmins, 2004), losses from wood removal would be lower than those

simulated. Fifth, the external components of the nutrient cycle (runoff, erosion, etc.)

should be adequately described to asses its relative contribution to other fluxes in our

model. Finally, simulated estimates of nutrient losses can be misleading. For example,

Blanco et al. (2005) 24/47

in unmanaged forests nutrient outputs can be naturally high (e.g. Ca in calcareous

locations), or there may be managed forests in which high nutrient reserves allow for

more intense exploitation. Thus, it may be more suitable to establish relative

comparisons among alternate thinning regimes to rank them in terms of potential risk.

All these limiting factors caution against our quantitative predictions. Despite these

limitations, our work shows how simple simulation models, based on parameters which

can be easily obtained from standard forest ecosystem studies (litter traps and

decomposition bags, chemical analyses of soil, needles, branches and wood) can be

used as diagnosis tools to compare the potential impact of alternate interventions on

long-term sustainability and suggest critical processes that may not be intuitively

obvious. The main challenge to developing explanatory and yet useful models of

ecosystem function from a management point of view is to achieve a reasonable balance

between model complexity, parameter observability and biological realism. We feel that

an iterative approximation of experimental studies and modeling within an adaptive

management context is the most promising direction.

5. Conclusions

It has been shown that suitability of recommended forestry practices in such a

heterogeneous region is very site dependent. Firstly, we caution against extrapolation

from one site to another without a specific impact evaluation. It is possible that many

studies from temperate P. sylvestris regions must be interpreted with caution. Secondly,

differences in ecosystem function in nutrient cycling between a Mediterranean climate

(Aspurz) and a more continental climate (Garde), are in part overridden by the impact of

biomass extraction. However faster decomposition rates under Mediterranean

Blanco et al. (2005) 25/47

conditions can make these forests more sensitive to human intervention. Thirdly, current

recommendations seem adequate for N, K, Ca and Mg but their consequences on P

reserves should be investigated in more detail. Finally, the single most important

silvicultural factor in a thinning operation is the tree harvesting method. For all the

cases analyzed in this study whole-tree removal should not be used as it may

compromise long-term sustainability. Results from simple models of this sort for

evaluation of long-term effects on current silvicultural practices must be evaluated for

management effects on regeneration, harvesting age or size as well as forest operation

costs to compile with current economic demands and ecological sustainability of forest

resource management.

Acknowledgements

Juan A. Blanco was supported by a research grant from CICYT (‘Ministerio de Ciencia

y Tecnología’) during his stay at the University of Alcalá. We thank ‘Gobierno de

Navarra, Departamento de Educación y Cultura" for financial support and

‘Departamento del Medio Ambiente’ for experimental setting of silvicultural treatments

and financial support. In particular we acknowledge Fernando Puertas, Carmen Traver

and Ana Iriarte for assistance at several stages of this work. We are grateful to the

scientific forest network GLOBIMED (http://www.globimed.net) for hosting meetings

where the cooperation between our universities was initiated. We are also grateful to Dr.

J.P. Kimmins, Tanya Seebacher and two anonymous reviewers for their useful

suggestions and comments on the manuscript.

References

Blanco et al. (2005) 26/47

Aerts, R., 1997. Climate, leaf litter chemistry and leaf litter decomposition in terrestrial

ecosystems: a triangular relationship. Oikos 79, 439-449.

Aerts, R., Verhoeven, J.T.A., Whigham, D.F., 1999. Plant-mediated controls on nutrient

cycling in temperate fens and bogs. Ecology 80, 2170-2181.

Agren, G.I., 1983. Nitrogen productivity of some conifers. Can. J. For. Res., 13, 494-

500.

Agren, G.I., Bosatta, E., 1996. Theoretical Ecosystem Ecology. Understanding element

cycles. Cambridge University Press, Cambridge.

Agren, G.I., Knecht, M., 2001. Simulation of soil carbon and nutrient development

under Pinus sylvestris and Pinus contorta. For. Ecol. Manage. 141, 117-129.

