HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Hengl, T. Articles by Husnjak, S. Search for Related Content PubMed Articles by Hengl, T. Articles by Husnjak, S. Agricola Articles by Hengl, T. Articles by Husnjak, S. Related Collections Soil Methods/Instrumentation

Pedology

Evaluating Adequacy and Usability of Soil Maps in Croatia

^a European Commission, Directorate General JRC, Land Management and Natural Hazards Unit, Ispra (VA), Italy
^b Soil Science Dep., Faculty of Agriculture, Univ. of Zagreb, Svetoimunska 25, 10000 Zagreb, Croatia

^* Corresponding author (tomislav.hengl{at}jrc.it)

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS CONCLUSIONS REFERENCES

 
The paper suggests a methodological framework for evaluating adequacy and usability of soil maps by analyzing the following 10 aspects: lineage, consistency, completeness, effective scale, attribute accuracy, thematic contrast, accuracy of legends, integrity, popularity, and accessibility. This framework was used to evaluate the national soil resource inventory in Croatia and to find out why the maps are not used to the full potential. Six 1:50K map sheets (of 185 in total), 3 control surveys, and 10 full profile descriptions in the main landscape regions of Croatia were used to estimate the effective map scale and thematic accuracy of profile observations. In addition, the existing digital data sets (digital soil map and database with 2198 profiles) were evaluated for thematic purity and contrast. The results show that the soil maps are of lower quality than planned and that their usability for spatial planning is limited. The average polygon size and the positional accuracy of primary soil boundaries correspond to the 1:150K scale, while the intensity of field observations corresponds to the 1:250K scale. Mapping units are heterogeneous for mapping of clay content, pH, and organic matter, with an average normalized variation of 68% within units and the mean thematic overlap of 66% between geographically adjacent units. This makes this inventory adequate for small-scale applications only. The major usability problems identified were lack of specific interpretations corresponding to user needs, unpopularity of soil survey concepts, inconsistency of methodology and unclear distribution policy. The evaluation can be used to decide how to improve the usability of the existing datasets and design methodological steps for a new survey by involving end users in the design of the soil information system.

Abbreviations: AD, area of disagreement • ASD, average size delineation • BSMC, Basic Soil Map of Croatia • CYS, Classification of Yugoslav Soils • ESN, effective scale number • GIS, Geographic Information System • ILWIS, Integrated Land and Water Information System • IMR, index of maximum reduction • ISRIC, International Soil and Reference Information Centre • ME, mean absolute error • MLD, minimum legible delineation • RMSE, root mean square error • SIS, Soil Information System • SMU, soil mapping units

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS CONCLUSIONS REFERENCES

 
SEVERAL STUDIES in the past 30 yr have shown that the technical quality and usability of soil maps, especially the ones produced through national soil surveys, has often been overestimated or neglected. Marsman and de Gruijter (1986) showed that the actual purity of soil mapping units in the Netherlands is usually considerably lower than the anticipated level of 70%. Groot (1993) estimated that 80% of soil information in the world is unusable due to incompleteness, unknown reliability, or inconsistant spatial referencing. A major problem of assessing the quality of soil maps is that, typically, the same teams that make the original surveys are responsible for adequacy evaluation. In the USA, quality control and quality assurance are increasingly being arranged by an external and unbiased team (USDA–NRCS, 2003; Part 609), which is not the case of most other countries in the world. A general impression is that the quality issue of soil maps has been underrepresented in the literature and definitively deserves more exposure (Bishop et al., 2001).

Five main elements determine the overall quality of a map: lineage, positional and attribute accuracy, logical consistency, and completeness. These correspond to the quality measures and standards approved by the International Cartography Association and applicable to any GIS (Guptill and Morrison, 1995). Soil surveyors have developed a concept of adequacy of a soil resource inventory, which was first introduced by a group at Cornell University (Forbes et al., 1982). This group proposed that adequacy should be evaluated using four aspects: (1) map scale and texture, (2) map legend, (3) base map quality, and (4) ground truth, also called thematic accuracy. The last aspect has attracted attention of Dutch mappers (de Gruijter and Marsman, 1985; Marsman and de Gruijter, 1986).

