Getting standardized spectral information about eroded soil by integration of gis and remotely sensed data

Jaromír Kolejka
Masaryk University, Faculty of Science
CZ-61137 Brno, Czech Republic

Nikos Silleos, Yiannis Manakos, Alexandros Konstadinidis
Aristotle University, Faculty of Agriculture
GR-54000 Thessaloniki, Greece

1. Introduction

The soil erosion is one of the most extended damaging phenomena acting in the environment. If the erosion exceeds the level that can be compensated with the pedogenetic process in the nature, the way for the soil degradation is open with consequences in the economy. The perspective goal of many research and applied activities is to preserve the production ability of the soil, in the best case its improvement. The basic problem is related with the insufficient inventory of consequences of the soil erosion in a real diverse environment (Šúri, 1996). The mitigation measures can be applied afterwards the data was collected, processed and evaluated.

2. Identification of eroded soil

Intesity of soil erosion varies very strongly in the landscape and it is very difficult to classify and evaluate it what leads to the certain subjectivism in the mapping of the eroded soil. The result insecurity is deepened with problems dealing with (1) the field works and application of other documents about erosional damages (e.g. insurance reports, agricultural maps, research reports), (2) the empiric mathematical formula applications for modelling erosional damages, and (3) not at least the processing of the remotely sensed data. The last mentioned one seems to be relatively the most reliable, especially from the territorial aspects. Various processing methods of remotely sensed data were developed for the identification of eroded soils in the past. Many of them are based on the application of necessary field data and capable processing technology. The digital image processing following the detection of the soil erosional process is common at present and deals with the supporting field data.

The better integration of remotely sensed and other field data is welcome from many theoretical and practical reasons:

  1. the soil erosion in not the process influenced by individual parameters of damaging phenomena and the background only but with the interference of many commonly antagonistic features,
  2. the form and intensity of soil erosion is produced by the combination of erional factors,
  3. no one territory on the Earth is a set of factors, but a harmonic and logical system of factors,
  4. the knowldge about the territorial units with hazardous parameters is the starting point for the qualificied assessment of erosional damages,
  5. the soil erosion is followed with the sedimentation of eroded material with similar optical parameters as eroded soil but in the area without hazardous parameters,
  6. any present landscape in any year season occure abbiotic surfaces with optical parameters similar to eroded soils,
  7. errors in the remotely sensed data processing are caused with the insufficient supporting knowledge,
  8. the digital image processing does not exclude the human impact on the processing progress in any case, it ensures the standard processing quality for the whole scene,
  9. an unified processing supports the application of the methodological and empirical experience in the other territories.

3. Application of standardized experience in detection of eroded soils

At present, some formula (Wischmeier, Stehlík, etc.) are available, to manifest relationships between the intensity of soil erosion (mainly water erosion) and the features of neighbouring environment. Although, all formulas represent detected relationships in a mathematical form, an abundant use of variable coefficients reduces their applicability usually to detail individual studies only. It is generally acceptable that present versions of "formulas" for the "calculation" of erosional damages are either very general or very detail in opposite and mostly do not leave the overview of factors to be ragarded during the assessment of soil erosion. Those very detail formulas need also very detail supporting field and laboratory information. This requirement cannot be fullfil totally because, regardless to some typology and classification procedures, the natural environment varies so much from place to place and also the values of representative parameters what in reality excludes the application of direct numerical data. A similar situation dominates among formulas for the assessment of the erosional risk. This way, it is necessary to base the research on the so called representative figures derived from the classification of original numerical data acquired by experimental measures (e.g. Solín, Lehotský, 1996). This type of data generalization is necessary because of it allows to work with the problem classes and not with the large number of individual cases in the next steps.

