prilagođeno pretraživanje po punom tekstu

ŠUMARSKI LIST 1-2/2011 str. 21     <-- 21 -->        PDF


UDK 630* 114.2 (001)



1 12



Bruno DE VOS , Primož SIMONČIČ

ABSTRACT: The data of 45 soil profiles from a 16 × 16 km grid across Slovenia
was analysed to develop a local pedotransfer function (PTF) for bulk density
(.b) estimation. In total, 106 soil horizons were considered. Concentration
of organic carbon (OC) was found to be well correlated (r = -0.861, p < 0.001)
with .b. Two separate line segments were fitted to the data, which was partitioned
into two intervals, based on OC content (below 36.0 g/kg and above

36.0 g/kg). Nearly 80 % of the variability in .b is explained with segmented regression.
The local PTF was compared with published PTFs and four validations
indices (MPE, SDPE, RMSPE and R2) confirmed the highest prediction
quality of the local PTF. The differences of carbon stock (Cpool) estimation, based
on usage of different PTFs could be higher than 160 t OC per hectare. Prediction
of carbon stocks could be substantially improved by calibration of the models
coefficients with data stratified according to each unique soil type.

Key words:pedotransfer function PTF, organic carbon OC, segmented
regression, forest soil, carbon stock Cpool

Since forest soil sampling and analyses of chemical Nimmo,2003), while PTFs for estimation of soil bulk
and physical properties of forest soils are time consu-density (.b)were introduced in the 1970s’ (e.g. Jeffrey,
mingand labor intensive, the development of alternative 1970).At first, bulk density was correlated only with
methods is indispensable. By using pedotransfer func-soil organic matter (SOM) (Adams,1973;Federer,
tions (PTFs), soil scientists are able to get information 1993;RawlsandBrakensiek,1985,Honeysett
on crucial soil properties, which are otherwise difficult andRatkowski,1989), but later the information on
(expensive or time consuming) toobtain. PTFs can be soil texture was added to some PTFs (Leonavičiute,
defined as statistical models for predicting soil physical 2000; Kauret al., 2002). Simple univariate models
(bulk density, soil hydraulic properties, etc.) and chemi-were supplemented with multiple regressions and diffecal
(e.g. cation exchange capacity) properties from other rent equations were developed separately for the organic
more available and routinely measured properties. and the mineral soil layers(e.g. Harrisonin Bocock,
1981), or even for different genetic soil horizons

The first PTF (for wilting coefficient) was develo

(e.g.Leonavičiute,2000). Recently, various techni

ped by Briggs and McLane 1907 (Landa and
ques of tree regressions were incorporated in PTFs de


Mr. sc. Milan Kobal, Slovenian Forestry Institute, Večna pot 2,

velopment (e.g.Martinetal., 2009).

SI-1000 Ljubljana,
Mihej Urbančič, dipl. inž., Slovenian Forestry Institute,
Soil bulk density (.b)is defined as the mass of a unit
Večna pot 2, SI-1000 Ljubljana,

volume of dry soil (105 °C), which includes both solids

Dr. sc. Primož Simončič, Slovenian Forestry Institute,

and pores and, thus, bulk density reflects the total soil

Večna pot 2, SI-1000 Ljubljana,



porosity (FAO, 2006). Usually, it is expressed in g/cm
Cvjetno naselje 41, HR-10450 Jastrebarsko,

Dr. sc. Nenad Potočić, Croatian Forest Research Institute,


or kg/dm. Soil bulk density is necessary for the asses


Mr. Bruno De Vos, Research Institute for Nature and Forest,

sment of soil carbon and nutrient pools (Tamminen

Gaverstraat 4, B-9500 Geraardsbergen, Belgium,

andStarr,1994) and for other mass-to-volume conver

ŠUMARSKI LIST 1-2/2011 str. 22     <-- 22 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY ... Šumarski list br. 1–2, CXXXV (2011), 19-27

sions. It is also needed when estimating soil water reten-cially recent studies evaluating existing PTFs (e.g. De
tion characteristics and is a required input parameter in Voset al., 2005; Martinet al., 2009) warn against
models of water, sediment and nutrient transport (Bouc-usage of PTFs without first testing their accuracy, and
neau et al., 1998).Additionally, soil bulk density is an in-stress the importance of local calibrations of coefficients
dicator of soil compaction, porosity and site productivity in the models.
(Tamminen andStarr,1994;Salifuetal., 1999).