Recently, the concept of spatial data usability, of which data quality is considered to be just one element, has been introduced (Wachowicz et al., 2002; Hunter et al., 2003). There are four general aspects of usability: (i) data quality (accuracy, completeness, logical consistency); (ii) data form(at); (iii) data accessibility and price, and (iv) quality of the metadata. The importance of each of these aspects may differ from user to user. For example, for environmental modelers, the incompatibility, low thematic contrast and detail of multi-source environmental geo-data will militate against their full usage. The difference between quality, adequacy, and usability is that quality is a set of constant technical characteristics, adequacy changes within the problem-solving context, while usability reflects all these elements in relation to the end-user satisfaction. Adequacy is related to the concept of effective scale—a less detailed soil map will show higher adequacy if used at smaller scales. Although usability is in practice hard to measure, Hunter et al. (2003) suggested some very concrete aspects of usability, such as data integrity, popularity, satisfaction, speed of access, reliability etc., which can be either measured directly or assessed through interviews. Soil mappers will increasingly need to find a balance between the availability of funds, models, tools, and users' demands. In fact, the key challenge to future soil mapping projects will be to fully identify and meet customers' needs (Indorante et al., 1996) and provide a better linkage with other earth and environmental sciences (Wysocki et al., 2005).

The objective of this research was to refine the methodology of evaluating the adequacy and usability of (traditional) polygon-based soil maps and then test it using a real case study. We decided to combine three methodological frameworks: (1) methodology to asses the adequacy of soil maps as described by Forbes et al. (1982) and Rossiter (2000); (2) methodology to assess the spatial data quality as described by Guptill and Morrison (1995); and (3) methodology to evaluate the usability of spatial databases as described by Hunter et al. (2003), which goes beyond pure technical measures. This framework was applied to a real case study: the national soil resource inventory in Croatia. This was a purely pedologic map, created by subjectively drawing boundaries on aerial photos and topo maps. Finally, we suggested a universal checklist of measurable aspects that can be used to determine the true adequacy and usability of soil maps.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS CONCLUSIONS REFERENCES

 
Case Study: the National Soil Resource Inventory in Croatia
During the late 1950s, soil scientists of the former Federal Republic of Yugoslavia agreed that an extensive general purpose inventory of soilscapes was needed to enhance spatial planning and nature conservation projects (Husnjak et al., 2006). Shortly after, each part of the federation formed their own mapping teams and slowly started to conduct surveys. In Croatia, mapping began in 1964 and lasted for 23 yr, finally ending in 1986. This project, also called "Basic Soil Map of Croatia" (BSMC), followed the soil survey methodology described by Kovaevi and Jaki (1964) and the classification system developed by Kovaevi et al. (1967), and further modified by kori et al. (1985). The project was conducted as a series of more or less independent subprojects, one for each 15' by 15' topographic map sheet covering approximately 20 km by 27.5 km (Fig. 1a ). As a result, each adjacent map sheet did not necessarily have to fit the neighboring sheets.

View larger version (57K): [in this window] [in a new window]   Fig. 1. Soil mapping in Croatia: (a) 1:50K map sheets (left) and an example of a digitized map sheet with soil profiles, and (b) part of the soil map with legend.

Control Surveys
We selected six map sheets (from a total of 185), performed three control surveys (each of an approximate size of 4 by 4 km) and redescribed 10 profiles of BSMC in the main landscape regions of Croatia. This was done to estimate the effective map scale, spatial accuracy of soil boundaries and thematic accuracy of profile observations (Table 1). Control surveys were small compared with the original data sets; they were, however, well distributed over the main geographical regions. We also used the Croatian soil profile database (Antoni et al., 2003) consisting of 2198 observations and the Digital Soil Map of Croatia, generalized to the scale of 1:300K (Bogunovi et al., 1998), to assess the sampling density and thematic contrast of soil mapping units.