The disponible formulas recommend to involve into erosion assessment the information about the local climate (if it varies in the study area), long-term soil humidity, geological structure, terrain, vegetation and especially about the soil, its genetic form and mechanics. Because of the recommended data represents in generally the overview of all compounds of the nature, it means, that the studied soil or its erosional degradation has to be put into the relationships with all the other jointly acting nature features. Some of them can be affected by humans, especially the vegetation cover. The dissection of the soil surface and reduction of its thickness represent the visual demonstration of the soil erosion. They both can be measured in the field by an immediate contact with the soil. The colour deviation is an accompanied demonstration of soil changes separating (convectionally) damaged soil from the other ones. The colour differences are visible in the field well, but the view from the above is the most reasonable for the mapping and classification. Both the "horizontal" dat acquired in the field and the "vertical" data got the birds eye are necessary for the identification of the level and territorial extend of eroded soil. This way only allows the area comparison of different classes of the same phenomenon.

4. Project targets

The international Czech-Greek project was focused on the identification of relationships between the various types of the natural background and the level of erosional damages in soil. The information expected will serve as a leading feature for damage localizing and quantification. Data on the soil and its environment (natural background and human utilizing) has been collected for the identification, localization, classification and evaluation of erosional damages.

The research was carried-out in two selected study areas represent by extracts from the topographic map sheets at the scale of 1:50 000 Hustopeèe (34-21) app. 30 km south-east from the City of Brno (Czech Republic) and Kilkis app. 25 km northward from the City of Thessaloniki (Greece). The Hustopeèe Area extension is 433 km2 and the same of the Kilkis Area is 301 km2.

5. Study areas

The Kilkis study area extends in the centre of the Kilkis Prefecture in Macedonia (Northern Greece) close to the borders to the FY Republic of Macedonia and Bulgaria.

The natural conditions of the Kilkis Area are varied, with regard to the complicated geological structure. NE is rich on paleozoic schichts and gneisses, accompanied with Permian sandstones, quarzites and agglomerates. Younger cover series consist of Triassic-Jurasic limestones. Besides the NE bedrock massiv, these rocks build also islands in other parts of the study area causing higher terrain dissection in the highland (max. elevation Mt.Antigoniz - 569 m) in SE. The bedrock area with a mediterranean climate characterized by annual average temperature 18 oC and 400 mm precipitations per year with extremal summer minimum cover typical cambisols or ranker on silicate rocks and rendzinas with lithosols on carbonates. The most study territory is covered with Quaternary terrace gravels and sands building three levels descending to the E to the River Gallikos valley - the main water stream of the area. The upper terraces are clayed and covered with vertisols. The loamy lower ones host mediterranean mollisols. The youngest fluvial sandy accumulations are covered with arenosols. The surroundings of the bitter Pikrolimni Lake are built with heavy lacustrine deposits with saline gleysols and solonets. Both the rock outcrops and sediments are dissected into a network of gullies. The small field parcelation (max. app. 5 ha) is prevailing in the area, especially where deposits dominate. The forest cover some slopes in dissected E part of the territory and also most of stabilized gullies. The mediterranean bush canopy of frygana type overgrows the abandoned stony parcels. The built-up areas belong to the dense network of small villages.

The study area of Hustopeèe posses a relatively homogenous geological structure. The Paleogene flysch rocks (sandstones and claystones, marlstones) dominate in the bedrock causing the rounded highland terrain in the N. Some prominent ridges are built with Triassic limestones of Pavlovské vrchy Hills in the SW or with hard sandstones in the Ždánický les Mts. in the N, where is the highest point Mt. Pøední kout - 410 m. The flysch bedrock is burried with Neogene marine deposits in the E where the hilly terrain dominates. The foots and partially slopes of hills are covered by loess. Loess, loamy weathering products a Neogene deposits are covered with mollisols and luvisols dependently on the elevation, slope aspect and humidity. Especially loess and flysch claystones were subjected to the intensive gully erosion. The wide riverine plain of the Dyje River in the S and the Trkmanka River from the N to the S are based on the variable fluvial deposits with alluvial soils (gleysols, fluvisols). The lower terraces are covered with gleyic mollisols, the higher ones with typical mollisols. The climate of the study area is moderate continental with the annual average temperature of 9 oC and yearly precipitations of 600 mm with the typical summer maximum. An important role in the soil cover differentiation is played by the humidity conditions. Since 1990, a part of the Dyje River alluvium is artifically flooded with water of the Nové Mlýny Lower Reservoir. The decidous forests cover the tops of highest hills and many of the more inclined northern slopes. The arable land dominates with the typical large parcel farming (max. 200 ha). Vineyards and orchards are also common. The settlement network consists of relatively large communities (towns and villages).

