Introduction

Over the past 50 years of international technical assistance, significant funds have been targeted towards research on problem-focused, field-scale, commodity constraints. The justification, however, for this and similar projects with an ecoregional perspective, is that demand-driven, "bottom-up" development means stakeholders, especially those only marginally connected with agriculture, will talk about all problems, not just problems related to basic commodities and livestock (Rhoades, 1998).

Members of rural communities are often eager to assess the future of their environments for a number of reasons that provoke a prevailing sense of venerability fueled by knowledge that neighboring villages are "doing better" and therefore "why can't we"; or from fear neighbors are "doing poorly" and "maybe we're next". Such reasons include stagnating yields, but also include deteriorating natural resources, declining markets, political and governance changes, dwindling employment opportunities, and endemic emigration. However, temporal and spatial interdependencies, which characterize many of their goals and problems, also contribute to a confusing range of temporal and spatial perspectives among stakeholders.

Our experiences as participants in planning workshops conducted in Colombia, Honduras, and Nicaragua led us to conclude that this confusing range of temporal and spatial perspectives among stakeholders has hindered progress toward goals of more productive, sustainable, and healthy rural environments. The symptoms have been a lack of clear objectives, a lack of precision in defining physical areas of interest, and a failure to quantify variables, all of which are indispensable for arriving at negotiated agreements for community action as well as reproducing results achieved.

An accepted premise of this Project is that natural resource management (NRM) goals and economic solutions related to agriculture often transcend field or farm boundaries, necessitating some form of collective action among landscape users; and that multiple goals can only be understood and negotiated, and problems corrected, through analyses and negotiations that explicitly consider multiple, transboundary effects.

This study focused on a "data-rich environment", which we defined in the original proposal as "locations where there exist socioeconomic data for the village administrative boundary or better, where satellite imagery and air photographs are available for multiple dates and seasons, where biophysical data such as soil type, elevation, slope, and climate exist, and where a number of sources are available of well-sampled household and farm data. Obviously such data will exist at different map scales and accuracy." It was determined to conduct a case study of one of Latin America's most impoverished nations - Honduras.

The CIAT scientists, William Bell, Ron Knapp and Grégoire Leclerc, presented a draft research proposal to a newly created funding consortium: The Trust Fund for Methodological Support to Ecoregional Programs. The principal investigators designed a project that would be synergistic with the CIAT project "Community-based Management of Watershed Resources" operating in Colombia, Honduras, and Nicaragua, and with the CIAT Land Management Project and GIS Unit, which focuses on land management using information technology, geographic information systems (GIS), and spatial analysis.

One justification for this Project, as it was developed jointly by CIAT and the Trust Fund, was to provide a contrast with another Trust Fund Project submitted by the International Potato Center (CIP). The contrast was conceived of in relation to a need to resolve methodological problems that might arbitrarily limit the range of "outside" information valuable for ecoregional research and development. This could result in narrowly conceived perceptions of development opportunities based on homogenous, historical thinking. This situation was recognized by no less an authority than a Task Force of the Consultative Group on International Agricultural Research (CGIAR), which concluded that there existed "an abysmal lack of data to validate conventional wisdom" (CGIAR, 1997).

Over the course of reporting and publishing results, it has become apparent that confusion surrounds the phrase "data-rich environment" that underlay an important raison d' être of this Project. "Data-rich" was never meant to suggest that data, in their original form and known to exist from national population and agricultural censuses, soils, and climate studies or even ad hoc farm surveys carried out by technical assistance organizations, were of sufficient resolution, quality, or general accessibility to be valuable ingredients for ecoregional studies. "Data-rich" has never meant "information-rich". Given that, the first objective of this Project was to assess the quality of potentially available data, the vast majority of which had remained "outside" the scrutiny of commodity-focused agriculturists. The potential value of these non-traditional data, however, has not gone unnoticed. Veldkamp and Fresco (1995), Verburg et al. (1999), De Koning et al. (1999), and Andriesse et al. (1994), among others, have all expressed warnings that multiple-scale ecoregional analyses that exclude demographic and socioeconomic variables do so at their own risk. This is not to say that biophysical data are any less important. On the contrary, strong arguments have been made that link the human condition of large populations of rural poor to "geographical capital" (Ravallion, 1996). This geographical capital incorporates biophysical variables such as topography, soils, and climate that are linked to cropping alternatives, market access, and so on, to development potential.

Objectives

The logical framework (Figure 1) and work breakdown structure state the Project's objective/purpose as being "To develop and document principles and procedures for building a scale-consistent database and for performing multi-scale characterizations of agro-ecosystems". This was to be accomplished by delivering four outputs.

Output 1: Quality controlled, multi-scale spatial databases for Honduras with associated methodology and training workbooks.

Output 2: Multi-scale characterization of Honduran agro-ecosystems for targeting problems, priority areas, and beneficiaries.