Agren, G.I., McMurtrie, R.E., Parton, W.J., Pastor, J., Shugart, H.H., 1991. State-of-the-

Art of models of production-decomposition linkages in conifer and grassland

ecosystems. Ecol.Appl. 1, 118-138.

Andersson, F.O., Agren, G.I., Führer E., 2000. Sustainable tree biomass production.

For. Ecol. Manage. 132, 51-62.

Baeumler, R., Zech, W. 1998. Soil solution chemistry and impact of forest thinning in

mountain forests in the Bavarian Alps. For. Ecol. Manage. 108, 231-238.

Battaglia, M., Sands, P.J., 1998. Process – based forest productivity models and their

application in forest management. For. Ecol. Manage. 102, 13-32.

Beets, P.N., Robertson, K.A., Ford-Robertson, J.B., Gordon, J., Maclaren, J.P., 1999.

Description and validation of C_Change: a model for simulating carbon content in

managed Pinus radiata stands. N.Z. J. For. Sci. 29, 409-427.

Berg, B., Lousier, J. D., 2000. Litter decomposition and organic matter turnover in

northern forest soils. For. Ecol. Manage. 133, 13-22.

Blanco et al. (2005) 27/47

Blanco, J.A., Imbert, J.B., Arricibita, F.J., Castillo, F.J., 2003a. Producción de hojarasca

en bosques de Pinus sylvestris L. sometidos a diferentes intensidades de clara. Actas

del VII Congreso Nacional de la Sociedad Española de Ecología Terrestre,

Barcelona, pp. 267-279.

Blanco, J.A., Imbert, J.B., Ozcáriz, A., Castillo, F.J., 2003b. Decomposition and

nutrient release from Pinus sylvestris L. leaf litter in stands with different thinning

intensity. Proceedings of the I.U.F.R.O. meeting. Silviculture and sustainable

management in mountain forests in the western Pyrenees, Navarre, Spain.

Bolker, B.M., Pacala, S., Parton, W.J., 1998. Linear analysis of soil decomposition:

insights from the CENTURY model. Ecol. Appl. 8, 425-439.

Botkin, D. 2001. The role of theory in forest science and forest management.

Proceedings of Forest modelling for ecosystem management, forest certification and

sustainable management conference. University of British Columbia, 12 nd-18th

August, Vancouver.

Bray, J., Kurtz, H. 1945. Determination of total organic, and available forms of

phosphorus in soils. Soil Sci. 59, 39-45.

Chertov, O.G., Komarov, A.S., Bykhovets, S.S., Kobak, K.I., 2001. Simulated soil

organic matter dynamics of the Leningrad administrative area, northwestern Russia.

For. Ecol. Manage. 169, 29-44.

Cole, D.W., Rapp, M., 1981. Elemental cycling in forest ecosystems. In: Reide, D.E.

(Ed.) Dynamic properties of forest ecosystems. International Biological Programme

23. Cambridge University Press, Cambridge, pp. 341-409.

Cooke, G.W., 1981. The fate of fertilizers. In: Greenland, D.J., Hayes, M.H.B. (Eds.)

Soil chemistry of soil processes. John Wiley & Sons, Chichester, New York.

Blanco et al. (2005) 28/47

Costanza, R., Gottlieb, S., 1998. Modelling ecological and economic systems with

STELLA: Part II. Ecol. Model. 112, 81-84.

Del Río, M., Montero, G., 2001. Modelo de simulación de claras en masas de Pinus

sylvestris L. Monografías I.N.I.A. Forestal Nº 3, Madrid.

Díez, M.R., Fernández-Golfín, J.I., 1997. Influencia de las claras en la calidad de la

madera estructural de Pinus sylvestris L. en una masa artificial. Montes 48, 54-57.

Edmonds, R.L., 1990. Organic Matter decomposition in Western United States Forests.

In: Harvey, A.E., Neuenschwander, L.F. (Compolrs.). Proceedings-Management and

productivity of western-montane forest soils. GTR-INT-280, USDA, Forest Service,

Intermountain Research Station, Ogden:UT, pp. 118-128.