 

6. Territorial data preprocessing

The contactless ways for the identification and evaluation of soil erosion can be focused on the accompanying or indicational demonstrations of the erosional process. These processes are depended on the soil, the surrounding environment and human activities. It is expected, that the erosion intensity and its visual demostrations are closely related to the environment. If the territory will be typified and classified, similarly the classification of demonstrations of the soil erosion can be done. The junction of these two formally separated classications is a technical, not gnozeological problem only, because of the relationships between the soil erosion and its environment are known but not definitely formalized. The first step is based on the location of detected soil erosion into classified geoecological units - geosystems, representing area homogenous from the viewpoint of parameters important for the development of erosional process, too. A later formalizing of the acquired knowledge about mutual relationships between geosystems and soil erosion can lead to the completing of a knowledge base enabling an advanced sensitive detection of the soil erosion, an evaluation of its importance and further development and the selection and addressing of mitigation measures.

7. Data integration

The research has been carried out in the map scale of 1:50 000, because of it combines the advantage of the sufficient resolution, the overview and the acceptable localization correctness.

The modified Wischmeier`s equation contains the main nature factors influencing the soil erosion:

RE = P + T + G + Sm + Sh + S + V,

where RE ...is an assessed level of the soil erosion risk, P...impact of position in terrain, T...slope inclination, G...geological structure, Sm...soil mechanics reflecting the composition of weathering products, Sh...soil humidity, S...genetic soil type, V...potential vegetation indicating the local bioclimate.

Data on the above mentioned variables was collected both for the two study areas. The individual information layers were of different origin and it was necessary to unify the from the viewpoint of the scale, resolution and logical relationships.

The data integration run practically as a consequence of map overlays and "cleanings" of the interproduct for the overlay with the next layer (Fig.1). The information layer overlaying followed the scheme (Fig.2) in the firm consequence: G + Sm + S + Sh, what basically follows the sinking worthiness of available information and the relative level of stability and independence of individual variable. The data on the position in the terrain was added to the area description at least and the direct climatic information is missing, because of it is by the certain way included in the parameters of the genetic soil type influenced by vegetation, what is narrowly linked with the bioclimate.

Fig.1. Data integration scheme

The information about the slope was joined afterwards as a result of the DTM analysis. This way, two mutually comparable networks of natural landscape units (geoecounits, geosystems) were identified both in two research areas as the background for localizing of the erosional phenomenon. The parameters of individual geoisystem types were organized in the tables (see Table 1 and 2) and applied in the further data processing.

Fig.2. Integrated data sets in advanced GIS database

The slope classes (0-2o, 3-7o, 8-15o, over 15o), important for the development of soil erosion, are common in the published papers (Jùva, Hrabal, Tlapák, 1977, Demek, Embleton, Kugler, eds., 1982, Kirchner, 1993, Èvancara, 1962). They represent basically the most common slope classes limiting the beginning and development of land forming processes, soil erosion inclusively, but regardless to the impact of the geological structure, soils, etc.

The compilled maps of geoecological units (geosystems) were stored in a digital form as a coverage in the GIS database of the system ArcView, Version 3.0, installed in Thessaloniki.

The both (Kilkis, Hustopeèe) digital terrain models (DTM) were developed based on the digitized contour lines contained in the appropriate topographic maps and applied later for the derivation data on the slope and aspect, also stored in ArcView 3.0.

Table 1: Natural geosystems in the "Kilkis" Area (extract)

No.

Name

P

T

G

S

Sh

Sm

Number

Code

1

water

B

P

-

-

-

-

1

1200

2

marsh

B

P

C

SG

W

C

1

1190

3

lake plain

B

P

C

SO

M

C

1

1010

34

steep flysch slope

S

S

SM

VE

D

C

37

1114

35

aggl. plateau

P

P

SS

CA

D

L

268

1161

36

flat aggl. slope

S

F

SS

CA

D

L

182

1162

37

gentle aggl. slope

S

G

SS

CA

D

S

284

1163

38

steep aggl. slope

S

S

SS

RK

R

K

159

1164

46

steep metam.slope

S

S

GN

RK

R

K

100

1174

 

Table 2: Natural geosystems in the "Hustopeèe" Area (extract)

No.

