224 Erosion Prediction in Ungauged Basins: Integrating Methods and Techniques (Proceedings of symposium

HS01 held during IUGG2003 at Sapporo, July 2003). IAHS Publ. no. 279, 2003.

Applicability of the Gavrilović method in erosion

calculation using spatial data manipulation
techniques

LIDIJA GLOBEVNIK1, DANKO HOLJEVIĆ2

GREGOR PETKOVŠEK1 & JOSIP RUBINIĆ2
1 University of Ljubljana, Faculty for Civil and Geodetic Engineering, Hajdrihova 28,
1000 Ljubljana, Slovenia
[email protected]
2 Croatian Waters VGO, Rijeka, Croatia

Abstract The Gavrilović method has been widely used for the prediction of
soil erosion and sediment yield on the basin scale in Slovenia and Croatia in
the last 30 years. The method has been developed for management practices in
erosion protection, mainly in forest management and torrent control. The
Gavrilović method involves a parametric distributed model, and is used for
predicting annual soil erosion rates and annual sediment yield. It uses
empirical coefficients (erodibility coefficient, protection coefficient, erosion
coefficient) and a matrix of physical characteristics of basin sub units. To
evaluate its applicability we analysed erosional processes in two neighbouring
drainage basins, the Dragonja River basin in Slovenia and the Botonega River
basin in Croatia. The results of this study show that GIS techniques can be
successfully incorporated into the Gavrilović method.
Key words erosion; sediment yield; Dragonja River; Botonega River; Gavrilović

INTRODUCTION
Erosion control is one of the biggest concerns in integrated water management. Soil
erosion is a natural process, but it can be accelerated by certain human activities. Agri-
culture, where soil loss also represents a loss in nutrients and productive capability,
construction, and mining are the most important (Hahn et al., 1994). In the first place,
soil erosion causes land degradation, but excessive sedimentation resulting from it can
also have an impact on river morphology. Another common problem is the pollutants,
eroded and transported alongside with the soil particles. If they are deposited in a
marine environment, they can cause serious pollution problems (Horvat et al., 1998).
To propose erosion control and stream regulation measures in a watershed, managers
should have an understanding of the processes of soil detachment, transport and
deposition. Knowing trends of those processes is equally important. Science has
always tried to offer practical methods for the evaluation of erosional processes for
managerial purposes. The first empirical soil erosion models were proposed by Cook
(1936), Zingg (1940) and Smith (1941). As quoted in Gavrilović (1972), in the late
1940s and early 1950s, methods for analysing the severity of erosion were developed.
Poljakov (1953) proposed an analytical expression with two parameters, average
annual water turbidity and slope of the basin, to determine the erosion development
coefficient. As these parameters should be regularly measured in the field, this method
has not been widely used. Herheulidze (1947) proposed four categories for the state of
 Applicability of the Gavrilović method in erosion calculation 225

torrent development to describe the degree of erosion in drainage basins with debris
torrents. For this categorization, data on slope and some geological parameters of
deposited material, are needed. At the same time, Browning (1947) proposed a relative
erosion factor (Re) as a measure of soil loss due to water erosion. Silvestrov (1955)
proposed an analytical equation for the determination of an ‘erosion coefficient’ (E)
for the analysed area. In this equation, some parameters relating to land use and relief
are needed. On the basis of the described theoretical work, Gavrilović (1962, 1970,
1972) developed a method for the analytical determination of erosion coefficients and
the quantification of erosion and average annual sediment yield. Using field
investigations on the Morava River (Serbia), and laboratory experimental work,
Gavrilović (1976) prepared detailed tables for the determination of parameters. This
method has been widely used in Slovenia and Croatia in the last 30 years to predict
erosional processes and implement torrent regulation and other erosion control works.
In this paper, we evaluate the applicability of the Gavrilović method for analysing
erosional processes in two sub-Mediterranean basins: the Dragonja River basin in
Slovenia and the Botonega River basin in Croatia. Since an array of empirical
coefficients based on the physical characteristics of the basin is used in the method, we
describe GIS techniques for data preparation and modelling.