The aim of our study was to develop a local PTF for

Several studies have investigated variation in forest the estimation of soil bulk density of(forest) mineral
soils properties at very detailed spatial scales (Phillips soils in Slovenia. Based on literature, we hypothesized
and Marion,2005; Scharenbrochand Bock -that (1) the bulk density.b correlated strongly with soil
heim 2007) and revealed that soil variability can be organic carbon concentration (OC) and (2) that our
high even on short distances and in small areas. Espe-local PTF perform better than published PTFs.

2 METHODS – Materijali i metode

2.1 Data sources and laboratory work – Izvori podataka i laboratorijski rad
The information on soil bulk
densities as well as physical and
chemical properties of soil horizons
wastaken from the soil database of
the Slovenian Forestry Institute
(SFI). Only the data on soil profiles
opened in year 2006 on the 16 × 16
km network across Slovenia were finally
selected; in total, 45 soil profiles
with 109 soil horizons (Figure
1). Summary information about soil
profiles is presented inTable 1.
Todescribe locations of the soil
profiles and evaluate morphological
and physical properties of the
soil horizons, FAO methodology
was followed (FAO, 2006).In each

soil horizon, separate soil samples
Figure 1 Locations of soil profiles (n = 45) across Slovenia from which the data for

were taken for bulk density estima

development of a local PTF for bulk density (.) estimation was derived.

tion and for chemical and physical

Slika 1. Položaj profila tla (n=45) u Sloveniji na osnovi kojih su dobivene lokalne

soil analysis. Samples for bulk den

pedotransfer funkcije (PTF) za procjenu gustoće tla.

sity estimation (ISO 11272) of a

fine earth fraction (< 2mm) were
obtained in five replicates by using
metal O-rings with volume of
5 cm.In the laboratory, soil samp les
were air dried (105 °C) and
weig hed. Variability of bulk density
estimation using metal O-rings
based on 5 replicates is presented in
Figure 2., where almost 80% of values
have a CVless than 10%. Soil
samples for chemical and physical
soil analysis were also air driedand
passed through a 2 mm sieve.The
fine earth fraction (< 2mm) wasre-
tained(UN/ECE ICP-Forests 2006,


Figure 2 Frequency distribution for coefficient of variation (CV) for bulk density measure

org/pdf/FINAL_soil.pdf) for furt


ments, obtained using 5 cm metal O-rings.

her chemical and physical analyses.

Slika 2. Distribucija frekvencija za koeficijent varijacije (CV) gustoća tala, izmjerenih ko-

The following methods were used:

rištenjem metalnih O-prstenova zapremine 5 cm3

ŠUMARSKI LIST 1-2/2011 str. 23     <-- 23 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY... Šumarski list br. 1–2, CXXXV (2011), 19-27

Table 1 Summary information of soil profiles included in the study, grouped
chloride following ISO 10390 on

pH was determined in calcium

according toWorld Reference Base soil reference groups (SRG)

Tablica 1.Zbirni podaci o profilima tla uključenim u studiju, grupiranima prema

automatic pH-meter Metrohm Ti-

WRB referentnim grupama tala (SRG)

trino, C and N content using dry

combustion using ISO 10694
and/or 13878 on Leco CNS-2000,
carbonates following ISO 10693
with Scheibler calcium-meter (Eijkelkamp)
and soil texture following
ISO 11277 with sedimentary method
and pipette according to Köhn.

