Soil monitoring in Germany

Review Articles Soil Monitoring in Germany Soil Monitoring in Germany Spatial Representativity and Methodical Comparability Winfried Schr6der*, Roland Pesch and Gunther Schmidt Institut fiir Umwehwissenschaften sowie Forschungszentrum fiir Geoinformatik und Fernerkundung der Hochschule Vechta, Postfach 15 53, D-49364 Vechta, Germany *Corresponding author ([email protected]) 1 Background and Goal Soil monitoring is a central component of ecological environmental monitoring that should be able to give a comprehensive description of the ecological situation in Germany (~ 12 German Nature Protection Law). This requires the fulfillment of the following three preconditions: (1) At each monitoring site, data on flora, fauna, soil, water a n d air are gathered by standardized methods (vertical integration). (2) Standardized data from monitoring nets that each concentrate on one ecosystem compartment, i.e. on flora o r fauna o r soil ..., and that are located at different sites, are united by means of geostatistical extrapolation (horizontal integration). (3) Monitoring sites, which correspond to criteria (1) or (2), should cover the most important landscapes (landscape representativity) (Schr6der et al. 2001). This article describes the aspects of the analysis and planning of environmental measuring nets specified under (2) and (3) using the the soil monitoring sites in Germany (Bodendauerbeobachtungsfl~ichen, BDF) as an example. JSS - J Soils & Sediments 4 (1) 49 - 58 (2004) 9 e c o m e d publishers, D-86899 Landsberg, Germany and Ft. Worth/TX 9 Tokyo | Mumbai o Seoul ~ Melbourne | Paris 49 Soil M o n i t o r i n g in G e r m a n y Soil monitoring on BDF should register the current soil condition, monitor its changes and should allow one to make a prognosis of the future development. Thus, BDF serve as an archive and calibration tool for soil protection (Bartels and Meiwes 1996, 6). In order to achieve these goals, the BDF are selected, installed and operated by the states of the Federal Republic in accordance with criteria worked out by soil monitoring experts (SAG / UAG 1991 ). These criteria should also be used for the determination of BDF, which (1) according to S 19 of the Federal Law for the Protection of Soils (Bundesbodenschutzgesetz, BBodSchG) should preferably provide data on the physical, chemical and biological soil condition~ and its changes, as well as on the contents and deposition of contaminants which, (2) according to ~ 12 BnatSchG, should be linked to the measuring data of other environmental programs. Those guidelines of SAG / UAG (1991) which correspond to the comparability of measuring data and the landcsape representativity of BDF, have been put into practice by the states of the Federal Republic ignoring methodical harmonization. Only Brandenburg and Schleswig-Holstein selected soil monitoring sites by means of statistical procedures that meet objectivity and reliability. Soil monitoring data, however, should be used for German-wide report purposes. Measuring data comparability and landscape representativity of BDF must therefore be determined by a standardized procedure. Tools had to be developed that enable the examination of the comparability and quality of soil monitoring data (section 3) as well as the representativity of monitoring sites for Germany's landscapes (sections 4.2 and 4.3). Since the landscape representativity is quantified by a relative measure, the absolute number of measuring sites which suffice for the estimation of areal data must be determined by geostatistical analysis of the spatial validity of measurement data (section 4.4). So, the article at hand describes an operationalized procedure that enables the investigation of the comparability of measuring data, of the landscape representativity of the m o n i t o r i n g sites and of the geostatistical representativity of the measurement data. This procedure has been developed and tested in three research projects (Schr6der et al. 1998, 1999, 2001). 2 Research Design The operationalization of the comparability of measuring data and of the landscape representativity of monitoring sites is comprised of two steps. In the first step, two data sets are linked: Metadata that was collected with a digital questionnaire about geographical locations, measured variables, methods and quality of the data acquisition of the BDF (section 3), on the one hand, as well as data on ecological landscape characteristics and statistical measures that were computed from these characteristics on the other hand (sections 4.2, 4.3), with the latter allow the ranking of all partial areas of Germany according to their landscape representativity. Finally, the geostatistic analysis of the spatial representativity of the measuring data is realized. Therefore, representativity analysis serves to answer the following questions: Does the number of meas- 50 Review Articles uring sites correspond to the surface portions of the landscape types (frequency-statistic landscape representativity)? Are the measuring sites located in typical neighborhood structures (structural landscape representativity)? Does the absolute number of measurement sites suffice to spatially generalize measuring data from these sites (geostatistical representativity) (section 4.4)? 