Name

P

T

G

S

Sh

Sm

Number

Code

1

riv. plain basin

B

P

C

GL

W

C

9

101

2

active riv. plain

B

P

L

FL

M

L

445

301

25

mollic ped.margin

F

F

L

MO

N

L

119

15032

26

gently mollic ped.

F

G

L

MO

N

L

1

15033

27

loess plateau

P

P

E

MO

N

L

491

15061

28

flat loess slope

S

F

E

MO

N

L

663

15062

29

gentle loess slope

S

G

E

MO

D

L

575

15063

30

steep loess slope

S

S

E

PS

R

L

21

15064

37

gen. flysch slope

S

G

SM

MO

D

L

338

15103

59

gen. limest. slope

S

G

LI

RA

D

K

1

23123

Table explanations

Position (P): B - valley or basin bottom, P - plateau, E - terrain edge, S - slope, F - foot

Terrain (T): P - plain (inclination 0-2o), F - flat slope (3-7o), G - gentle slope (8-15o), S - steep slope (over 15o)

Geology (G): C - clay, L - loam, S - sand, G - gravel, E - loess, B - limestone breccia, SM - flysh sandstones and marlstones, NC - Neogene clays and marls, LI - limestone, SS - sandstones and quazites, GN - gneisses and phyllites,

Soil type (S): PS - primitive soil, RK - ranker, RA - rendzina, PR - pararendzina, CA - cambisol, AR - arenosol, FL - typic fluvisol, FG - gleyic fluvisol, VE - vertisol, LU - luvisol, MO - typic mollisol, MG - gleyic mollisol, MS - salinic mollisol, SO - solonets, GM - mollic gleysol, GL - typic gleysol, SG - salinic gleysol

Soil humidity (Sh): W - wet, M - moist, N - normal, D - drying out, R - dry

Soil mechanics (Sm): C - clayic, L - loamy, S - sandy, G - gravelly, K - stony,

Abbreviations: aggl. - agglomerate, metam. - metamorphite, riv. - riverine, gen.- gently, ped. - pediment

The database for both study areas was completed with the layers for the built-up areas, drainage network and lakes, road and railway network.

8. Satellita data preprocessing

The disponible satellite images for both the study areas were provided from different sources. The SPOT HRV multispectral image (bands 1, 2, 3) from June 6, 1987 covers the "Kilkis" area. The "Hustopeèe" study territory is covered with a Landsat TM image from August 1, 1994 in 7 bands. SPOT bands 1,2,3 and TM bands 2,3,4 are similar from the spectral viewpoint. These bands only were applied for the next data processing. The falso colour composites were created following the 1(2)B+2(3)G+3(4)R band/colour consequence and used for the geometric corrections.

Because of the large part of theterritory is covered with the vegetation in the time of satellite overpassing, the corresponding areas is excluded from the immediate soil research. The vegetation covered areas must be separated from the further processing. The calculation of the normalized differentiated vegetation index using by spectral information from the red and infrared bands is the best way for the identification of vegetation covered surface.

9. Preliminary assessment of geoecological data

Every geosystem is defined and described with a set of parameters important for the development of soil erosion. It is also possible to assess the level of the sensitivness of every one to the soil erosion. An expert team has been set up to reduce the subjectivism in the evaluation. The experts decided to compile a 4-grade scale (0 for the lowest sensitivness of the geosystem parameter to the soil erosion, 3 for the maximum sensitivness). This way, all the geosystem parameters have been assessed from this viewpoint. The assessment results are presented in the table (see Table 3).

Table 3: Erosional risk evaluation of geosystem parameters in the "Kilkis" a "Hustopeèe" study area

Variable risk value

Kilkis Hustopeèe

POSITION

bottom 0 0

plateau 1 1

edge 3 3

slope 2 2

foot 1 1

TERRAIN

plain 0 0

flat slope 1 1

gentle slope 2 2

steep slope 3 3

GEOLOGY

clay 2 2

loam 3 3

sandk 1 1

gravel 0 0

loess . 3

neogene clays/marls . 2

limestone breccia 2 .

sand/marlstones 2 2

limestones 3 1

sandstones, quarzites 1 1

gneisses, phyllites 2 .