METHODS

Gavrilović (1962, 1970, 1972) proposed an analytical equation for determining the
annual volume of detached soil due to surface erosion:

Wp = π * P * Fw * Kt * K z (1)
3

where P is the average yearly precipitation (mm), Fw is the drainage area (km2), Kt is
the temperature coefficient, and Kz is the erosion coefficient. The equation for Kt is as
follows:

Kt = (0.1 + T ) (2)
10
where T is the average yearly temperature (°C). The erosion coefficient Kz can be
estimated using corresponding tables or calculated from:
Kz = Ky * Kx * (Ko + Fsl ) (3)

where Fsl is the average slope of the basin (%), Ky is the soil erodibility coefficient, Kx
is the soil protection coefficient and Ko the erosion and stream network development
coefficient. Gavrilović suggested calculating the average slope as the total contour line
length in the analysed basin multiplied by the contour interval divided by the drainage
area. If a digital elevation model is available, the average slope is simply calculated as
the average slope of the cells in the basin. When the drainage basin is not uniform with
respect to the erosion coefficients, Gavrilović suggested that the basin should be
divided into smaller sub areas (hydrographic units). After the annual soil erosion rates
Wp are calculated for each hydrographic unit, they are summed to obtain the soil
erosion rate for the whole basin.
226 Lidija Globevnik et al.

Gavrilović (1976) has suggested the following equation for determination of the
sediment delivery ratio:
(O * D) 0.5
Ru =
0.25 * ( L + 10) (4)
where O represents the perimeter of the basin (or sub unit) (km), D is the average
height distance of the basin (or sub unit), expressed in (km), and L length of the basin
(km). Average height distance is calculated as:
n

åf i * hi
D= i =1
− H min (m) (5)
Fw
where fi represents area between two contour lines (km2), hi average altitude between
the contour lines (m), and Hmin the minimal altitude of the basin (m). The actual
sediment yield is then calculated as:
G y = Ru * Ws (m3 year-1) (6)

DESCRIPTION OF THE RESEARCH AREA

The research basins are located on the Istria peninsula along the North Adriatic Sea
(Fig. 1). The Dragonja River basin has an area of 91 km2; the river is 29 km long and
flows from east to west. The river at its lower section forms the state border between
Slovenia and Croatia, and much of its drainage basin lies in Slovenia. The Botonega
River is 24 km long, and has a 98-km2 drainage area, 73 km2 of which drains into the

Fig. 1 Map of the research area, showing the Dragonja and Botonega River basins.
 Applicability of the Gavrilović method in erosion calculation 227

Botonega Reservoir. It is a tributary of the Mirna River, which is one of the main
rivers in the Istria Peninsula in Croatia. The bedrock in both basins consists of flysch
substratum, impermeable and easily erodible, being composed of sandstone and
marlstone with some limestone. The climate is sub-Mediterranean with dry, warm
summers and wet winters. The annual precipitation of the area is 900 mm near the
coast (west) and 1200 mm in the inland portions of the peninsula. The daily maximum
rainfall in the area is 175 mm. The annual average temperature is 12°C, with a summer
and winter average of 21 and 4°C, respectively.