3 3.1 Methodical Comparability of Soil Monitoring Data Background and target The aim of the collection of metadata is to investigate the methodical comparability of soil monitoring data which is a prerequisite for statistical analysis and ecological evaluation. Therefore, with the help of this metadata, it should be possible to answer the following questions: Where do monitoring activities take place? What are they aiming at? What variables are measured with which methods? How is the quality of measurement assured and documented? How are the data put in archives? Is the measurement data suitable for soil and environmental monitoring purposes according to environmental law? The existing metadata systems on environmental monitoring in Germany, i.e. the 'Umweltdatenkatalog' (UDK) and the 'German Environmental Information Network' (GEIN), are not detailed enough for the analysis of the comparability of data and of representativity criteria because methods of data acquisition and of statististical evaluation are not taken into consideration. Furthermore, the degree of detail of the UDK objects depends on the respective editor and there is no specific description of monitoring sites. Thus, it is essential to supplement the information of the UDK and the GEIN by the metadata collection described in section 3.2. 3.2 Metadata collection and evaluation The metadata which should supplement the UDK is collected by way of digital questionnaires that have been answered by the authorities operating the soil monitoring sites. Structure and technical features of the questionnaires all enable the integration of the collected metadata into the UDK and GEIN. The supplementary metadata concretize the comparability of measuring data and the landscape representativity with regard to ten aspects which have been derived from the guidelines for selection and operation of soil monitoring sites (SAG / UAG 1991). These aspects were elaborated in great detail in the digital questionnaire (Table 1). The revised version of the BDF guidelines (Barth et al. 2000) could not be considered. This, however, is not substantial with respect to the development and application of software tools and statistical methods which this article focuses on. For each soil monitoring program of the states of the Federal Republic, a part of the questionnaire asking for general information had to be filled out. This part corresponds to the UDK and contains formal and administrative questions to the entire monitoring program. Data which is specific for each of the soil monitoring nets (measured parameters, JSS - J Soils & Sediments4 (1) 2004 Review Articles Table 1: Structure of the digital questionnaires for the collection of metadata Soil Monitoring in Germany 3.3 Results All answered questionnaires were united in one database. This database may be linked to areal data on soil texture, soil types, land use, climate, topography, potential natural vegetation and a landscape regionalization which was derived from these data layers (section 4.2). The evaluation of all questionnaires shows that most states of the Federal Republic implemented the mandatory criteria of the selection and operation of BDE Between 70 and 98% of the 133 mandatory criteria of the guideline (SAG / UAG 1991) are taken into consideration. The criteria for technical facilities (II, 80%), sampling (VI, 98%), analytics (VII, 77%) and quality assurance (91%), which are substantial for the comparability and statistical evaluation of the data, are kept to a considerable degree. Quality control and assurance is crucial for the Germanwide comparability of soil monitoring data (Beier 2000, Frfinzle et al. 1995, Kluge and Heinrich 1994, Schr6der et al. 1991, Wagner et al. 1997). Most states of the Federal Republic apply the guidelines for sampling, sample preparation and analytics. In 99%, the analytical methods used are comparable. In most cases, quality assurance of the analytical results is performed by interlaboratory tests or DIN / ISO guidelines. measuring methods, quality assurance, etc.) are treated in the following way: (1) yes / no, (2) one of many choices, (3) multiple choice, (4) numbers and (5) texts. For evaluation purposes the questionnaires may be merged in a desired way, e.g. for a measuring variable of interest. The BDF metadata raised with the questionnaire is supplemented by metadata of other programs relevant for environmental monitoring (Schr6der et al. 2001). The questionnaires have been answered by all states of the Federal Republic and contain information on 630 BDE Meanwhile the metadatabase has been extended for approximately 850 BDF (Spatz 2001). This latest version, however, has not been available for the evaluation presented in this article. Only two of the aspects listed in Table 1 can be discussed here. For a complete analysis, refer to Schr6der et al. (2001). JSS - J Soils & Sediments 4 (1) 2004 The complex of questions I (selection of monitoring sites, see Table 1) deals with representativity criteria like landscape, soil and land use representativity. The results of the evaluation of these questions show that most states of the Federal Republic fulfill the representativity