SOIL TYPE

primitive soil 3 3

lithosol 2 .

ranker 2 2

arenosol 1 1

cambisol 2 2

rendzina 3 3

pararendzina . 2

fluvisol typic 0 0

fluvisol gleyic . 0

vertisol 2 2

solonets 0 .

mollic gleysol . 0

gleysol typic . 0

gleysol salinic 0 .

luvisol . 1

mollisol typic 2 2

mollisol gleyic . 0

mollisol salinic . 0

SOIL HUMIDITY (long-term)

water 0 0

wet 1 1

moist 0 0

normal 1 1

drying out 2 2

dry 3 3

SOIL MECHANICS

clayic 2 2

loamy 3 3

sandy 1 1

gravelly 0 0

stony 0 0

The risk values of the individual parameter assessment were agglomerated using by the function "addition" and the total sum was presented as a result. The sum values varied between 1-14 in the geosystems of the "Kilkis" area and between 4-17 in the "Hustopeèe" area. These results were classified into 5 classes separately for both two areas (Table 4).

 

Table 4: Erosional risk classes of geosystems in the "Kilkis" and "Hustopeèe" study area

geosystem risk class agglomerated evaluation

Kilkis Hustopeèe

very low 1 - 2 4 - 5

low 3 - 5 6 - 8

medium 6 - 8 9 -11

high 9 -11 12 -14

very high 12 -14 15 -17

Fig.3. Hustopeèe Area (SE extract). Geosystems in erosional risk classes (1 - none or very low erosion risk, 2 - low erosion risk, 3 - moderate erosuion risk, 4 - high erosion risk, 5 - very high erosion risk)

The geosystems within individual erosional risk classes were also demonstrated in computer maps of both study areas (Fig.3). The areas with the highest two erosional risk classes were subjected to the satellite image processing to get spractral values of the most endangered baren soils. This approach makes the research more efficient because of the attention is focused on the selected territory only where the highest expectation of soil erosion is possible.

10. Remotely sensed and geoecological data integration

The preprocessed satellite data about the bare land (NDVI) was corrected using other data sets on built-up areas and communication networks. The fresh forest clearings in the "Hustopeèe" area with spectral parameters similar to bare soils were also excluded from the processing by the application of a mask for forests taken from the topographic map. It was not necessary to separate the water areas from the data set on the "bare land" because of they belong to the "accumulation" areas in the lowest erosional risk classes.

The bare soils in the highest two erosional risk classes were identified by the overlay of (1) the "cleaned" layer of the "bare land" and (2) the areas with geosystems belonging to the two highest erosional risk classes detected by the expert assessment. This way, the "clean" data on bare soils in the erosionally active areas was acquired and addressed to the known natural units.

11. Estimation of classes of eroded soils in selected geosystem types

The most bare soils identified in the high risk areas in the time of satellite overpassing belong regardless to some rare exceptions to slope geosystems. The most common are those on flat and gentle slopes where the agricultural use is possible without large limits and the land is penetrateable for machines.

The "bare soil" occured in 20 types of geosystems in the "Kilkis" area only (June 1987) with a total pixel number app. 1200 in the high risk areas. In three geosystem types only, the total pixel number exceeded 100. The spectral values of the "bare soil" were identified in these 3 geosystem types (flat agglomerate slope with cambisols - code 1162, gentle agglomerate slope with cambisols - code 1163, steep agglomerate slope with rankers - code 1164). The spectral demonstration of the bare soil in these geosystems are shown in the histogramms for 3 analyzed bands. The forms and positions of curves in the axis system are different for these 3 geosystems what makes a good starting point for the application of a more advanced image classification method (e.g. MLC).