Profiles were described using the standard national soil survey methodology and the local classification system—Classification of Yugoslav Soils (CYS), which was used in BSMC and which is still in use in all republics of the former Yugoslavia (Bogunovi et al., 1998; Husnjak et al., 2006). Laboratory analyses were done in the Soil Science Department in Zagreb using the same methods as in the original survey.

Map Scale
The effective map scale was evaluated by estimating the average size delineation (ASD):

[1]

where A_j is the area of the jth polygon and m is the total number of polygons (Forbes et al., 1982). The index of maximum reduction (IMR), that is, the factor by which the scale of the map could be reduced before ASD was equal to the minimum legible delineation (e.g., MLD—should be 0.4 cm² on the map), was then derived by:

[2]

From this, the effective scale number (ESN) was computed as (Forbes et al., 1982):

[3]

where NSN is the nominal scale number.

To describe the general geometry of soil polygons, we used a shape complexity index S, which is the perimeter-to-boundary ratio:

[4]

where P is the polygon perimeter, A is the polygon area, and r is the radius of the circle with the same surface area (Hole, 1978). A value of S close to 1 means that the polygon is rather compact and simple. Higher values describe narrower or more branching polygons, which often means higher positional accuracy and larger effective scale (D'Avello and McLeese, 1998).

Soil Boundaries
In the case of the national soil inventory in Croatia, all soil polygons were delineated manually following the concept of free survey, that is, using an irreproducible method. However, a careful study of the legend and comparison of the definitions of adjacent units make it possible to infer which soil boundaries, showing an obvious topographic break (e.g., abrupt change from a floodplain to a sloping hill) and a change in parent material, should have been followed (Wysocki et al., 2005). These primary boundaries, drawn independently by several surveyors, should definitively match within the soil survey standards (Bie and Beckett, 1973).

Maps of control surveys (photo-interpretation overlays) were georeferenced with an ortho-correction to horizontal precision of 3 to 12 m. The methodology is explained in detail in Rossiter and Hengl (2002). Spatial accuracy of soil boundaries was then assessed by delineating the area of disagreement (AD), which is the area of the polygon produced as the intersection between the original and control surveys. In this case, the ‘true’ line is assumed to be in the middle. Positional accuracy, that is, the mean absolute error (ME) can then be derived as half the average width of the AD:

[5]

where l is the boundary length of the mapped delineations, and l' is the boundary length of the control delineation.

Profile Observations
Ten random profiles from the original soil survey with positions shown on soil maps were selected to assess how well the profile data corresponds to the control. We navigated to these points using the GPS-GIS system with the georeferenced original soil map in the background. It was always possible to be clearly within the square representing the original observation, as the profiles were shown by a 4-mm² square on the soil map (100 x 100 m on the ground). The exact location of the control observation within the 1-ha square was determined in the field by experienced surveyors, who were looking for the same type of site as described in the original survey report. We then compared seven physical and chemical soil properties: sand, silt and clay contents (%), pH (H₂O), pH (MKCl), organic matter (%), and carbonates (%) in all horizons, by calculating the root mean square error (RMSE) and normalized error (RMSE_%):

[6]

[7]

where x is the given value of the soil attribute and x' is the control, n is the number of control measurements, _x is the global standard deviation for the whole study area, and R is the range (Ott and Longnecker, 2001). The dimensionless RMSE_% allows a comparison of accuracy for variables of different types and with different ranges of variation (Park and Vlek, 2002). Values of RMSE_% < 40% indicate a high correlation between the original and control measurements, while values close to 100% mean that we measured any equally probable value, which also means that the profile data are practically useless for spatial prediction. Note that the Eq. [7] is applicable to only the variables with a distribution close to normal. Otherwise a transformation needs to be applied.