Referentna grupa N
Soil depth, cm
Dubina tla, cm
Elevation, m
Nadmorska visina, m
tala prema WRB mean SD min max

2.2 Statistical analyses and model comparison

Statističke analize i usporedba modela

In total, 109 soil samples were included in the statistical
analyses.Three influential points (soil samples)
according to Cook’s distance were excluded from further
analysis.The simple and multiple regression models
were used to predict.b from different explanatory
variables. According to PTFs, developed by Hoekstra
and Poelman (1982), van Wallenburg
(1988) and Reinds et al. (2001), regression models
with segmented relationships were also tested. Only
variables that show statistical significance at the 0.05
level were included in the models. Models were compared
using partial F-test.

From the literature, four different published PTFs
were selected (Jeffrey, 1970; Harrison et al.,

1981; Tamminen, 1994; Kaur et al., 2002) using
following equations:



and Loss-On-Ignition method, the equation according
toCraftet al.(1991)was used:

The local PTF was compared with published PTFs
using four validation indices: mean predicted error
(MPE), standard deviation of the prediction error
(SDPE), root mean square prediction error (RMSPE)
and coefficient of determination(R2 ).These indices are
defined as:


where.bis soil bulk density (g/cm),OCis OC concentrationby
dry combustion method,LOIis organic


matter content (g/kg) by Loss-On-Ignition method, Where.b,i is measured bulk density of soil sam-
Clayis percentage of a clay fraction (0-2 µ) andSiltis ple,.bp,i is predicted bulk density ofith soil sample,nis
percentage of silt fraction (2-63 µ). For conversion of the number of soil samples,covis the covariance and
the data on OC obtained by dry combustion method varis the variance.

2.3 Carbon stock calculation (C )

Izračun zalihe ugljika (Cpool)

Carbon stock per given area, hectare in our case, WhereOCiis organic carbon concentration of ith soil

was calculated using following equation: horizon, is thickness of soil horizon (in m), is

di .b,i

ith 3

bulk densityof soil horizon (in g/cm),stoni is a corith

rection factor for stoniness in horizon and n is the
number of soil horizons for a given soil profile.

ŠUMARSKI LIST 1-2/2011 str. 24     <-- 24 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY ... Šumarski list br. 1–2, CXXXV (2011), 19-27
Statistical analyses were carried out using the R
2.9.3 software environment (R Development Core
Team, 2009). Package ‘’segmented’’was used to fitre-
gression models with segmented relationships(Mug-
3 RESULTS AND DISCUSSION – Rezultati i rasprava
3.1 Development of local PTF for predicting soil bulk density of mineral part of soil
Razvoj lokalnih pedotransfer funkcija za predviđanje gustoće mineralnog sloja tla
Soil bulk densityand concentration of organic car-
bon were strongly correlated (r = -0.86, p < 0.001).
Other chemical soil properties, except the concentra-
tion of total nitrogen (N), were less correlated with
bulk density (Figure 3).The correlation between bulk
density and base saturation (BS) and the correlation
between bulk density and clay content were not statisti-
cally significant (p > 0.05).
Figure 3 Relationship between bulk density (.b) and concentration of organic carbon (OC), concentration of total nitrogen (N), cation
exchange capacity (CEC), base saturation (BS) and clay content (Clay) for 106 soil samples.
Slika 3. Odnos gustoće tla i koncentracije organskog ugljika (OC), ukupnog dušika (N), pH, kapaciteta za izmjenu kationa (KIK),
sume baza (BS) i sadržaja gline (glina) za 106 uzoraka.
More than 73 % of the total variability of bulk den-
sity was explained byOC (model SFI 1,Table 2).Ad-
ding other chemical properties as explanatory variables
in the multiple regression models(modelsSFI 2, SFI 3,
Table 2 Regression relationship between soil properties as predictors and bulk density as response for 106 soil horizons
(OC – organic carbon, BS – base saturation, CEC – cation exchange capacity, CLAY- clay content.
Tablica 2.Regresijski odnos karakteristika tla kao prediktora i gustoće tla kao odziva za 106 horizonata tla (OC – organski
ugljik, BS – suma baza, KIK - kapacitet za izmjenu kationa, glina – sadržaj gline).
Response variable
Intercept OC pH* BS*
SE Adj. R
Varijabla odziva KIK glina
SFI 1 .b 1.3983 -0.0734 0.1403 0.7384
SFI 2 .b 1.4509 -0.0720 -0.0115 0.1404 0.7379
SFI 3 .b 1.3752 -0.0749 0.0004 0.1399 0.7398
SFI 4 .b 1.3902 -0.0788 -0.0011 0.1402 0.7385
SFI 5 .b 1.3438 -0.0734 0.0019 0.1390 0.7431
.b for OC < 3.6 % 1.4842 -0.1424
0.1257 0.7958.b for OC . 3.6 % 1.1253 -0.0452
* denotes not statistically significant variable in the model
* označava nesignifikantnost varijable u modelu
Bulk density/gustoća tla [g/m3
Bulk density/gustoća tla [g/m3
Bulk density/gustoća tla [g/m3
Bulk density/gustoća tla [g/m3
Bulk density/gustoća tla [g/m3
Bulk density/gustoća tla [g/m3