criteria. The realization of landscape representativity (93% of the BDF) is based predominantly (80%) on the regionalization after Meynen et al. (1962). Less considered are climate (20%) and relief (10%). Soil representativity has been considered with respect to soil associations (86%), substrate of soil development (73%) and soil type (62%). Special sites have particularly been selected according to the degree of contamination (78%) or the proximity to emitters (78%), although less with regard to the cultivation intensity (11%) and atmospheric deposition (13%). Practicability criteria have frequently been important for the selection of soil monitoring sites (95%), and administrative criteria have rarely been used (15%). Although the requirements for the selection of soil monitoring sites are fulfilled to a great extent, the procedure applied cannot be put in concrete terms. It is only in Brandenburg and Schleswig-Holstein where BDF have been selected in a statistically based manner (Cordsen 1993, Daschkeit et al. 1993~ Kothe and Schmidt 1994, Kuhnt 1989). In the remaining states of the Federal Republic, the soil monitoring sites have been selected predominantly by the evaluation of maps, e.g. maps on soil distribution and typical landscapes. Lower Saxony used the soil information system NIBIS to calculate spatial portions of soil units. In section 4, a standardized procedure for the determination of the representativity of environmental monitoring nets is described in detail. 51 Soil Monitoring in Germany 4 4.1 Review Articles 4.2 Representativity Analysis Concept 4.2.1 Environmental monitoring is supposed to quantify the elements of ecosystems and their relations considering technical sufficiency and economic efficiency. The efficiency criteria are fulfilled if monitoring sites are representative in terms of the inventory and spatial structure of landscapes and if the measured data are geostatistically valid. Geostatistical analysis presupposes that the considered data is comparable regarding methodical aspects and has been submitted to the usual procedures of internal and external quality control (Frfinzle et al. 1995, Mohnen 1996). A comprehensive analysis of the representativity of environmental monitoring nets encompasses three steps: calculation of landscape representativity of monitoring sites by means of frequency statistics (section 4.2) and neighborhood analysis (section 4.3) as well as geostatistical analysis of the measuring data (section 4.4). Slightly different methods are used in hydrology (Kleeberg, 1992), in soil protection (Bachmann et al. 1994, Beckmann et al. 1993, Daschkeit et al. 1993, Hinterding and Streit 2000, Kothe and Schmidt 1994, Kuhnt et al. 1992, Utermann et al. 1999, Utermann and DOwel 2000), in forest monitoring (Wolff et al. 2000) and in environmental monitoring (Fr/inzle et al. 1987, Schr6der et al. 1992, 2001, Vetter et al. 1991). Fig. 1 illustrates that the examination of the representativity of monitoring sites is based on a landscape regionalization which cannot be described in detail. To determine geostatistical representativity, measuring data and geographical locations of the measuring sites are needed (section 4.4). Frequency of statistical representativity of soil monitoring sites Fundamentals The interrelations between biotope and biocoenosis lead to spatial structures that can be recognized in macroscopic dimensions up to the landscape scale and be interpreted as integral indicators of material and energy flows in ecosystems. These spatial structures should be covered by environmental monitoring and their measuring data should be valid not only for the measuring sites but also for the surrounding areas. The determination of the landscape representativity of environmental monitoring nets presupposes differentiating between areas that are homogeneous regarding characteristics that allow a conclusion on the material household of the respective area in a certain scale. If landscapes are indicators of material budget types, then the anthropogenetic influence on the material household should be seized by environmental monitoring. Even if the monitoring sites are representative of the inventory and spatial pattern of landscapes, monitoring data should not be spatially extrapolated without geostatistic examination, except in the sense of a geoscientific analogy reasoning (Martin et al. 2000). Because landscapes are receptors of impacts, the knowledge of climate, vegetation and soils of landscape units, in which environmental monitoring is operated, is essential for the interpretation of the measuring data. 4.2.2 Method On the condition that geographical coordinates and empirical measurement data are available, geostatistic procedures like variogram analysis and kriging may be used to examine Representativity analysis of environmental measurement nets ] areal data Areal and measurement data Areal estimations Classification (variogram analysis, kriging) with CART t Frequencylandscape representativityanalysis ] neighbourhoodanalyticalevaluation (Schr6deret al. 2001) (Wolffet al. 2000) (Utermannet al. 1999) landscape description 9 ecological potential anthropogenic stress Fig. 1 9Framework of representativity analysis 52 JSS - J Soils & Sediments 