Soil Mapping Units
A soil surveyor usually tries to delineate soil bodies in such a way that the contrast between adjacent soil mapping units (SMUs) is maximized, which reflects the idea of the maximum amount of information in a system. In the case of categorical data, separability of attribute values between mapping units is the key measure of the thematic map quality (Lilburne, 2002, p. 404–413). In soil resource inventories, the standard method of assessing classification efficiency is to compare the within-class variances with the between-class and total variances (Webster and Oliver, 1990, p. 63–70). Bishop et al. (2001) discussed some information-content criteria to optimize the choice of grid resolution for spatial interpolation. We decided to assess two aspects of the thematic quality of SMUs: (1) thematic purity or homogeneity of SMU composition and attribute values within the SMUs (Beckett and Burrough, 1971), and (2) thematic separability of geographically adjacent SMUs. These are different issues because SMUs can show imprecise distributions of attributes, while at the same time the thematic contrast between adjacent units can still be fairly high, and vice versa.

These two measures were assessed using 2198 profile observations and the 1:300K digital soil map of Croatia consisting of 65 SMUs. The mean value and standard deviation were calculated for each of the 65 SMUs of the 1:300K digital soil map and for three soil parameters: clay content (%), pH (measured in H₂O), and organic matter (OM) in topsoil (%). The homogeneity within the SMUs was expressed using normalized standard deviation (see also Eq. [7]):

[8]

where s_x,j is the standard deviation of a xth property inside the jth SMU, k is the total number of units, and R_x is the range of variation.

Thematic separability of adjacent SMUs was assessed using thematic overlap (see also Lilburne, 2002, p. 404–413). This was calculated as the average probability of thematic overlap between geographically adjacent SMUs. This measure quantifies the uncertainty in attribute maps for site-specific decisions. If properties within units are normally distributed, the overlap between the two mapping units (µ₁, ₁) and (µ₂, ₂) can be quantified using the t test (Ott and Longnecker, 2001), with null hypothesis that the two samples belong to the same population:

[9]

where P(x₁ x₂) is the normal probability of thematic overlap, P_n is the one-way normal cumulative distribution and is the average value of the xth property. Finally, the average probability of overlap P(x) for all SMUs for a given property is:

[10]

where m is the total number of SMUs. The lower the average probability of overlap, the more SMUs differ from one another, that is, the more contrasting are the delineations, and vice versa. If P(x) > 95%, there is no significant difference in attributes between adjacent SMUs, that is, we can measure similar properties within adjacent SMUs. This means that the SMUs are overspecified and can be simplified.

The neighboring SMUs (geographically adjacent polygon pairs) were derived in ILWIS using the neighbor polygon operation (Unit Geo Software Development, 2001). We then sorted the pairs of SMU polygons using the longest length of the neighborhood boundary between the SMUs and calculated the overlap using Eq. [9] and [10].

Usage and Usability
In a GIS, usability can be defined as the property of a given dataset that expresses (1) how well it helps users to arrive at a correct decision within their problem-solving context, and (2) how easily it can be accessed and made ready for use. This aspect is subjective and difficult to quantify. We decided to assess usability by applying three criteria:

a) How well can the product be integrated into an existing GIS?

b) How popular is the product—what is the number of users compared with the potential number of users, what is the degree of their satisfaction?

c) How easily can potential users access the product?

We first made a small inventory of all existing users and then performed unstructured interviews with several existing users in the land-use planning offices in the cities of Osijek, Karlovac, and Split. In addition, we discussed the cartographic and GIS issues with cartographic departments in major cities in Croatia. Finally, to see both sides of the story, we discussed the usability problems with the original surveyors.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS CONCLUSIONS REFERENCES

 
Lineage
The adequacy and data quality analysis was difficult because the data were not fully integrated and detailed metadata were missing. In some cases, we were able to understand the exact methodological steps used to produce maps only by talking to the surveyors involved. A good example is the problem of switching from the old to the new coordinate system, for which we could not find any information on maps. The soil map of Croatia was produced using old military 1:50K topomaps and had to be transformed to the new system of ground control points. This means that without the help of land surveyors, old soil maps cannot be accurately integrated into a GIS. We first overlaid digitized soil boundaries over the original georeferenced maps and concluded that the discrepancy could be estimated with a simple systematic shift. This example clearly shows how lack of metadata can lead to usability problems.