ŠUMARSKI LIST 1-2/2011 str. 25     <-- 25 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY... Šumarski list br. 1–2, CXXXV (2011), 19-27
Figure4 Segmented regression relations-
hip between soil bulk density (.
and organic carbon content (OC)
in the mineral soil [g/kg]
Slika4. Segmentirana regresija odnosa
gustoće tla (.
) i sadržaj organ-
skog ugljika (OC) u mineralnom
sloju tla [g/kg]
SFI 4 and SFI 5 inTable 2) did not significantly im-
prove the prediction of SFI 1 (partial F-test, p > 0.05).
Unexpectedly, soil texture, especially clay content, was
not statistically significant variable in the models;
contrary tomany studies revealing that clay content is
related with soil bulk density (Kaur, 2002; Leo na-
The segmented regression method (SFI 6) improved
prediction of .b (partial F-test, p < 0.001). The inde -
pendent variable OC was partitioned into two intervals
and a separate line segment was fitted to each interval.
The boundary between two segments (breakpoint) was
36.0 g/kg OC (Figure 4).Nearly 80 % of the total va riabi-
lity in.b was explained by using segmented regression.
Figure 5 Evaluation indices for published PTFs and local PTF (SFI 6)
Slika 5. Indeksi evaluacije za objavljene i lokalni PTF (SFI 6)
3.2 Validation of local and published PTFs for mineral part of soil
Validacija lokalnih i objavljenih pedotransfer funkcija za mineralni sloj tla
For the validation of local and published PTFs,bulk
density was calculated using four published PTFs and
predicted values were compared with local PTF (SFI
6).The prediction quality of all five PTFs is presented
in Figure 5.All four validation indices confirmed the
highest prediction powerof our local PTF (Figure 5)by