The separation of the erosionally damaged soil has been done according to the knowledge collected in the field. The attention was focused on the strongly damaged soil only (notice: the "strongly damaged soil" was not defined clearly in the "Kilkis" area because of the absence of results of the soil sample laboratory analysis) visually identified in the field. The visited sites with strongly eroded soils were found precisively in the image and their spectral values registered. Such class of the eroded soil was separated from the other pixels of the "bare soil" using by the density slice method for any of three geosystem type individually (see Table 5). The areas with the same class of eroded soil (strongly eroded soil) were presented in the map.

 

Table 5: Spectral and geoecological parameters of strongly eroded soils in the "Kilkis" area in SPOT B2 (red band)

geosystem type spectral value pixel number

1162 59 - 67 192

1163 57 - 72 279

1164 56 - 72 128

A different situation dominates in the "Hustopeèe" area because of the high portion of the "bare soil" in the high risk geosystems (August 1, 1994). The total pixel number of the bare soil was app. 30 000 and they were dispersed in 27 geosystem types. The total pixel number more than 1000 was identified in 4 geosystem types only (mollic pediment margin - code 15032, flat loess slope with mollisol - code 15062, gentle loess slope with mollisols - code 15063, gentle flysch slope with mollisols - code 15103). The histogramms were compiled for the two of these 4 geosystems (code 15062 and 15103) as examples (Fig.4) where surface soil samples were taken and analysed before as the strongly erosionally damaged soil (see Kolejka, Shallal, 1997). Fig.4. Example of density slicing in TM B3 for detection of strongly damaged soil

The laboratory, GIS and remotely sensed data make an efficient image classification and a result quantification really possible. The density slide method was applied here in the Landsat B3 band as well to get similar results as they are from the "Kilkis" area.

Table 5: Spectral and geoecological parameters of strongly eroded soils in the "Hustopeèe" area in TM B3 (red band)

geosystem type spectral value pixel number

15062 60 - 120 9155

15103 70 - 95 1347

 

Pixels with corresponding values were found in the image and displayed in the map (Fig.5).

Fig.5. Hustopeèe Area. Strongly eroded soil in geosystems: loess flat slope and flysch gentle slope with mollisols

The comparison both of spectral values in these two geosystem types shows that the same damage class in the soil on loess includes a wider spectral range in TM B3 as it is with data from the same soil on the flysch bedrock. The slope aspect does not play any important role probably because of the low slope incinations in geosystems compared. The impact of the aspect will be tested in the future.

12. Conclusion

The completed phase of the research shows, as expected, that the same classes of eroded soil posse different spectral demonstrations. The histogramms of the bare land can be accepted as introductory standards of soils, incl. eroded soils at least.

The results acquired can be used for the preparation of a formalized knowledge for the further correlation and integration of the laboratory, field, GIS and remotely sensed data for the better detection and quantification of eroded soils and finally for the selection and location of

Acknowlegements: The authors are thankfull both to the Czech and Hellenic Ministries of Education for the financial support in 1996 and 1997 to cover the research expenses.

References

  1. Èvancara, F., 1962: Zemìdìlská výroba v èíslech. Vol.I., SZN, Praha, 1170 p.
  1. Demek, J., Embleton, C., Kugler, H. (edits), 1982: Geomorphologische Kartierung in mittleren Ma3staben. VEB Hermann Hack, Gotha, 254 p.
  1. Jùva, K., Hrabal, A., Tlapák, V., 1977: Ochrana pùdy, vegetace, vod a ovzduší. SZN, Praha, 180 p.
  1. Kirchner, K., 1993: Pøíspìvek k hodnocení reliéfu okresù Èeské republiky. Sbornik ÈGS, Vol.98, No.1, pp.13-24.
  1. Kolejka, J., Shallal, J.K., 1997: Identifikace a dvojstupòová klasifikace erozního poškození na bázi analýzy pùdních vzorkù a družicových snímkù. Geografie-Sborník ÈGS, Vol. 102, No. 1, pp. 17-30.
  1. Solín, Ľ., Lehotský, M., 1996: Susceptibility of the Jablonka catchment to soil erosion. Geografický èasopis, Vol. 48, No. 2, pp. 153-170.
  1. Šúri, M., 1996: Analýza a hodnotenie možností využitia diaľkového prieskumu zeme vo výskume erózie pody. Geografický èasopis, Vol. 48, No. 1, pp. 73-92.