Logical Consistency and Completeness
As regards consistency and completeness, the national soil inventory shows a number of discrepancies, illustrated by the following two examples. Figure 2a shows how the calculated average density of mapped profiles per 1000 ha varies per map sheet, indicating rather different sampling densities in two parts of the country. The right part of the same figure shows printed and incomplete reports, that is, reports that are currently maintained only as working materials, such as sketches, drawings, handwritten notes, etc. (Fig. 2b).

View larger version (31K): [in this window] [in a new window]   Fig. 2. Two examples of methodological inconsistency (a) and incompleteness (b) of the national soil inventory in Croatia.

Inspection density for the 1:50K maps ranged from 0.7 to 2.2 profiles per 1000 ha, which is an order of magnitude lower than the suggested minimum of 50 (Avery, 1987; USDA–NRCS, 2003) for this scale, and two orders of magnitude lower than the ideal inspection density of four observations per map cm², that is, 160 per 1000 ha. According to the original surveyors, the total inspection density was higher because there were up to 10 times more mini-pits that were visited but not recorded on the original map sheets. Ultimately, we decided not to use this information, since no record of it is available.

Although the soil maps and survey reports from the national inventory seem to be of high quality and with a lot of detail, it is clear that the aimed scale of 1:50K was not achieved in more than 95% of sampled map sheets. We estimated the effective scale to be about: (a) 1:150K based on the ASD; (b) 1:250K according to the inspection density; and (c) 1:150K according to the spatial accuracy of soil boundaries. Similarly, the shape complexity index showed rather simple geometries of delineations. Thus, the national soil inventory in Croatia can be classified as a small scale or medium intensity survey (Avery, 1987), which is not applicable for agricultural extension, civil engineering projects, or county level land-use planning.

Spatial Accuracy of Soil Boundaries
A comparison of soil boundaries in the original and control surveys is given in Fig. 3 . The original maps generally show less detail and follow the master lines only approximately. Only in the case of the "Kalinovac" area was the level of detail similar (Fig. 3c). In this case, we were very confident that the original surveyors should have exactly followed the master lines indicated on Fig. 3. The observed mean absolute error estimated at six evaluated master lines (±40 m) is about three times worse than the typical map accuracy standard for the 1:50K scale (Davies, 1981), which gives an effective map of 1:150K. The other secondary boundaries show even higher disagreement with the control surveys. Very often we could not conclude on which basis the surveyor had drawn the original boundaries. Some neighboring SMUs were described as consisting of the same soil types, often only with minor differences in composition. This makes many of the soil boundaries, other than the master lines, even more relative.

View larger version (35K): [in this window] [in a new window]   Fig. 3. Comparison of the original soil boundaries (left) and control survey areas (right): (a)"Gustirna", (b) "Popovac" and (c)"Kalinovac". The six master lines are bolded. In this case, we were very confident that the original surveyors should have followed the master lines on the right.

Comparison of lab data showed a general correspondence with what we described in the field, for example, the texture class, clay content, and clay increase with an average normalized error of ±62.8% for analytical data (Table 2). The highest accuracy of measuring the same property was for organic matter (±34.8%), while the most inaccurate property was pH.

Thematic Contrast between Soil Mapping Units
Analysis of the thematic contrast between SMUs with regard to clay content (%), pH, and OM % in topsoil showed that the SMUs from the 1:300K soil map are fairly heterogeneous (Table 3). The SMUs showed a high normalized standard deviation (s_%), of 68% on average, for these three variables. Nevertheless, the analysis of variance showed that there were significant (p > 0.01) differences between the 65 SMUs.

View larger version (62K): [in this window] [in a new window]   Fig. 4. Attribute map of (a) clay content (%),(b) histogram, map of relative variation within SMUs (c), and two examples of high and low thematic overlap between neighboring SMU pairs– density histograms produced using the S-Plus package (d). P is the average probability of thematic overlap (P(x)).