4 (1) 2004 Review Articles whether number and geographical arrangement of monitoring sites grasps the spatial structure of the measured variables reliably (Baler 2000, Heinrich 1994 a, 1994 b, Piotrowski et al. 1994). If the measuring data is not available, it is still possible to determine whether the number of monitoring sites in each landscape unit (ecoregion) corresponds to the spatial extent of this landscape unit. This requires a definition of the term landscape that contains the method for its empirical-statistic determination. Such an operational definition is the precondition for the reproducibility of the ecoregionalization. Because the ecoregionalizations of Meynen et al. (1962) and Renners (1992) do not fulfill this criterion, and because they do not comprise a statistical description of the ecoregions, a new ecoregionalization of Germany by means of multivariate statistics and GIS-techniques was worked out. This was done in cooperation with: Bundesamt Rir Naturschutz (BfN), Bundesanstalt fiir Geowissenschaften und Rohstoffe (BGR), Statistisches Bundesamt (StaBa) and Umwehbundesamt (UBA). This ecoregionalization considers the following landscape-ecological characteristics: potential natural vegetation (67 classes; Bohn et al. 2000), soil types (25 classes; Hartwich et al. 1995), elevation (UNEP GRID) as well as the monthly values of air temperature, precipitation, evaporation (arithmetic means 1961-1990) and global radiation (1981-1999) (DWD). For statistic computations, these data layers are integrated in ArcView GIS and gridded on a 2 km x 2 km Raster. The resulting 88,400 grid cells cover the whole area of Germany and are grouped by means of Classification and Regression Trees (CART) according to the similarity of ecological characteristics mentioned above (Breiman et al. 1984). CART is able to compute very large matrices of categorical, ordinal and metric data (Stevens 1946). The results of the CART computations suggest an ecoregionalization with 73 classes as optimal, which were reduced to 21 classes in order to calculate the landscape representativity of soil monitoring sites. The computation of the landscape representativity of soil monitoring sites supplies quantitative answers to the following questions: a) Does the number of monitoring sites correspond to the area that is covered by the respective ecoregion? b) Are the monitoring sites embedded into typical landscape patterns? By examining these two aspects, it is possible to analyze whether the measuring net includes the landscape structures sufficiently. For the soil monitoring sites this is at first investigated regarding question (a). 4.2.3 Results By intersecting soil monitoring sites and the ecoregionalization mentioned in section 4.2.2 in ArcView GIS, it is possible to.quantify the number of BDF within each German ecoregion. We assume that the number of BDF should be directly proportional to the areal portion of an ecoregion in the entire area of Germany. Deviations indicate a lack or a surplus of representarivity of monitoring sites. This is shown in Fig. 2. We can see that the geographical distribution of BDF fits quite well according to the areal portions of the ecoregions. The maximum deviation is about § / - 6%. In case of the monitoring of metal accumulation in mosses (1026 sites), it is even less (about + / - 4%). For the moni- JSS - J Soils & Sediments 4 (1) 2004 Soil Monitoring in Germany Fig. 2: Landscape representativity of soil monitoringsites (BDF) toring of air quality, deviations o f - 5% to + i 8 % are calculated (Schr6der et al. 2001). These findings reveal a different need for optimizing the geographical distribution of monitoring sites. Hence, it follows that further tools are needed: geostatistics for the evaluation of the spatial validity of measuring data (section 4.4) and neighborhood analysis to determine the best possible location for monitoring sites (section 4.3). The calculation of landscape representativity is based on the spatial distribution of a discrete characteristic (here monitoring sites) within the ecoregions. The results of this frequencystatistic landscape representativity analysis do not include information about the spatial landscape pattern of the surrounding areas of the monitoring sites. This may be complementarily done via neighborhood analysis (section 4.3). 4.3 4.3.1 Neighborhood analysis of landscape patterns Method The investigation of maps that describe spatially differentiated phenomena by means of neighborhood analysis results in a quantitative description of the spatial structure of the sourroundings of each location regarding certain characteristics (Vetter 1989, Vetter and Maass 1994). In this way, areas can be identified whose environment is typical with regard to the design and spatial arrangement of chosen charact- 53 Soil Monitoring in Germany Review Articles The RI / MNR is a dimensionless number that expresses the similarity of the spatial structure of a certain grid cell being described by n landscape ecological characteristics to the environment of all other grid cells being described by the same landscape ecological characteristics. RI / MNR ranges from 1 (complete similarity) to 0 (no similarity). Therefore, the calculation of the RI / MNR enables the ranking of all or a sample of the 88,400 grid cells. On account of the calculated indices existing monitoring nets can be thinned out or complemented and new monitoring nets may be designed. In section 4.3.2 we describe the application and results of neighborhood analysis according to the procedure shown in Fig. 3. 