ŠUMARSKI LIST 1-2/2011 str. 26     <-- 26 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY ... Šumarski list br. 1–2, CXXXV (2011), 19-27
Figure 6 Performance of two local PTFs and published PTFs for the total dataset: estimated versus observed bulk densities with
references to the 1:1 line.
Slika 6. Kvaliteta predviđanja dvije lokalne i objavljenih pedotransfer funkcija za ukupni zbir podataka: procijenjene u odnosu
na izmjerene gustoće s linijama izjednačenja.
3.3 Carbon stock calculation using different PTFs
Izračun zalihe ugljika korištenjem različitih pedotransfer funkcija
Carbon stock (C
) per hectare was calculated for
different soil profiles, based on the usage of different
PTFs (Table 4). Four different soil profiles were ran-
domly selected from our soil databaseof the 16 × 16 km
grid: Zajama, Lubnik, Besnica and Merljaki (Table 3).
In the calculation of C
, the stone content in soil hori-
zons was considered, while the root portion was not.We
assumednosurface rock outcrops.
Soil profile “Zajama” was excavated in the Pokljuka
plateau and is classified as Leptosol, soil profile “Lub-
nik” was dug near Škofja Loka and is classified as Cam-
bisol, profile “Besnica” was excavated near Ljubljana
and is classified also as Cambisol, whereas soil profile
“Merljaki” is classified asAcrisol and was excavated
near Nova Gorica. Morphological, physical and chemi-
cal properties are presented in detail inTable 3.
The calculation of the C
, based on the PTF of
Kaur etal. (2002)gives highly underestimated values
for all four soil profiles.The differences between cal-
culated C
using PTF SFI 6and measured C
are not
unambiguous, i.e. for soil profile “Lubnik” and “Be-
snica” the carbon stock is underestimated, while for
soil profile ‘’Zajama’’carbon stock is overestimated.
The calculations of C
revealed that differences of
calculated carbon stock per hectare could be quite large
and arestrongly dependent upon the PTFs algorithm.
However, the lowest difference of the C
based on
measured and calculated bulk density was found for
profile ‘’Merljaki’’. Both chemical and physical pro-
perties of this profile are close to average soil proper-
ties, included in this study, i.e. lower OC concentration
and high bulk density (Figure 3). Consequently, the
having the lowest value ofbias of the regression model
(MPE), the lowest random variation of the predictions
after correction for the global bias (SDPE),the lowest
overall error of the predictions (RMSPE) and the hig-
hest coefficient of determination (R2 ).
In the case of high bulk density, local SFI 6 PTF
seems slightly less accurate (Figure 6). Probably, that
could be explained because of not including informa-
tion onclay content, which is normally the highest just
forthe soil horizons with high bulk densities (Urban-
čič et al., 2005).For other PTFs the systematic error in
predictions is evident from the scatterplots of Figure 6.

ŠUMARSKI LIST 1-2/2011 str. 29     <-- 29 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY... Šumarski list br. 1–2, CXXXV (2011), 19-27

Green World Research, Wageningen, The Netherlands,
55 pp.

Salifu, K. F., W. L. Meyer,and H. G. Murchison,
1999. Estimating soil bulk density from organic
matter content, pH, silt and clay. J. Tropic.
For. 15:112–120.

Scharenbroch, B.C., J. G.Bockheim,2007. Pedodiversity
in an old-growth northern hardwood
forest in the Huron Mountains, Upper Peninsula,
Michigan. Canadian Journal of Forest Research
37, 1106–1117.

Tamminen,P.and M.Starr,1994. Bulk density of
forested mineral soils. Silva Fennica 28:53–60.

Urbančič, M,P.Simončič,T.Prus and L.Kutnar,
2005.Atlas gozdnih tal Slovenije. Zveza
gozdarskih društev Slovenije: Gozdarski vestnik:
Silva Slovenica: Gozdarski inštitut Slovenije,

Van Wallenburg,C.,(1988)The density of peaty
soils (in Dutch). Internal Report, Soil Survey Institute,
Wageningen, The Netherlands, 5 pp.

SAŽETAK: S obzirom na vremensku zahtjevnost i veliku količinu rada potrebnog
za uzorkovanja i analize kemijskih i fizikalnih svojstava šumskih tala,
razvoj alternativnih metoda je vrlo važan. Korištenjem pedotransfer funkcija
(PTF), znanstvenici koji se bave proučavanjem tala mogu dobiti informaciju o
najvažnijim svojstvima tala koja je inače teško (skupo ili vremenski zahtjevno)
dobiti. PTF se mogu definirati kao statistički modeli za predviđanje fizikalnih
(gustoća, hidraulička svojstva, itd.) i kemijskih (npr. kapacitet za izmjenu kationa)
svojstava tla iz drugih, dostupnijih ili rutinski analiziranih svojstava.

