344 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 31, NO. 3, AUGUST 2001 Integrating Expert Knowledge in Environmental Site Characterization Melek Demirhan and Linet Özdamar Abstract—The site characterization issue is the most essential task to be undertaken prior to the reclamation of a potentially con- taminated site and it is composed of sampling, laboratory analysis, and data evaluation phases. We are primarily concerned with the data evaluation phase and we utilize a recently developed adaptive areal partitioning algorithm to characterize the site. Here, we en- hance this approach by integrating expert knowledge (expert be- lief) into the fuzzy areal assessment scheme which derives informa- tion from sample data. We propose to allocate an adaptive weight to expert belief during the assessment. We compare the belief-inte- grated approach with the nonintegrated one on synthetically gen- erated sites where both uniform and biased sampling have been applied independently. In biased sampling, the zones claimed to be highly contaminated (by the expert) are allocated a higher sam- pling density. We demonstrate that the belief-integrated approach outperforms the nonintegrated one both when the expert is cor- rect or mistaken in his/her judgment irrespective of the sampling methodology. Index Terms—Belief theory, environmental site characteriza- tions, fuzzy data analysis, partitioning algorithms. I. INTRODUCTION T HE ISSUE of preserving the natural environment becomes more and more crucial in developing countries as well as the developed ones as the process of industrialization gains speed. Joint projects (see [1]) are being taken up in developed countries by concerted actions involving related units in the public administration, scientific, and industrial parties. One of the main themes in concerted actions is brownfield reclama- tion. Brownfield reclamation is important in the preservation of groundwater resources and air quality as well as the devel- opment and re-use of the field. Furthermore, the very obvious requirement that housing settlements, green parks, and agricul- tural land need to be situated on soil whose contaminants are well below tolerance levels, reemphasizes the crucial issue of land protection and reclamation. The investigation of a potentially contaminated site involves the following phases. 1) Selection of the sampling strategy (spatial distribution of samples). 2) Collection of samples. 3) Laboratory analysis. 4) Evaluation of data. Manuscript received August 12, 1999; revised April 10, 2000. This paper was recommended by Associate Editor M. Shahidehpour. The authors are with the Department of Systems Engineering, Yeditepe Uni- versity, Istanbul, Turkey (e-mail: lozdamar@hotmail.com). Publisher Item Identifier S 1094-6977(01)09792-9. 5) Assessment of the site’s contamination levels (site char- acterization). Here, we concentrate on the last phase, namely, the identifi- cation of the contaminant distribution over the site, which is the most essential task to be undertaken prior to the reclamation of the site. The aim in site characterization is twofold: 1) to identify the contaminated regions in the site as precisely as possible (the inability to identify the contaminated zones may lead to high health hazards) and 2) to carry out this task so that it results in minimum possible reclamation costs. These two contradictory goals can be satisfied only if contaminated areas are identified without leading to overestimated zone boundaries (false posi- tives). Consequently, the objective of site characterization is ex- pressed in the following statement. “Maximize the percentage of correctly recognized contaminated areas with minimum areal cover.” From now on, we call false positives “SCRAP” and cal- culate its value by dividing the total area of false positives by the site’s area. The percentage of areas (with respect to all con- taminated areas) recognized correctly as contaminated is called “COVER.” Common techniques utilized in identifying contaminated zones (hot spots) in potentially contaminated sites are collected under the general term spatial interpolation methods (SIM). These approaches are usually interpolation-based techniques used for estimating the value of characteristics at unsampled locations within an area where observations exist. Hence, SIM are used to estimate some property of the surface at a given location and they have become standard techniques used in the conversion of sample-based information into areal inferencing during environmental site investigation [2], [3]. SIM are well-established methods discussed in the literature in many other contexts [4]–[18]. SIM include popular methods such as radial basis functions (RBFs) [19], Kriging [20]–[22], minimum curvature [23], Shepard’s Inverse Distance Method [24], and triangulation [25]. An in-depth discussion and classi- fication of SIM is found in [17]. Recently, a fuzzy areal site assessment (FASA) approach has been proposed as a site characterization approach alternative to SIM [26]. FASA is distinguished from SIM by the following characteristics. 1) It does not assume an underlying random process [27], [28] explaining the distribution of concentration values from which the collected data result (such as Kriging). 2) Unlike all SIM, it does not interpolate unobserved loca- tions using the information of the neighboring data but develops inferences which involve whole areas. 1094–6977/01$10.00 © 2001 IEEE