4.3.2 Results In the following, the BDF distributed all over Germany are submitted to a four-step neighborhood-analytical evaluation (see Fig. 3). Tables 2 to 5 document part of the results of these working steps (for tables containing the calculation results for 100 BDF see Schr6der al. 2001). Fig. 3: Calculation of representativity of monitoring sites for landscape patterns eristics. This landscape pattern representativity may be expressed by an index. If this index is calculated for a single landscape characteristic it is called Representativity Index (RI). For more than one characteristic, it is called Multidimensional Neighborhood Representativity Index (MNR). Applying this method to the mentioned GIS-layers in section 4.2.2, the RI and the MNR quantify the representativity of each of the approximately 88,400 2 km x 2 km grid cells for each of n landscape chraracteristics depicted in n maps. To calculate the neighborhood-statistical representativity of monitoring sites, the ecoregionalization was complemented by GIS-maps of the spatial distribution of soil types association and land use. After the merging of these three layers in ArcView GIS, the relative areal proportions of each combination of the characters of ecoregion, soil type and land use were calculated. Table 2 lists some of the corresponding resuits. Soil types association 42, for example, is most frequent in Germany (7.56%, column B) with a main focus in ecoregions 62 (30.57%, column D) and 47 (27.52%, column D) where arable land and broadleaved forests dominate (columns E, F). The combination of soil types association, ecoregion and land use described above covers 1.7% of the area of Germany (column G). In the second step, the maps on soils assosciations, ecoregions and land use are overlayed with the soil monitoring locations. These are assigned to the site characteristics and combinations listed in Table 3. The number of BDF are then compared to the areal portion of the respective combination of site characteristics (section 4.2). Table 2: Areal proportions of combinations of soil type associations, ecoregions and landuse 54 JSS - J Soils & Sediments 4 (1) 2004 Review Articles Soil Monitoring in Germany Table 3: Frequency-ranking of BDF with respect to combinations of soil types, ecoregions and land use If the number of BDF is not proportional to the area covered by the respective combination of site characteristics, these areas are either complemented or thinned out by means of the results of neighborhood analysis (Table 4). Soil monitoring sites should be added where the MNR are highest and removed where MNR are lowest. From Table 4, we see that BDF 7 shows the highest representativity for landscape pattern (MNR = 0.83). The respective combination of site characteristics should include 6 soil monitoring sites and indeed it does. These regions are therefore well represented by soil monitoring sites and there is no need to add or remove any BDE A statistical procedure for the plannning of new and evaluation of existing monitoring nets was presented. With this Table 4: MNR-ranking of BDF with respect to combinations of soil types, ecoregions and land use procedure, information can be gained to either complement or thin out existing monitoring nets. Because of financial reasons, there is a great pressure to reduce monitoring systems in Germany. Decisions to reduce the spatial density of monitoring sites should never be met on the basis of the landscape representativity of monitoring nets (sections 4.2 and 4.3) alone, but should additionally be supported by geostatistical analysis of measuring data (section 4.4). 4,4 4.4.1 Geostatistical representativity of monitoring data Background and target Our concept of representativity analysis (section 2.1) focuses on the proportionality of the number of monitoring sites and area of ecoregions (section 4.2), on the representativity of the monitoring sites for the landscape pattern (section 4.3), and finally on the geostatistical validity of monitoring data in the surroundings of the monitoring sites. The analysis of spatial data validity summarized in section 4.4.2 is essential for the sufficiency and effiency of monitoring nets, because it tackles the following two aspects: Can the data measured be transferred to other sites, at which no monitoring takes place? Can the monitoring data be interpolated into the surface, and can the reliability of these estimations be quantified? Both these questions are examined using the example of BDF data in a Land of the Federal Republic. Here, the number of the BDF was reduced to 21% without substantially changing the relative frequency of the BDF within the ecoregions. Even if the landscape representativity of BDF did not change, the influence of thinning out the measuring net's density on the spatial validity of measuring data should be proved (section 4.4.2). 