Cilj ovog rada je bio razviti lokalnu PTF za procjenu gustoće mineralnog dijela
šumskih tala Slovenije. Na osnovi literature, hipoteza je bila da (1) gustoća
snažno korelira s konce4ntracijom organskog ugljika (OC) i (2) lokalna PTF
daje bolčje vrijednosti od objavljenih pedotransfer funkcija.

Podaci 45 profila tla s bioindikacijske 16 x 16 km mreže u Sloveniji su analizirani
s ciljem razvijanja lokalne pedotransfer funkcije za procjenu gustoće tla.
Ukupno je obrađeno 106 profila tla.

Uzorci za procjenu gustoće tla uzeti su u pet ponavljanja korištenjem metalnih
O-prstenova zapremine 5 cm3. U laboratoriju su uzorci tla osušeni na 105 °C
i izvagani za daljnje kemijske i fizikalne analize. Korištene su sljedeće analitičke
metode: pH je određen u KCl prema ISO 10390 na automatskom ph-metru Metrohm
Titrino, sadržaj C i N je određen prema ISO 10694 i/ili 13878 na elementarnom
analizatoru Leco CNS-2000, karbonati prema ISO 10693 Scheiblerovim
kalcimetrom a mehanički nsastav tla prema ISO 11277 sedimentnom metodom i
pipetom prema Köhnu.

Jednostavna i multipla regresija korištene su za predviđanje .b korištenjem
različitih zavisnih varijabla, a testirani su također i regresijski modeli sa segmentnim

Koncentracija organskog ugljika (OC) dobro korelira (r = -0.861, p < 0.001)
s gustoćom tla. Dva odvojena segmenta linije izjednačenja uklopljeni su u podatke
koji su razdijeljeni u dva intervala prema sadržaju OC (ispod i iznad
36,0 g/kg). Gotovo 80 % varijabiliteta gustoće tla objašnjeno je segmentnom regresijom
(Slika 4.).

Lokalna pedotransfer funkcija uspoređena je s objavljenim funkcijama a četiri
indeksa validacije (MPE, SDPE, RMSPE and R2) potvrdila su najveću kvalitetu
predviđanja lokalne pedotransfer funkcije (Slika 5.).

Razlike u procjeni zalihe ugljika u tlu (Cpool) različitih pedotransfer funkcija
bile su veće od 160 t/ha (Tablica 4.). Predviđanje zaliha ugljika moglo bi biti
značajno unaprijeđeno kalibracijom koeficijenata u modelima pomoću podataka
razvrstanih prema vrsti tla.

Ključne riječi:pedotransfer funkcija PTF, organski ugljik OC, segment na
regresija, šumsko tlo, zaliha ugljika Cpool

ŠUMARSKI LIST 1-2/2011 str. 27     <-- 27 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY... Šumarski list br. 1–2, CXXXV (2011), 19-27

Table 3 Morphological, physical and chemical properties of four soil profiles for evaluating C estimations.


Tablica 3.Morfološka, fizikalna i kemijska svojstva četiri profila tla za ocjenu kvalitete predviđanja zalihe ugljika

Horizon boundary
Granica horizonta
Upper Lower
Gornja Donja
Physical soil properties
Fizikalna svojstva tla
Clay Silt Sand Stoniness
Glina Prah Pijesak Kamenitost .b
Chemical soil properties
Kemijska svojstva tla
-Cm Cm % % % % g/cm
% % -cmol/kg %
Zajama AC
0 13
13 33
9.9 48.2 42.0 15
20.2 57.0 22.8 40
9.47 0.78 6.99 83.89 100
6.52 0.56 7.24 68.28 100
Lubnik AC
0 15
15 45
32.5 51.9 15.6 15
40.7 43.1 16.1 40
10.70 0.73 7.13 66.72 100
8.47 0.68 7.18 58.05 100
Besnica A
0 3
3 29
29 49
49 82
20.4 27.7 51.9 5
12.2 40.3 47.5 8
18.1 36.1 45.8 30
18.9 34.0 47.2 65
8.38 0.5 3.47 12.62 45.7
1.06 0.06 3.89 5.39 15.7
0.45 0.03 4.01 4.16 10.8
0.49 0.03 4.00 4.43 12.5
Merljaki A
0 10
10 47
47 85
85 122
25.4 40.8 33.8 5
21.5 50.6 27.9 10
26.5 43.2 30.3 13
32.9 46.5 20.6 13
6.86 0.46 3.69 10.95 40.3
0.67 0.06 3.81 7.18 6.6
0.63 0.05 3.84 5.95 5.6
0.69 0.05 3.95 5.65 12.3