DATA PREPARATION AND GIS MANIPULATION TECHNIQUES

A study of erosion processes using the Garilović method was undertaken in the
Dragonja River basin in 1971. An evaluation of the average annual sediment yield in
1971 was performed for sub-basin units. On the basis of this study, erosion control
measures have been applied in the basin since the mid 1970s. After abandonment of
agricultural land and subsequent natural reforestation during the period 1975–1995, the
extent of forest cover increased from 25% to 65% over this period (Globevnik et al.,
1998; Globevnik, 2001). To compare changes in erosion between 1971 and 1995, the
same approach as used in the 1971 study was applied. In Fig. 2, the Dragonja River
basin is shown with the sub areas (units) used for the calculations in the 1971 study.
To analyse the suitability of the Gavrilović method for use with GIS techniques,
we prepared cartographic data on geology, pedology and land use in digital form.
Geological maps with a scale of 1:100 000 have been digitized and prepared in
AutoCad for the Dragonja River basin and in ArcView for the Botonega River basin.
The same was done for pedology maps, which are available at a scale of 1:25 000. The
Corine Land Cover Data (European Environmental Agency, 1999) have been used to
determine Kx coefficients for both basins. For the upper part of the Dragonja River
basin (NE part, Rokava area), the Corine Land Cover map (source: http://nfp-
si.eionet.eu.int/ewnsi/index.htm), representing the situation in 1995, and pedology are
shown in Fig. 2. The corresponding Ky and Kx coefficients, derived from data on
pedology and land cover data, are given in Table 1.
The applicability of the Gavrilović method in the AutoCad Map GIS environment
has been tested in the 20.4 km2 Rokava sub-basin (Fig. 1), which is part of the

Table 1 Values for soil erodibility coefficient, Ky and soil protection coefficient, Kx.
Selected pedological units Ky Selected Corine land cover classes Kx
Carbonatic rendzina on flysch 1.2 1—artificial surfaces; 5.1—inland water 0
Eutric brown soil on flysch 0.9 3.1.1—broad-leaved forest; 0.05
3.1.3—mixed forest
Eutric brown soil on flysch, pseudogleic 1.1 2.4— heterogeneous agricultural areas 0.4
Eutric brown soil, colluvial, on flysch 0.2 3.2.4—transitional woodland shrub 0.5
Fluvisol on carbonatic alluvium 0.2 2.3.1—pastures; 3.2.1—natural grassland 0.6
2.2—permanent crops 0.7
2.1—arable land 0.9
3.3.2—bare rocks; 3.3.6—areas under erosion 0.95
 228 Lidija Globevnik et al.

Dragonja River basin. The thematic layers, representing coefficients Ko, Ky and Kx,
were overlayed and average slopes for each new polygon were determined. For this
purpose we imported Slovene digital elevation data (raster 25 m × 25 m) into terrain
modelling software (Quicksurf) and determined the average slope for each polygon on
the Rokava basin. A similar procedure was used in ArcView for the Botonega basin.
The slope was calculated from the Croatian digital elevation data in the grid of 50 m ×
50 m, using the terrain modelling software package Spatial Analyser.

RESULTS

Erosion rates for the Dragonja basin as calculated for 1971 and 1995 in the traditional
way are given in Table 2 (Globevnik, 2001). Rates were calculated for 29 hydrograp-
hic sub areas (flow direction considered), using one value of each parameter for one
sub area. For 1971, specific rates per unit area varied between 200 and 550 m3 km-2,
with the average being 325 m3 km-2. For 1995, rates varied between 30 and 310 m3 km-2,
with an average rate of 130 m3 km-2. The result of the overlay analysis of Corine land
cover and pedological data for the 20.4 km2 large Rokava sub-basin (Fig. 3) is 74
polygons. The sum of erosion rates for the Rokava sub-basin, calculated using
automatically-derived Gavrilović coefficients, is 1213m3 year-1 (Table 3). The average
erosion rate is marginally lower when calculated with spatially distributed parameters

Table 2 Erosion rates in 1971 and 1995 for the Dragonja River basin (for units, Fig. 2).
Units F Fsl P T Kt Ky Kx Ko Kz Wp (m3 year-1)
(km2) (%) (mm) (°C) 1971 1995 1971 1995 1971 1995 1971 1995
Sum of units 1 to 12: 9945 3515
13 13.4 23.7 1000 12 1.14 0.64 0.35 0.2 0.2 0.12 0.154 0.078 2895 1039
18 0.66 37.5 1000 12 1.14 0.6 0.35 0.2 0.2 0.12 0.171 0.088 166 62
Rokava (sum of units 13 to 18): 4690 1597
Sum of units 19 to 29 13 534 5990
Sum of all units (F = 86.8 km2) 28 168 11 102