4.4.2 Method Based on the theory of regionalized variables (Matheron 1971) geostatistics helps to examine the spatial validity of measurement data. The basic assumption of geostatistics is JSS - J Soils & Sediments 4 (1) 2004 55 Soil Monitoring in Germany Review Articles that measurements of neighboring sites are more similar, i.e. exhibit less variance, than those from more remote locations (= autocorrelation). This is examined by means of variogram analysis and requires a two-step procedure: estimation of the experimental variogram and fitting of a mathematical model. With the experimental variogram, spatial processes that underly the measured values are estimated. Two input parameters have to be determined: distance intervals and maximum spatial expansion of the variogram. Within defined distance intervals, the semivariances are calculated for pairs of measurement points. It is recommended that the number of pairs within one distance interval should not be less than 30 (Journel & Huijbregts 1978). Two examples of experimental semivariograms are part of Figs. 4 and 5o The quality of any geostatistical estimation depends particularly on the resolution and on the accuracy of the experimental variogram near its origin (Olea 1999). According to Webster and Oliver (2001), a good rule of thumb is to set the distance intervals equal to the average distance between neighboring measuring points. Furthermore, the maximum distance between measuring points, for which semivariograms are still computed, should not exceed half of the maximum horizontal expansion of the area under investigation (Journel & Huijbregts 1978). If the semivariances rise with increasing measuring point distance, a distance-dependent structure of the measured phenomenon can be assumed. The range of spatial autocorrelation of data can be determined by projecting the vertex of the model variogram that is fitted to the experimental variogram onto the distance axis. Within this range, the interpolation between measuring points can be seen as statistically meaningful. If the model variogam cuts the ordinate above the origin, either small scale variabilities or measurement errors can to be assumed (= nugget effect). If measuring data is spatially autocorrelated, then kriging procedures are recommended for interpolation. The spatial extrapolation of soil monitoring data was performed with ordinary kriging. Estimations are calculated and distanceweighted for the cells of a defined grid. The model function derived by the variogram analysis enters the kriging interpolation in form of the extrapolation radius (kriging window) and weighting factors. For the regionalization of data on metal contents of soils, Utermann et al. (1999) stratified the data according to soil texture. To estimate the means of metal contents within each stratum, only those metal values were used there that lie beyond the autocorrelation range. If the aim is to compute, illustrate and interprete the spatial variance and not to calculate a surface mean, the use of autocorrelated data for interpolation should be preferred (Fotheringham et al. 2002, Matheron 1971, O'Sullivan and Unwin 2002). The kriging interpolation optimizes the estimate results by minimizing the estimation variance. Regions can be identified where the spatial data density is too low for a valid regionalization of the measured phenomenon. 56 Fig. 4: Variogram and histogram Pb-contents in the topsoil (complete soil monitoring net) 4.4.3 Results In an exemplary fashion, the results of the geostatistical analysis of Pb-contents in top soils are presented. The variogram that was computed on basis of the complete BDF monitoring net is included in Fig. 4. Fig. 5 shows the results for the BDF-net that was reduced to 21% of the original number of measurement sites. Figs. 4 and 5 demonstrate that both the complete and reduced number of BDF show high semivariances in the first distance intervals (nugget effect). Nevertheless, the variogram of the measured values of all BDF (see Fig. 4) shows an autocorrelation structure of Pb-contents in top soils. This can hardly be recognized in the variogram that was calculated on behalf of the reduced BDF monitoring net (see Fig. 5). On the basis of both variograms, ordinary kriging was performed to derive surface estimations from each sample. Figs. 4 and 5 include histograms of the calculated grid cells of both surface estimations which are not shown here. In the histogram of Fig. 5, almost all high values of Pb-contents are missing. The complete monitoring net registers five main areas in which the Pb content in top soils are higher than 60 to 70 ppm, while the reduced monitoring net registers only one. JSS - J Soils & Sediments 4 (1) 2004 Review Articles Soil Monitoring in Germany References Fig. 5: Variogram and histogram Pb-contents in the topsoil (reduced soil monitoring net) 5 Conclusion and Perspective The analysis of metadata reveals that soil monitoring is a kind of environmental monitoring of great importance. Next to its ecosystematic concept, a considerable degree of methodical harmonization can be shown for German soil monitoring sites. Investigations show that not only the proportional distribution of monitoring sites in landscape-ecological units (landscape representativity) is