Usability Aspects
We concluded from our interviews that most of the government departments involved in physical planning, agricultural extension, and environmental protection are not using the existing soil data to their fullest potential. The problem appears to be two-fold. First, most map units are compound and map users (planners or extensionists) are rarely capable of finding the SMU components in the field solely on the basis of their landscape relations. Second, there are no interpretations about suitability/limitations for a specific land-use system.

The principal small-scale users of soil data in Croatia are: (i) Croatian Waters, a government agency, (ii) county planning offices (8 of 21), and (iii) departments of the Ministry of Nature Protection and Physical Planning. Together these account for only about one quarter of all users that might benefit from soil geoinformation.

Finally, we have identified the three main causes that inhibit the growth of applications:

• Legend and soil survey concepts not understandable to end users. In particular, the soil classification system and the concept of compound mapping units used in Croatia are unfamiliar to most users and therefore unpopular.
  
• Inventory products not adjusted to users. The users in principle ask for interpretations, that is, maps of soil attributes and land characteristics, while soil maps typically show only the distribution of soil types with an accompanying legend.
  
• The accessibility and distribution policy for the soil data in Croatia is neither clear nor transparent. Most often, the choice of using or ignoring information about soils is left to the local decision makers.
  

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS CONCLUSIONS REFERENCES

 
In the sections above, we have presented a methodological framework for evaluating the adequacy and usability of traditional soil maps based on concrete statistical measures. The final checklist of the suggested measures for estimating the true adequacy and usability of soil maps is given in Table 4. This aspect of soil survey should deserve more exposure in the coming years, since the results of our case study confirm that soil maps are often of lower quality than planned. The remaining open issues are discussed below.

View larger version (41K): [in this window] [in a new window]   Fig. 5. Mismatch between the boundaries in adjacent map sheets. Note that most of the adjacent soil mapping units have also showed different topologies. Problems like this can seriously reduce confidence in the soil survey products.

Why Are Soil Maps Not Used?
The quality and usefulness of polygon-type soil maps (area partitions) has for decades been a subject of argument (Webster and Beckett, 1968; Groot, 1993). Many GIS professionals working on data integration claim that the critical layer in a multi-thematic GIS, particularly when utilized for land management decisions, is soil survey information (Maclean et al., 1993). Vegetation, relief, land cover, geological, and other maps, on the other hand, have been considered to be easier to interpret and use by GIS users than maps of soil types. There are several probable reasons why soil maps are so unpopular outside soil science: (1) the concept of soil types cannot be compared with more clearly identifiable individuals such as plant species since soil bodies are hidden, often mixed and fuzzy (Burrough et al., 1997), not to mention the problem of too many active soil classification systems used in the world; (2) reproducible analytical procedures are missing in some phases of soil mapping, for example, during photo-interpretation. Because of this, the soil survey profession is still considered, by some, to be more of an art than a science (Hudson, 1992).

Many usability problems—lack of metadata, inconsistent methodology, incompleteness, unpopular concepts used—discovered in this case study are typical of national inventories conducted in Eastern European countries after the Second World War but also in some more developed countries. This, however, does not mean that new soil maps need to be produced from scratch or that this data is completely unusable. One solution to save this large amount of high quality soil field data is to use auxiliary maps of terrain morphology, remote sensing images, and new quantitative techniques, to produce more accurate prediction maps.

	ACKNOWLEDGMENTS

 
The authors thank the Croatian Ministry of Science and the International Institute for Geo-Information Science and Earth Observation (ITC) for funding and supporting this research. We also thank Dr. D.G. Rossiter, former head of the ITC Soil Science Division, for his comments, suggestions and valuable discussion on the statistical measures of adequacy; Prof. Dr. M. Bogunovi of the Department of Soil Science, University of Zagreb, and Dr. N. Pernar of the Faculty of Forestry in Zagreb, who provided the digital soil database with 2198 full profile observations.