Table 4 Estimated carbon stock (C ) tilldepth of parent material for four different soil profiles based on measuredand


calculated bulk densities.

Tablica 4.Zaliha ugljika do dubine matičnog supstrata za četiri profila tla, procijenjena na osnovi izmjerenih i izračunatih
gustoća tla.

bulk density
gustoća tla
Carbon stock C
in t/ha
Zaliha ugljika Cpool u t/ha
Bulk density calculated using PTF
Gustoća tla izračunata pomoću PTF
SFI 6 Jeffrey Harrison Tamminen Kaur

model is nicely predicting the.bof the profile ‘’Merljaki’’,
whereas differences for other soil types are larger;
i.e. even higher than 25t of OC per hectare (profile
‘’Zajama’’).Using non local PTFs drawn from litera

ture may resulted in high differences between measured
and calculated C up to 160 t of OC per hectare


(profile ‘’Lubnik’’, PTF Kaur).

4. CONCLUSIONS – Zaključci

Using national data from a 16 × 16 km plot network,
we developed a pedotransfer functionfor bulk density
of mineral forest soils of Slovenia. Most of the variability
in soil bulk density can be explained by concentration
of organic carbon. Adding other chemical (pH, N,
CEC, BS) and physical soil properties (soil texture) in
the regression equation did not significantly improve
the prediction quality.The prediction quality of all five
PTFs (Jeffrey, Harrison, Tamminen, Kaur and local SFI
6) were tested using four validation indices (MPE,
SDPE, RMSPE,R2 ), the result being that local PTF SFI

6 gives the most accurate prediction of soil bulk density.

The PTFs were also used for prediction of carbon

stocks in forest soils. Unexpectedly, using the local

PTF SFI 6 could still lead to possible inaccuracies of
the C calculation higher than 25t of OCper hectare.


Weassume that the main reason for that is a high pedodiversity
of Slovenian forest soils, requiring additional
soil .b sampling, especially for the main forest soil

ŠUMARSKI LIST 1-2/2011 str. 28     <-- 28 -->        PDF

M. Kobal, M. Urbančič, N. Potočić, B. De Vos, P. Simončič: PEDOTRANSFER FUNCTIONS FOR BULK DENSITY ... Šumarski list br. 1–2, CXXXV (2011), 19-27


The study was financed by the Ministry of Higher
Education, Science and Technology, through the research
programme P4–0107 and young researcher programme,
BioSoil Module Soil (2006-2008) within
Forest Focus programme, national CRP project
V2-0511. Special thanks to Damijana Kastelec for
methodological improvement.

6. REFERENCES – Literatura

Adams,W.A., 1973. Effect of Organic-Matter on Bulk
andTrue Densities of Some Uncultivated Podzo

lic Soils. Journal of Soil Science 24, 10–17.

UN/ECE ICP-Forests2006.Annex 1, Methods for Soil
Analysis. ICPForests, IIIa Sampling andAnalysis
of Soil – update 2006. http://www.icp-forests.

Benites,V.M., P.L. O.A.Machado, E. C. C.Fidalgo,
M. R.Coelho, B. E.Madari,2007.
Pedotransfer functions for estimating soil bulk
density from existing soil survey reports in Brazil.
Geoderma 139, 90–97.

Boucneau,G., M. Van Meirvenne and G.Hofman,
1998. Comparing pedotransfer functions
to estimate soil bulk density in northern Belgium.
PedologieThemata 5:67–70.