Table 3 Examples of erosion rate calculation with automatically derived Gavrilović coefficients for the Rokava
and Jukani test areas and sediment yield for the Jukani.
F P T Kt Ky Kx Ko Kz Wp (m3 O L D Ru G (m3
2
(km ) (mm) (°C) year-1) (km) (km) (km) year-1)
Calculation for units on the Rokava test area (Slovenia, on the Dragonja basin), Fig. 3:
251 0.18 1000 12 1.14 0.1 0.05 0.12 0.002 0.1
309 0.26 1000 12 1.14 0.6 0.4 0.12 0.138 47.2
Tot 20.4 1213
Calculation for units on the Jukani test area (Croatia, on the Botonega basin), Fig. 4:
J1 3.33 1043 12.5 1.16 0.36 2736 8.8 2.36 0.19 0.4 1130
J7 1.71 1043 12.5 1.16 0.50 2300 5.77 1.29 0.10 0.26 606
Tot 26.7 28 615 10 666
 Applicability of the Gavrilović method in erosion calculation 229

Fig. 2 Pedology, the Corine Land Cover (Table 1) and sub area reference numbers for
the Dragonja River basin.

Fig. 3 Digital terrain model and spatial units for the Rokava sub-basin, used for GIS
calculation of the Gavrilović coefficients.

and no flow directions are taken into consideration. The annual erosion rate for the
Rokava sub-basin calculated with this method was 60 m3 km-2, whereas it was 78 m3
km-2 if calculated with the traditional method.
230 Lidija Globevnik et al.

Fig. 4 The Botonega Reservoir basin area with the test area (Jukani), divided into
hydrographic units (J1–J7) and results of the erosion coefficient Kz.

In Table 3, erosion rates and sediment yield for hydrographical units in the Jukani
test area (on the Botonega Reservoir basin) are presented. The spatially distributed
Gavrilović coefficient Kz for the Botonega Reservoir basin as determined with GIS
techniques (Holjević, 2002) is presented in Fig. 4. A comparison to the measured value
of 19 000 m3 of deposited material in the period 1989–2000 was also carried out. The
computed annual sediment yield of 10 666 m3 was divided into bed load (1446 m3) and
suspended load (9220 m3). Based on measurements at high flows and regional
experience, the trapping efficiency for suspended sediment was assumed to be 10%.
This resulted in a total volume of deposited sediment of 26 000 m3 in the specified
period of 11 years, which can be considered a reasonable agreement between the
measured and computed values.

DISCUSSION

Three interesting differences can be observed from the results of this study: the
temporal change in soil erosion rates in the Dragonja basins between 1971 and 1995;
the remarkable difference in soil erosion rates between two otherwise similar Dragonja
and Botonega basins; and finally, the difference in the results when average or
distributed values of the parameters are used in the selected hydrographic unit (the
Rokava test area).
In 1971, the annual soil erosion rate in the Dragonja basin amounted to 325 m3 km-2,
which is close to the average for the Republic of Slovenia (Mikoš, 1995). In 1995, this
figure was reduced by more than half to 128 m3 km-2. There are two main reasons for
 Applicability of the Gavrilović method in erosion calculation 231