Received for publication April 19, 2004.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS CONCLUSIONS REFERENCES

 

• Antoni, O., N. Pernar, and S.D. Jelaska. 2003. Spatial distribution of main forest soil groups in Croatia as a function of basic pedogenetic factors. Ecol. Modell. 170:363–371.[CrossRef]
• Avery, B.W. 1987. Soil survey methods: A review. Soil Survey & Land Resource Centre, Silsoe.
• Beckett, D.H.T., and P.A. Burrough. 1971. The relation between cost and utility in soil survey IV: Comparison of the utilities of soil maps produced by different survey procedures, and to different scales. J. Soil Sci. 22:466–490.
• Bie, S.W., and P.H.T. Beckett. 1973. Comparison of four independent soil surveys by air-photo interpretation, Paphos area (Cyprus). Photogrammetria 29:189–202.
• Bishop, T.F.A., A.B. McBratney, and B.M. Whelan. 2001. Measuring the quality of digital soil maps using information criteria. Geoderma 103:95–111.
• Bogunovi, M., . Vidaek, S. Husnjak, and M. Sraka. 1998. Inventory of soils in Croatia. Agriculturae Conspectus Scientificus 63:105–112.
• Burrough, P.A., P.F.M. Gaans, and R. van Hootsmans. 1997. Continuous classification in soil survey: Spatial correlation, confusion and boundaries. Geoderma 77:115–135.[CrossRef][ISI]
• D'Avello, T.P., and R.L. McLeese. 1998. Why are those lines placed where they are? An investigation of soil map recompilation methods. Soil Survey Horiz. 39:119–126.
• Davies, R.E. 1981. Surveying: Theory and practice. 6th ed. McGraw-Hill, New York.
• de Gruijter, J.J., and B.A. Marsman. 1985. Transect sampling for reliable information on mapping units. p. 150–163. In D.R. Nielson and J. Bouma (ed.) Soil spatial variability: Proceedings of a workshop of the ISSS and SSSA. Las Vegas, NV. 30 Nov.–1 Dec. 1984. PUDOC, Wageningen.
• ESRI. 2000. Using ArcPad. p. 43. In ArcPad Tutorial, New York.
• Forbes, T.R., D. Rossiter, and A. Van Wambeke. 1982. Guidelines for evaluating the adequacy of soil resource inventories. SMSS Tech. Monogr. 4. Cornell University Department of Agronomy, Ithaca, NY.
• Goodchild, M.F. 2000. Communicating the results of accuracy assessment: Metadata, digital libraries, and assessing fitness for use. p. 3–16. In H.T. Mowrer and R.G. Congaltone (ed.) Quantifying spatial uncertainty in natural resources: Theory and Applications for GIS and remote sensing. Ann Arbor Press, Chelsea, MI.
• Groot, R. 1993. Making information technology work. ITC J. 3:228–235.
• Guptill, S.C., and J.L. Morrison (ed.). 1995. Elements of spatial data quality. p. 1–197. Elsevier Science Ltd.
• Hole, F.D. 1953. Suggested terminology for describing soil as three-dimensional bodies. Soil Sci. Soc. Am. Proc. 17:131–135.
• Hole, F.D. 1978. An approach to landscape analysis with emphasis on soils. Geoderma 21:1–23.
• Hudson, B.D. 1992. The soil survey as paradigm-based science. Soil Sci. Soc. Am. J. 56:836–841.[Abstract/Free Full Text]
• Hunter, G.J., M. Wachowicz, and A.K. Bregt. 2003. Understanding spatial data usability. Data Sci. J. 2:79–89.
• Husnjak, S., D.G. Rossiter, T. Hengl, and B. Milo. 2006. Soil inventory and soil classification in Croatia: Historical review, current activities, and future directions. Country report, ISRIC, Wageningen. [Online] Available: http://www.itc.nl/personal/rossiter/docs/Overview_Soil_Survey_HR_2004_334.pdf. (Verified 17 Feb. 2006).
• Indorante, S.J., R.L. McLeese, R.D. Hammer, B.W. Thompson, and D.L. Alexander. 1996. Positioning soil survey for the 21st century. J. Soil Water Conserv. 51:21–28.