Fisher, F., Binkley, D. 2000. Ecology and Management of Forest Soils. John Wiley &

Sons. New York.

Friend, A.D., Shugart, H.H., Running, S.W., 1993. A physiology-based gap model of

forest dynamics. Ecology 74, 792-797.

Gilmanov, T.G., Parton, W.J., Ojima, D.S., 1997. Testing the ‘CENTURY’ ecosystem

level model on data sets from eight grassland sites in the former U.S.S.R.

representing a wide climatic/soil gradient. Ecol. Model. 96, 191-211.

Grigal, D.F., 2000. Effects of extensive forest management on soil productivity. For.

Ecol. Manage. 138, 167-185.

Grote, R., Suckow, F., Bellmann, K., 1998. Modelling of carbon-, nitrogen- and water

balances in Scots pine stands. In: Hüttl, R.F., Bellmann, K., (Eds.), Changes of

atmospheric effects on forest ecosystems. Kluwer Academic Publishers, London, pp.

251-281.

Hall, D. O., 1997. Biomass energy in industrialised countries – a view of the future. For.

Ecol. Manage. 91, 17-45.

Blanco et al. (2005) 29/47

Harmon, M.E., 2001. Carbon cycling in forests: simple simulation models. H.J.

Andrews Research Report Number 2. (http://www.fsl.orst.edu/lter/pubs/webdocs/

reports/ccycleforest.cfm).

Harrington, T.B., Edwards, M.B., 1999. Understory vegetation, resource availability,

and litterfall responses to pine thinning and woody vegetation control in longleaf

pine plantations. Can. J. For. Res. 29, 1055-1064.

Harwitte, H. (Ed.), 1980. Official methods of analysis of the Association of Official

Analytical Chemists. A.O.A.C., Washington D.C., pp. 127-129.

High Performance Systems Inc., 1997. STELLA research software, v 5.0. Hanover,

New Hampshire, USA.

Hunt, E.R., Running, S.W., 1992. Simulated dry matter yields for aspen and spruce

stands in the North American boreal forest. Can. J. Remote Sensing 18, 126-133.

Jacobson, S., Kukkola, M., Malkonen, E., Tveite, B., 2000. Impact of whole-tree

harvesting and compensatory fertilisation on growth of coniferous thinning stands.

For. Ecol. Manage. 129, 41-51.

Johnson, D.W., Todd, D.E., 1987. Nutrient export by leaching and whole-tree

harvesting in a loblolly pine and mixed oak forest. Plant and Soil 102, 99-109.

Kimmins, J.P., 2004. Forest Ecology. A foundation for sustainable management and

environmental ethics in forestry, 3rd Edition, Prentice Hall, New Jersey.

Kimmins, J.P., Mailly, D., Seely B., 1999. Modelling forest ecosystem net primary

production: the hybrid simulation approach used in FORECAST. Ecol. Model. 122,

195-224.

Kirschbaum, M.U.F., 1999. CenW, a forest growth model with linked carbon, energy,

nutrient and water cycles. Ecol. Model. 118, 17-59.

Blanco et al. (2005) 30/47

Klemmedson, J. O., Meier, C. E., Campbell, R. E., 1990. Litter fall transfers of dry

matter and nutrient in ponderosa pine stands. Can. J. For. Res. 20, 1105-115.

Komarov, A., Chertov, O., Zudin, S., Nadporozhskaya, M., Mikhailov, A., Bykhovets,

S., Zudina, E., Zoubkova, E., 2003. EFIMOD 2 - a model of growth and cycling of

elements in boreal forest ecosystems. Ecol. Model., 10, 373-392.

Landsberg, J., 2003. Modelling forest ecosystems: state of the art, challenges, and future

directions. Can. J. For. Res. 33, 385-397.

Litton, C.M., Ryan, M.G., Tinker, D.B., Knight, D.H., 2003. Belowground and

aboveground biomass in young postfire lodgepole pine forests of contrasting tree

density. Can. J. For. Res. 33, 351-363.

MAFF, 1986. The analysis of agricultural materials. Ministry of Agriculture, Fisheries

and Food. Reference Book 427, HMSO, London, pp. 156-157.