this decrease. The first is a conversion of agricultural land to bush and forest land due
to the abandonment of previously intensively farmed fields in the area (Globevnik et
al., 1998; Globevnik & Sovinc, 1998). In addition, in 1971 systematic erosion control
measures were introduced in the area (Globevnik, 1998). Consequently, natural
reforestation commenced and, as a result, erosion rates decreased (Globevnik, 2001).
The average annual erosion rate in part of Botonega basin, the Jukani test area,
calculated by the Gavrilović method in a GIS environment (Table 4) is 1070 m3 km-2
(Holjević, 2002). This rate is several times higher than the rates in the Dragonja basin
or the Rokava sub-basin. The likely reason is the selection of the values of the
coefficients, in particular Ko. Another example of extreme erosion in Croatian Istria is
the Raša basin (205 km2) in central Istria, which is similar to the Botonega basin, and
has an estimated average annual erosion rate of 1270 m3 km-2. A detailed survey of the
river’s estuary between 1938 and 1968 showed that 1 450 000 m3 of sediment was
deposited (Juračić et al., 1995), equivalent to an average annual erosion rate of 230 m-3
km-2. Conversely, monitoring of the suspended and bed load in the Rokava sub-basin
of the Dragonja basin showed the annual sediment yield of around 50 m-3 km-2
(Petkovšek, 2002). These data indicate that differences in annual erosion rates between
basins can be high, even when they are located close to each other and have similar
climatic and geological characteristics.
It is also interesting to observe the difference between the soil erosion rates
calculated with the Gavrilović method in a traditional way, using the values of the
coefficients for hydrographic units (Table 2), and with distributed values of
coefficients for sub units on the basis of land use and pedology (Table 3). The
difference between the two approaches on the Rokava sub basin is approximately 20%.
For the Jukani basin, Holjević (2002) found that the difference between traditional
(58 000 m3 year-1) and GIS methods (28 000 m3 year-1) is even higher. The difference
may be explained by the fact that with the GIS approach we delineate the basins into
much smaller sub units in the agricultural area, but into larger units in the forested
area. The agricultural areas that are prone to more erosion are located on lower slopes,
where stream networks are not fully developed, which accord with the lower Fsl and Ko
coefficients. The forest areas that usually have larger Fsl and Ko, have very low Kx
(high protection). In the traditional way, the average coefficients of the hydrographic
units are used, which do not take into account specific conditions of land use. Within
the hydrographic units, frequently the erosion coefficient Kz is not uniform. In this
case, its mean value should be calculated. Considering the structure of equation (1), the
correct expression for mean Kz is (Petkovšek, 2000):
æ å K 1.5 z ,i ⋅ Fi ö
2/3

Kz = ç ÷ (8)
ç
è å F i
÷
ø
Applying the correct expression for the mean Kz (equation 8) we might get different
results.

CONCLUSION
The approaches for using the Gavrilović method in combination with GIS techniques
as presented here demonstrate a great potential for soil erosion estimation in basins.
232 Lidija Globevnik et al.

The application of such methods in two similar, sub-Mediterranean basins showed

both similarities and differences in the determination of key coefficients between the
two different approaches. The main difference was the estimation of the erosion
development coefficient Ko, where the severity of erosion is to be assessed. The
problem of averaging spatially-variable parameters is also stressed. A possible solution
to the latter is the use of homogeneous units that have the same value for all soil-
erosion-related parameters. These units can be constructed with GIS techniques, as
shown in the given examples.
The Gavrilović method and the presented GIS procedures can be further refined by
developing a more representative factor for the impact of precipitation (e.g. R factor
from RUSLE equation) and its spatial variability. The Gavrilović method can also be
modified in terms of determination of hydrographic units. Gavrilović assumed
(Gavrilović, 1972) that the ratio of the perimeter of a hydrographic unit (sub-basin) to
its slope length is equal to π (3.14). However, using GIS tools, the ratio can be
calculated from the geometric properties of a subunit as derived from pedological and
land use characteristics. Following Gavrilović’s proposals, the subunit for which
coefficients are determined should be a hydrographic unit (sub-basin with developed
hydrographic network). How those two approaches interact should be further analysed.
In this respect, we should also take into account lateral and riverbed erosion. For
further method validation, measurements on both basins have already commenced to
monitor actual sediment yield.