• Kovaevi, P., and V. Jaki. 1964. Handbook for soil survey methodology (in Croatian). Inter-institutional commission for the production of the Soil map of Yugoslavia, Zagreb.
• Kovaevi, P., M. Kalini, and V. Pavli. 1967. A detailed classification of Soils and elaboration of Soil map of the Croatian part of the Sava river valley, in the scale 1:50 000. (In Croatian.) Zemljite i biljka, Beograd. p. 247–259.
• Lilburne, L. 2002. Quantifying map confusion due to uncertainty. p. 404–413. In Proceedings of the 5th International Symposium on Spatial Accuracy Assessment in Natural Resources and Environmental Sciences (Accuracy 2002), Melbourne, Australia.
• Maclean, A.L., T.P. D'Avello, and S.G. Shetron. 1993. The use of variability diagrams to improve the interpretation of digital soil maps in a GIS. Photogramm. Eng. Remote Sens. 59:223–237.
• Marsman, B.A., and J.J. de Gruijter. 1986. Quality of soil maps: A comparison of soil survey methods in a sandy area. Soil Survey Papers 15. Netherlands Soil Survey Institute, Wageningen.
• MathSoft Inc. 1999. S-PLUS 4 guide to statistics. MathSoft Inc., Seattle.
• Ott, R.L., and M. Longnecker (ed.). 2001. An introduction to statistical methods and data analysis. Duxbury Press, Belmont.
• Park, S.J., and P.L.G. Vlek. 2002. Environmental correlation of three-dimensional soil spatial variability: A comparison of three adaptive techniques. Geoderma 109:117–140.[CrossRef][ISI]
• Rossiter, D.G., and T. Hengl. 2002. Technical note: Creating geometrically correct photo-interpretations, photomosaics, and base maps for a project GIS. ITC, Department of Earth System Analysis, Enschede, Netherlands. [Online] Available at Http://www.itc.nl/personal/rossiter/(verified 15 Feb. 2006).
• Rossiter, D.G. 2000. Methodology for soil resource inventories. 3rd ed. ITC, Enschede, the Netherlands. [Online] Available: http://www.itc.nl/personal/rossiter/teach/ssm/SSM_LectureNotes2.pdf (verified 17 Feb. 2006).
• kori, A., G. Filipovski, and M. iri. 1985. Classification of Yugoslav soils (in Serbo-Croatian). Department of Sciences and Arts of Bosnia and Herzegovina, Sarajevo.
• Unit Geo Software Development. 2001. ILWIS 3.0 Academic user's guide. International Institute for Geo-information Science and Earth Observation (ITC), Enschede, the Netherlands.
• USDA–Natural Resources Conservation Service, 2003. National Soil Survey Handbook, title 430-VI. [Online] Available: Http://soils.usda.gov/technical/handbook/ (verified 10 Apr. 2004).
• Van Reeuwijk, P. 1984. Exchangeable bases, base saturation and pH. ISRIC Technical paper 8: Part II. International Soil and Reference Centre, Wageningen, the Netherlands.
• Wachowicz, M., C. Riedemann, W. Vullings, J. Suárez, and J. Cromvoets. 2002. Workshop report on spatial data usability. p. 429–436. In Proceedings of the 5th Agile Conference on Geographic Information Science, Palma Mallorca, Spain. Universitat de les Illes Balears.
• Webster, R., and P.H.T. Beckett. 1968. Quality and usefulness of soil maps. Nature 219:680–682.[CrossRef]
• Webster, R., and M.A. Oliver. 1990. Statistical methods in soil and land resource survey. Oxford University Press, Oxford.
• Wysocki, D., P.J. Schoeneberger, and H.E. LaGarry. 2005. Soil surveys: A window to the subsurface. Geoderma 126:167–180.

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Hengl, T. Articles by Husnjak, S. Search for Related Content PubMed Articles by Hengl, T. Articles by Husnjak, S. Agricola Articles by Hengl, T. Articles by Husnjak, S. Related Collections Soil Methods/Instrumentation