Mäkipää, R., Karjalainen, T., Pussinen, A., Kukkola, M., Kellomäki, S., Mälkonen, E.,

1998. Applicability of a forest simulation model for estimating effects of nitrogen

deposition on a forest ecosystem: Test of the validity of a gap-type model. For. Ecol.

Manage. 108, 239-250.

Malkönen, E., 1974. Annual primary production and nutrient cycle in some Scots pine

stands. Comm., Ins. For. Fenn. 84.5, 87 pp.

Margolis, H., Oren, R., Whitehead, D., Kaufmann, M.P., 1995. Leaf area dynamics of

conifer forests. In: Smith, W.K., Hinckley, T.M. (Eds.), Ecophysiology of coniferous

forests. Academic Press. San Diego.

Ministerio de Medio Ambiente, 2003. Deposición atmosférica en la estación de

Burguete (Navarra). Servicio de Protección Contra Agentes Nocivos en los Montes,

Madrid.

Blanco et al. (2005) 31/47

Montero, G., Ortega, C., Cañellas, I., Bachiller, A., Elena, R., San Miguel, A. 1999

Aboveground productivity and nutrient dynamics in a reforestation of Pinus pinaster

Ait. with different intensities of thinning. Investig, Agrar., Sist. Recur. For., Fuera de

Serie No. 1, 175-206.

Morris, D.M., Kimmins J. P., Duckert, D.R., 1997. The use of soil organic matter as a

criterion of the relative sustainability of forest management alternatives: a modeling

approach using FORECAST. For. Ecol. Manage. 94, 61-78.

Olson, J.S., 1963. Energy storage and the balance of producers and decomposers in

ecological systems. Ecology 14, 322-331.

Parton, W.J., Schimel, D.S., Cole, C., Ojima, D.S., 1987. Analysis of factors controlling

soil organic matter levels in great plains grasslands. Soil Sci. Soc. Am. J. 51, 1173-

1179.

Parton, W.J., Coughneour, M.B., Scurlock, J.M.O., Ojima, D.S., Gilmanov, T.G.,

Scholes, R.J., Schimel, D.S., Kirchner, T.B., Meneaut, J.-C., Seastedt, R., García

Moya, E., Kamnalrut, A., Kinyamario, J.I, Hal,l D.O., 1996. Global grassland

ecosystem modelling: development and test of ecosystem models for grassland

systems. In: Breymeyer, A.I., Hall, D.O., Melillo, J.M., Agren, G.I. (Eds.), SCOPE

56-Global change: effects on coniferous forests and grasslands. Chapter 11. John

Wiley & Sons. Chichester, New York, pp. 229-270.

Pérez-Batallón, P., Ouro, G., Merino, A., Macías, F., 1998. Descomposición de materia

orgánica, biomasa microbiana y emisión de CO2 en un suelo forestal bajo diferentes

manejos selvícolas. Bol. Soc. Esp. Cienc. Suelo, 5, 83-93.

Piene, H., Van Cleve, K. 1978. Weight loss of litter and cellulose bags in a thinned

white spruce forest in interior Alaska. Can. J. For. Res. 8, 42-46.

Blanco et al. (2005) 32/47

Potter, C., Bubier, J., Crill, P., Lafleur, P., 2001. Ecosystem modelling of methane and

carbon dioxide fluxes for boreal forest sites. Can. J. For. Res. 31, 208-223.

Puertas, F., 2003. Scots pine in Navarre: Economic importance and production.

Proceedings of the I.U.F.R.O. meeting Silviculture and sustainable management in

mountain forests in the western Pyrenees. 15th-19th September, Navarre, Spain.

Rolff, C., Agren, G.I., 1999. Predicting effects of different harvesting intensities with a

model of nitrogen limited forest growth. Ecol. Model. 118, 193-211.

Seely, B., Welham, C., Kimmins, H., 2002. Carbon sequestration in a boreal forest

ecosystem: results from the ecosystem simulation model, FORECAST. For. Ecol.

Manage. 169, 123-135.