REFERENCES

Browning, G. M. (1947) A Method for Estimating Soil Management and Requirements. Iowa, USA.
Cook, H. L. (1936) The nature and controlling variables of the water erosion process. Soil Sci. Soc. Am. Proc. 1, 60–64.
European Environmental Agency (1999) Corine Land Cover—a key database for European integrated environmental
assessment. G.I.M.-Geographic Information Management NV, Belgium.
Gavrilović, S. (1962) Proračun srednje godišnje količine nanosa prema potencijalu erozije. Jaroslav Černi Institute,
Beograd, Serbia.
Gavrilović, S. (1970) Savremeni načini proračunavanja bujičnih nanosa i izrada karata erozije (Proceedings of Erozija,
bujični tokovi I rečni nanos), 85–100.Jaroslav Černi Institute, Beograd, Serbia.
Gavrilović, S. (1972) Inženjering o bujičnim tokovima i eroziji. Izgradnja. Beograd, Serbia.
Gavrilović, S. (1976) Bujični tokovi i erozija (Torrents and erosion, in Serbian), 159–311. Građevinski kalendar 1976,
Beograd, Serbia.
Globevnik, L, Fazarinc, R. & Sovinc, A. (1998) Land degradation and environmental changes in the Slovenian
Submediterranean (The Dragonja River Catchment). Geoökodynamik XIX, 281–291. Bensheim, Slovenia.
Globevnik, L. & Sovinc, A. (1998) The impact of catchment land use change on river flows: the Dragonja river, Slovenia.
In: Hydrology in a Changing Environment, vol. I. (ed. by H. Wheater & C. Kirby). (Proc. British Hydrol. Soc. Int.
Conf., Exeter, UK), 525–533. John Wiley & Sons, Chichester, UK.
Globevnik, L. (1998) Analysis of river morphological and environmental changes with the integration of historical data
and image processing. In: Modelling Soil Erosion, Sediment Transport and Closely Related Hydrological Processes
(ed. by W. Summer, E. Klaghofer & W. Zhang) (Proc Vienna Symp., July 1998), 279–285. IAHS Publ. 249.
Globevnik, L. (2001) Celosten pristop k urejanju voda v povodjih. PhD Thesis, University of Ljubljana, Slovenia.
Hahn, C. T., Barfield, B. J. & Hayes J. C. (1994) Design Hydrology and Sedimentology for Small Catchments. Academic
Press Inc., San Diego, USA.
Herheulidze, I. (1947) Ovranie i selevie vinosi. Moskva, Russia.
Holjević, D. (2002) Procjena erozije tla i pronosa nanosa primjenom GIS-tehnologije. Masters Thesis, University of
Zagreb, Faculty of Civil Engineering (unpublished)
Horvat, M., Faganeli, J. ,Planinc, R. Logar, M., Mandić, V., Rajar, R., Širca, A., Žagar, D. & Covell, S. (1998) The impact
of mercury minig on the Gulf of Trieste. In: Coastal Environment (Proc. 2nd Int. Conf., Cancun), 11–20. WIT Press.
Southampton. UK.
 Applicability of the Gavrilović method in erosion calculation 233

Juračić, M., Sondi, I., Rubinić, J. & Pravdić, V. (1995) Sedimentacija u neravnoteznom estuariju pod utjecajem rijeke:
krski estuarij Raše (Hrvatska) (Sedimentation in an unbalanced estuary under river influence, in Croatian). (Proc.
First Croatian Geological Congress), 265–268.
Mikoš, M. (1995) Soodvisnost erozijskih pojavov v prostoru (Spatial interrelations of erosion-related phenomena, in
Slovenian). Gozdarski vestnik 9, 342–351.
Petkovšek, G. (2000) Process based soil erosion modelling. Acta Hydrotechnica Ljubljana 18/28, 41–60.
Petkovšek, G. (2002) Kvantifikacija in modeliranje erozije tal z aplikacijo na povodju Dragonje (Quantificaion and
modelling of soil erosion - application to Dragonja catchment, in Slovenian). PhD Thesis, University of Ljubljana,
Slovenia.
Poljakov, B. V. (1953) Problemi ruslovih processov. Moskva, Russia.
Silvestrov, S. I. (1955) Reljef i zemljedelije v erozionih rajonah. Moskva, Russia
Smith, D. D. (1941) Interpretation of soil conservation data for field use. Agric. Eng. 22, 173–175.
Zingg, A. W. (1940) Degree and length of land slope as it affects soil loss and runoff. Agric. Eng. 21, 59–64.