Craft, C.B., E. D.Seneca, S.W.Broome,1991.
Loss on Ignition and Kjeldahl Digestion for Estimating
Organic-Carbon and Total Nitrogen in
Estuarine Marsh Soils – Calibration with Dry
Com bustion. Estuaries 14, 175–179.

DeVos, B., M. VanMeirvenne,P.Quataert, J.
Deckers, B.Muys,2005. Predictive quality
of pedotransfer functions for estimating bulk
density of forest soils. Soil Science Society of
America Journal 69, 500–510.


FAO. 2006. Guidelines for soil description. 4 edition.
FAO, Rome.

Federer, C. A., D. E. Turcotte, C. T. Smith,
1993.The Organic Fraction – Bulk-Density Relationship
and the Expression of Nutrient Content
in Forest Soils. Canadian Journal of Forest
Research 23, 1026–1032.

Harrison,A.F., K. L.Bocock,1981. Estimation of
Soil Bulk-Density from Loss-On-Ignition Values.
Journal ofApplied Ecology 18, 919–927.

Heuscher, S.A., C. C. Brandt, P. M. Jardine,
2005. Using soil physical and chemical properties
to estimate bulk density. Soil Science Society
ofAmerica Journal 69, 51–56.

Hoekstra C, Poelman JNB (1982) Density of soils measured
at the most common soil types in the Ne

therlands (in Dutch). Report 1582, Soil Survey

Institute,Wageningen, The Netherlands, 47 pp.

Honeysett, J.L., D.A.Ratkowsky,1989. TheUse

of Ignition Loss to Estimate Bulk-Density of Forest
Soils. Journal of Soil Science 40, 299–308.

Jeffrey, D.W.,1970.ANote on Use of Ignition Loss
asAMeans forApproximate Estimation of Soil
Bulk Density. Journal of Ecology 58, 297–&.

Kaur,R., S.Kumar, H.P.Gurung,2002. Apedotransfer
function (PTF) for estimating soil bulk
density from basic soil data and its comparison
with existing PTFs. Australian Journal of Soil
Research 40, 847–857.

Landa, E.R., J. R.Nimmo,2003.The life and scientific
contributions of Lyman J. Briggs. Soil
Science Society ofAmerica Journal 67, 681–693.

Leonavičiute, N.,2000. Predicting soil bulk and particle
densities by soils pedotransfer functions
from existing soil data in Lithuania. Geografijos
metrastis 33:317–330.

Martin, M.P., D.Lo Seen, L.Boulonne, C.Jolivet,
K. M. Nair, G. Bourgeon, D. Arrouays,
2009. Optimizing Pedotransfer
Functions for Estimating Soil Bulk Density
Using Boosted Regression Trees. Soil Science
Society ofAmerica Journal 73, 485–493.

Muggeo,V.M.R., 2008. Segmented: an R package
to fit regression models with broken-line relationships.
Rnews 8/1, 20–25.

Phillips, J.D., D.A. Marion, 2005. Biomechanical
effects, lithological variations, and local pedodiversity
in some forest soils of Arkansas. Geoderma
124, 73–89.

R Development CoreTeam (2009). R: Alanguage and
environment for statistical computing, reference
index version 2.9.1. R Foundation for Statistical
Computing, Vienna, Austria. ISBN 3-90005107-
0, URL

Rawls,W.J., and D. L.Brakensiek,1985. Prediction
of soil water properties for hydrologic modelling.
p. 293–299. In E. Jones and T.J. Ward
(ed.) Watershed Management in the Eighties.
Proc. of a Symp.ASCE, Denver, CO. 30 Apr.–
2 May 1985.ASCE, NewYork.

Reinds, GJ, M. Posch, W. De Vries, (2001) A
semi-empirical dynamic soil acidification model
for use in spatially explicit integrated assessment
models for Europe.Alterra Report 084,Alterra