Sverdrup, H., Svensson, M.G.E., 2002. Defining sustainability. In: Sverdrup, H.,

Stjernquist, I. (Eds.), Developing principles and models for sustainable forestry in

Sweden. Kluwer Academic Publishers, Dordecht, pp. 21-32.

Swift, M.H., Heal, O.W., Anderson, J.M., 1979. Decomposition in terrestrial

ecosystems. University of California, Berkeley.

Thornley, J.H.M., Cannell, M.G.R., 2000. Managing forests for wood yield and carbon

storage: a theoretical study. Tree Physiol. 20, 477-484.

Tiktak, A., Van Grinsven, H.J.M., 1995. Review of sixteen forest-soil-atmosphere

models. Ecol. Model. 83, 35-53.

Verbug, P.S.J., Johnson, D.W., 2001. A spreadsheet-based biogeochemical model to

simulate nutrient cycling processes in forest ecosystems. Ecol. Model. 141, 185-200.

Vitousek, P.M., Reiners, W.A., 1975. Ecosystem succession and nutrient retention: a

hypothesis. BioSci. 25, 376-381.

Waring, R.H., Running, S.W., 2001. Forest Ecosystems, analysis at multiple scales.

Academic Press, San Diego.

Blanco et al. (2005) 33/47

Zavala, M.A., Oria, J.A., 1995. Preserving biological diversity in managed forests: a

meeting point for ecology and forestry. Land. Urb. Plann. 31, 363-378.

Zhu, Z., Arp, P.A., Meng, F., Bourque, C.P.-A., Foster, N.W., 2003. A forest nutrient

cycling and biomass model (ForNBM) based on year-round monthly weather

conditions, part II: calibration, verification, and application. Ecol. Model. 170, 13-27.

Blanco et al. (2005) 34/47

 Tables

Table 1. Parameter values used in the model and source by nutrient and experimental site, A: Aspurz and G: Garde. After the first thinning mean

total tree biomass was 243,850 Kg ha-1 (P0), 181,404 Kg ha-1 (P20) and 164,868 Kg ha-1 (P30) in Aspurz, and 218,396 Kg ha-1 (P0), 158,265

Kg ha-1 (P20) and 135,813 Kg ha-1 (P30) in Garde (see experiment design section for treatment description).

Parameter N P K Ca Mg Units Reference

A G A G A G A G A G
Concentration
green needles 1.65 1.22 0.11 0.12 0.74 0.78 0.35 0.45 0.11 0.12 % Experimental
dry needles 0.87 0.83 0.05 0.07 0.28 0.33 0.58 0.67 0.08 0.10 % Blanco et al. (2003a)
branches 0.69 0.48 0.06 0.09 0.58 0.48 0.24 0.36 0.09 0.09 % Blanco et al. (2003a)
stems 0.09 0.04 0.01 0.01 0.09 0.06 0.15 0.10 0.03 0.03 % Experimental
soil 2225 2450 20 19 63 125 707 1614 89 164 ppm Experimental
Decomposition rate
needles 0.517 0.299 0.517 0.299 0.517 0.299 0.517 0.299 0.517 0.299 years-1 Blanco et al. (2003b)
branches 0.028 0.028 0.028 0.028 0.028 0.028 0.028 0.028 0.028 0.028 years-1 Agren and Bosatta (1996)
roots 5 5 5 5 5 5 5 5 5 5 % year-1 Mälkonen (1974)
Litterfall fraction
needles 39 38 39 38 39 38 39 38 39 38 % year-1 Experimental
branches 6 8 6 8 6 8 6 8 6 8 % year-1 Experimental
Ministerio de Medio
Atmospheric deposition 9.2 9.2 0.1 0.1 39.7 39.7 24.5 24.5 8.9 8.9 Kg ha-1 year-1
Ambiente (2003)
Kimmins (1997),
Weathering 0.0 0.0 0.2 0.2 2.0 2.0 25.0 25.0 4.0 4.0 Kg ha-1 year-1 Fisher and Binkley
(2000)
Ratio Above / below-
0.19 0.19 0.19 0.19 0.24 0.24 0.19 0.19 0.21 0.21 % Mälkonen (1974)
ground biomass
Initial above-ground
502 292 41 37 392 284 383 285 89 74 Kg ha-1 Experimental
nutrient pool
Initial soil nutrient pool 9758 12119 79 69 276 390 3134 7695 427 604 Kg ha-1 Experimental

Blanco et al. (2005) 35/47

Table 2. Results of verification regressions (R2) of observed data vs. simulated data for nutrient amount (kg ha-1) in five different fluxes (* =

P<0.01, ** = P<0.001).