important for the assessment of environmental monitoring nets (section 4.3). The number of monitoring sites should rather be sufficient to guarantee a spatial representation of the respective measurement variable. Their geographical distribution should be based on the spatial model of landscape-ecological units. Additionally, particular criteria that is important for the object of investigation, like for example the distance to emitters, should be considered. It is strongly recommended that activities for the integration of data in a central German environmental information system are intensified so criteria like methodological aspects and representafivity criteria can be examined for other environmental monitoring networks. Internet and GIS technologies should be used, furthermore, to assist in the environmental data acquisition in Germany. A prototype of such an instrument, the so called Internet and GIS based Environmental Monitoring System (IGUS) was already established and tested in the moss monitoring programme 2000 (Schroeder et al. 2002 c). JSS - J Soils & Sediments 4 (1) 2004 Anger M (2002): Nitrat-Austr~ige auf intensiv und extensiv beweidetem Grimland, erfasst mittels Saugkerzen und Nmi,-Beprobung. II: Variabilit~it der NO 3- und NH4-Werte und Aussagegenauigkeit der Messmethoden. Journal of Plant Nutritition and Soil Science 165 (5) 648-657 Bachmann S, Stolz W, Kantor W und Kuhnt G (1994): Begleitstudie zur bundesweiten Bodenzustandserhebung im Walde. UBA-Texte 6 / 94, Berlin Bartels R, Meiwes KJ (1996): Bodendauerbeobachtungsfl~ichen (BDF) in Niedersachsen. Bodentechnologisches Institut des Nieders~ichsichen Landesamtes ffir Bodenforschung Bremen Barth N, Brandmer W, Cordsen E, Dann T, Emmerich K-H, Feldhaus D, Kleefisch B, Utermann J (2000): Boden-Dauerbeobachtung. Einrichtung und Betrieb yon Boden-Dauerbeobachtungsfl~ichen, in Rosenkranz D, Bachmann G, K6nig W, Einsele G: Bodenschutz. Erg/inzbare Loseblattsammlung, Blatt 9152; BOS 32. Lfg. XI / 00 Beckmann A, Kantor W, Adler G, Khosrawi K, Kothe P, Kuhnt G (1993): Erfassung des Bodenzustandes in Deutschland auf~erhalb der Forstfl~ichen (BZE-F) einschliefglich Statusbericht Boden neue L~inder. Phase I: Ergebnisse der Recherchen, hrsg. v. Bundesanstalt fiir Geowissenschaften und Rohstoffe. Umweltforschungsplan des Bundesministers fiir Umwelt, Naturschutz und Reaktorsicherheit. Forschungsbericht 107 07 001 / 02 im Auftrag des Umweltbundesamtes), Hannover Beier R (2000): Die Validit~it von Umweltdaten. In: Fr~inzle O, Miiller F, Schr6der W: Handbuch der Umweltwissenschaften. Grundlagen und Anwendungen der Okosystemforschung. Landsberg am Lech, Kap. V-1.2 (6. Erg.Lfg.) Bohn U, Gollub G, Hettwer C, Weber H (eds) (2003): Map of the natural vegetation of Europe. Scale 1:2.5 million. Miinster Breiman L, Friedman JH, Ohlson RA, Stone CJ (1984): Classification and Regression Trees. Pacific Grove, Kalifornien Cordsen E (1993): Boden-Dauerbeobachtung in Schleswig-Holstein. Mitteilungen der Deutschen Bodenkundlichen Gesellschaft 72, 859-862 Daschkeit A, Kothe P, Schr6der W (1993): Repr~isentanzanalyse zur Auswahl von Bodendauerbeobachtungsflfi.chen in Brandenburg. Abschlussbericht im Auftrag des Zentrums f/.irAgrarlandschafts- und Landnutzungsforschung e.V., Institut fiir Bodenforschung, Eberswalde-Finow Fotheringham AS, Brunsdon C, Charlton M (2002): Geographically weighted regression. The analysis of spatially varying relationships. John Wiley & Sons, London 57 Soil Monitoring in Germany Fr/inzle O, Kuhnt D, Kuhnt G, Z6litz R (1987): Auswahl der Hauptforschungsr~.ume ffir das Okosystemforschungsprogramm der Bundesrepublik Deutschland. Umwettforschungsplan des Bundesministers fiir Umwelt, Naturschutz und Reaktorsicherheit. Forschungsbericht 101 04 043 / 02 I im Auftrag des Umweltbundesamtes, Berlin Fr~inzle O, Schmotz W, Krinitz J, Hertling Th, Miiller J, Florinski M, Permitin V, Pochlebkina L (1995): Handlungsanleitungffir Schadstoffuntersuchungen in B6den. UBA-Texte60 / 95, Berlin Hartwich R, Behrens J, Eckelmann W, Haase G, Richter A, Roeschmann G, Schmidt R (1995): Bodeniibersichtskarte der Bundesrepublik Deutschland 1 : 1 000 000. Karte mit Erl~iuterungen, Textlegende und Leitprofilen. Bundesanstalt ffir Geowissenschaften und Rohstoffe, Hannover Heinrich U (1994 a): Fl~ichensch/itzungmit geostatistischen Verfahren - Variogrammanalyse und Kriging. In: Schr6der W, Vetter L, Fr~inzle O: Neuere statistische Verfahren und Modellbildung in der Geo/Skologie.Braunschweig, Wiesbaden, 145-164 Heinrich U (1994 b): Fl~icbenhafte Ableitung der Klimaparameter Niederschlag und Temperatur mittels geostatistischer Verfahren. In: Schr6der W, Vetter L, Fr~inzle O: Neuere statistische Verfahren und Modellbildung in der Geo6kologie. Braunschweig, Wiesbaden, 283-295 Hinterding A, Streit U (2000): Grunds~itzliche Ans~itzezur Auswertung punktbezogener Daten und der Ermittlung des Fl/ichenbezugs, in Umweltbundesamt: Fl~ichenhafteDarstellung punktbezogener Daten fiber Stoffgehalte in B6den. UBA-Texte 49 / 00, Berlin, 40-49 Kleeberg H-B (Hrsg.) (1993): Regionalisierung in der Hydrologie. Weinheim Kuhnt G, Garniel A, Kothe P, Schr6der W (1992): Umsetzung des Bodeninformationssystems: Begleitstudie zur bundesweiten Bodenzustandserhebung im Walde (BZE-W). Band 3: Standortbestimmung ffir die begleitende Bodenprobenahme und -analyse sowie lDberprfifung der Messnetzvalidit/it. Bundesanstalt ffir Geowissenschaften und Rohstoffe, Hannover Martin W, Auflendorf M, Suttner T (2000): Obersichtskarte von Hintergrundwerten anorganischer Schadstoffe in B6den Bayerns. In: Umweltbundesamt: Fl~ichenhafte Darstellung punktbezogener Daten fiber Stoffgehalte in B6den. UBA-Texte49 / 00, Berlin, 75-79 Matheron G (1971): The theory of regionalized variables and its application. Fontainebleau Meynen E, SchmithfisenJ, Gellert J, Neef E, Mfillerminy H, Schultze JH (1962): Handbuch der naturr/iumlichen Gliederung Deutschlands, Bad Godesberg Mohnen V (1996): Quality assurance / assurance control as prerequisite for harmonization of data collection and interpretation within the WMO-Global Atmosphere Watch (GAW)Programme, in Schr6der W, Fr~inzleO, Keune H, and Mandry P: Global monitoring of terrestrial ecosystems. Ernst & Sohn, Berlin, 77-84 O'Sullivan D, Unwin D (2002): Geographic information analysis. John Wiley & Sons, London Piotrowski JA, Bartels F, Salski A, Schmidt G (1994): Fuzzy kriging of imprecise hydrogeological data. In: International Association of Mathematical Geology. Proceedings of the Annual Meeting Mont Tremblant, Quebec, Canada, 282-288 Renners M (1992): Geo6kologische Raumgliederung der Bundesrepublik Deutschland. Forschungen zur deutschen Landeskunde 235, Trier SAG / UAG (1991): Konzeption zur Einrichtung von Boden-Dauerbeobachtungsfl~ichen. Arbeirshefte Bodenschutz 1, Ivlfinchen Schr6der W, Fr~inzleO, Daschkeit A, Barrels F, Kaske A, Kerrines A, Schmidt G, Stech C (1998): Organisation und Methodik des Bodenmonitoring. Texte des Umweltbundesamtes 21 / 98, Berlin Schr6der W, Garbe-Sch6nberg CD und Fr~nzle O (1991): Die Validit/it von Umweltdaten. Kriterien ffir ihre Zuverl~issigkeit:Repr/i- 58 Review Articles sentativit~it, Qualit~itssicherung und -kontrolle. Umweltwissenschaften und Schadstoff-Forschung. UWSF- Z Umweltchem Okotox 4 (7) 237-241 Schr6der W, Schmidt G (2001): Defining ecoregions as framework for the assessment of ecological monitoring networks in Germany by means of GIS and classification and regression trees (CART). Gate to EHS 2001, 1-9 (DOI: http://dx.doi.org/ l O.10651ehs2001.03.006) Schr6der W, Schmidt G, Pesch R, Matejka H und Eckstein Th (2001): Konkretisierung des Umweltbeobachtungsprogrammes im Rahmen eines Stufenkonzeptes der Umweltbeobachtung des Bundes und der L~inder.Teiivorhaben 3. Umweltforschungsplan des Bundesministers ffir Umwelt, Naturschutz und Reaktorsicherheit. Abschlussbericht FuE-Vorhaben 299 82 212 / 02, im Auftrag des Umweltbundesamtes, Berlin Schr6der W, Pesch R, Anhelm P, Bau H, Matter Y, Mitze R, Mohr K, Pelter A, Peronne T, Roostai H, Roostai Z, Schmidt G, Siewers U (2002): Untersuchungenvon Schadstoffeintr~igenanhand von Bioindikatoren: Aus- und Bewertung der Ergebnisse aus dem Moos-Monitoring 1990/91, 1995196 und 2000/01Tell IlL Berlin (Umweltforschungsplan des Bundesministers ftir Umwelt, Naturschutz und Reaktorsicherheit - Umweltprobenbank einschliefllich Human- und Biomonitoring, UFOPLANFkz 200 64 218) Schr6der W, Vetter L, Fr/inzle O (1992): Einflufgstatistischer Verfahren auf die Bestimmung repr~isentativer Standorte ffir Umweltuntersuchungen. Petermanns Geographische Mitteilungen 136 ( 5 / 6), 309-318 Stevens SS (1946): On the theory of measurement. Science 103, 677-680 Spatz P (2001): M/Sglichkeiten der l~inderfibergreifenden Auswertung an Standorten der Bodendauerbeobachtung, ausgehend yon der Zusammenstellung der Metadaten aus den L/indern. UBATexte 22 / 01, Berlin Utermann J, Dfiwel O, Fuchs M, G~iblerH-E, Gehrt E, Hindet R, Schneider J (1999): Methodische Anforderungen an die Fl~ichenrepr/isentanz von Hintergrundwerten in Oberb6den. UBA-Texte 95 / 99, Berlin Utermann J, Dfiwel O (2000): R/iumliche und inhaltliche Harmonisierung yon Datens//tzen, in Umweltbundesamt: Fl~ichenhafte Darstellung punktbezogener Daten fiber Stoffgehalte in B6den. UBA-Texte 49 / 00, Berlin, S 65-74 Vetter L (1989): Evaluierung und Entwicklung staristischer Verfahren zur Auswahl von repr~isentativenUntersuchungsobjekten ffir 6kotoxikologische Problemstellungen. Dissertation, Geographisches Institut Kiel Vetter L, Maass R (1994): Nachbarschaftsana|ytische Verfahren, in Schr6der W, Vetter L, Fr/inzle O: Neuere statistische Verfahren und Modellbildung in der Geo6kologie. Braunschweig, Wiesbaden, 103-107 Vetter L, Maass R, Schr6der W (1991): Die Bedeutung der Repr~.sentanz f~r die Auswahl von Untersuchungsstandortenam Beispiel der Waldschadensforschung. Petermanns Geograhische Mitteilungen 135 (3) 165-175 Wagner G, Klein B, Klein R, Miiller P, Paulus M (1997): Umweltprobenbank, in Fr/inzle O, Mfiller F, Schr6der W: Handbuch tier Umweltwissenschaften. Grundlagen und Anwendungen der Okosystemforschung. Landsberg am Lech, Kap. VI-3.2 Wolff B, Barbitz R, Riek W (2000): Regionalisierungvon Bodendaten auf der Grundlage von Inventurdaten des Forstbereiches, in Umweltbundesamt: Fl~ichenhafteDarstellung punktbezogener Daten fiber Stoffgehalte in B6den. UBA Texte 49 / 00, Berlin, 80-90 Received: January 30th, 2002 Accepted: July 21 st, 2003 OnlineFirst: July 23rd, 2003 JSS - J Soils & Sediments 4 (1) 2004