Nutrient Extracted stems Extracted branches Extracted needles Leaf litterfall Woody litterfall

N 0.832** 0.866** 0.898** 0.725* 0.760**

P 0.821** 0.836** 0.900** 0.715* 0.761**

K 0.840** 0.862** 0.888** 0.720* 0.755**

Ca 0.831** 0.871** 0.891** 0.716* 0.761**

Mg 0.812** 0.860** 0.890** 0.723* 0.765**

Blanco et al. (2005) 36/47

Figure legends

Fig. 1. Model fluxes and nutrient pools. Nd: needle demand. Bd: branch demand. Sd: stem

demand. Rd: root demand. Ab: root uptake. Re: retranslocation. Wl: branch litterfall. Nl: needle

litterfall. Wm: branch mineralization. Nm: needle mineralization. Rm: root mineralization.

Fig. 2. Information fluxes followed to evaluate sustainability of thinning practices with our

model. Required inputs are shown in italics and model outputs in bold letters.

Fig. 3. Mass evolution of N, P, K, Ca and Mg in every pool along stand life in Aspurz. Left

column is a non-managed forest and right column is a simulation under recommended thinning

prescription (thinning of 30% basal area every 10 years, stem removal).

Fig. 4. Mass evolution of N, P, K, Ca and Mg in every pool along stand life in Garde. Left

column is a non-managed forest and right column is a simulation under recommended thinning

prescriptions (thinning of 20% basal area every 15 years, stem removal).

Fig. 5. Potential N losses caused by management (thinning + lixiviation) in the two experimental

sites, with stem (top) and whole-tree (bottom) removal as a function of thinning intensity and

rotation. The point labelled by a solid dot represents recommended thinning intensity and

frequency. The horizontal solid line represents accumulated external N inputs by deposition and

mineral weathering.

Fig. 6. Potential P losses caused by management (thinning) in the two experimental sites, with

stem (top) and whole-tree (bottom) removal as a function of thinning intensity and rotation. The

37
point labelled by a solid dot represents recommended thinning intensity and frequency. The

horizontal solid line represents accumulated external P inputs by deposition and mineral

weathering.

Fig. 7. Potential K losses by management (thinning + lixiviation) in the two experimental sites,

with stem (top) or whole-tree (bottom) removal as a function of thinning intensity and rotation.

The point labelled by a solid dot represents recommended thinning intensity and frequency.

Accumulated K inputs are 2836 kg ha -1 in Aspurz and 2627 kg ha-1 in Garde (beyond figure

limits).

Fig. 8. Potential Ca losses by management (thinning + lixiviation) in the two experimental sites,

with stem (top) or whole-tree (bottom) extraction as a function of thinning intensity and rotation.

The point labelled by a solid dot represents recommended thinning intensity and frequency.

Accumulated Ca inputs are 3366 kg ha-1 in Aspurz and 3069 kg ha-1 in Garde (beyond figure

limits).

Fig. 9. Potential Mg losses by management (thinning + lixiviation) in the two experimental sites,

with stem (top) or whole-tree (bottom) extraction, depending on thinning intensity and

frequency. The point labelled by a solid dot represents recommended thinning intensity and

frequency. Accumulated Mg inputs are 877 kg ha-1 in Aspurz and 813 kg ha-1 in Garde (beyond

figure limits).

38
Fig. 1.

39
Fig. 2

40
Fig. 3

41
Fig. 4

42
Fig. 5

43
Fig. 6

44
Fig. 7

45
Fig. 8

46
Fig. 9

47