Output 3: Institutional capacity to supervise and guide change using multi-scale spatial analysis.

Output 4: Administration, management, and monitoring of the Project.

In the view of Project personnel, the Project objectives, its purpose, and outputs remained unaltered over the course of the Project (see Figure 1 Project logframe). It is important to note that some that are acquainted with the Project believe there was a change in Project strategy and conceptual design over its course. During the Project's course, specific activities were given greater or lesser emphasis as we gained knowledge and reassessed their contributions towards the Project outputs and final purpose.

Project outputs and products integrate easily into the Fund's Guidelines for Submitting Proposals as seen in Figure 2 of the "Generalized five-point scheme for ecoregional programs into which methodological research projects should fit". (This is taken from Figure 22, page 150, ISNAR [1998]). Also shown in red are Project achievements associated with each point.

Program strategies

Activities conducted as part of this Trust Fund Project were influenced by CIAT's approach to global activities for NRM. In general, CIAT, like its sister International Agricultural Research Centers (IARCs), tends to balance strategic and applied research activities based upon the extent to which alternative "suppliers" exist, and on donor interest. What this meant in terms of the research strategy for this Trust Fund Project was that we made efforts to forge partnerships throughout the life of the Project ranging from local Honduran government organizations, such as the General Directorate for Statistics and Census (DGEC), responsible for some of the original data sets, to advanced research organizations, such as The Royal Agricultural College (UK) and University of Florida (USA), responsible for decision support and dynamic modeling applications. Where there was no obvious and available alternative source to carry out a specific activity, Project personnel were responsible. Examples of this ranged from weeks of on-site ground truthing of remote-sensed imagery to programming application tools, such as the "Accessibility Wizard", which was developed in conjunction with the CIAT Indicators Project.

The Role of Institutional Partners

Building partnerships has been essential to this Project. Reciprocity has been a keystone. Since the beginning, the Project has made available all data sets that the Government of Honduras permitted it to release to other interested researchers. Examples of this use of the data sets range from the work of a Ph.D. candidate from Wageningen Agricultural University, to a World Bank Project, to the work of numerous nongovernmental organizations (NGOs) that used the database for planning responses to Hurricane Mitch, to an Inter-American Development Bank (IDB) project prioritizing watersheds countrywide. This will be elaborated on in the section "Dissemination, use of outputs, and impact".

Partnerships have been developed in three ways. The first phase of the Project has involved close collaboration with advanced research institutions to acquire the most up-to-date knowledge and skills. The Project was designed in collaboration with Drs. Barry Dent and Louise Fresco, and this was expected to continue. At the time, Dr. Dent was a Professor at the University of Edinburgh and a leading scientist in modeling socioeconomic ecoregional variables; and Dr. Fresco (with Dr. Veldkamp) had just finished work on the CLUE model for across-scale analysis of land use. Although both Drs. Fresco and Dent left their respective university positions, it was possible to develop other collaboration, which more than met the requirements of the Project for advanced research collaboration. We have formal research agreements with Dr. Barry Dent of the Royal Agricultural College (RAC), and with Dr. Jim Jones of the University of Florida (UFL). The RAC collaboration is based on work by various staff, while the UFL collaboration is primarily through a Dutch postdoctoral student, Dr. J.C. Luijten. Collaboration has also occurred with Wageningen Agricultural University through visiting scientist interactions by Kasper Kok and Andy Nelson. Dr. Kok has had total access to all the Honduran data, has been using the Accessibility Wizard outputs, and in return has improved the quality of the data significantly and has produced an exhaustive study of land use change in Honduras, and a calibrated model. The ITDEA was initially developed in close collaboration with the Agricultural Engineering Department of the University of Georgia.

The second way in which partnerships were implemented was in the process of building spatial databases. The Project owes a tremendous debt to the Government of Honduras and specifically to the DGEC. They believed in the Project and made data available to us that have never before been made available to anyone outside the government. Various other governmental organizations (GOs) have supplied road maps, topographic maps, and soils information, to mention only three types of information. Honduran government analysts have visited CIAT for training and participated in the training activities for use of the databases. This Project worked with the Red Nacional de Sistemas de Información Geográfica (RENASIG) network of GIS analysts (Honduras) to run one workshop; and with the Universidad Nacional Agraria (UNA) to conduct another workshop in Nicaragua. When Hurricane Mitch struck Honduras in October 1998, the Project worked in collaboration with a large body of GOs and NGOs to generate data of immediate use.

The third way in which partnerships were created was through formal capacity building. Partners in capacity building contributed to the in-country costs of workshops (travel, per diems, and staff time) and range from the donor community (Canadian and Danish), to Honduran Ministry representatives (Health, Agricultural, and Planning), to NGOs, and local political leaders. CIAT's "Community-based Management of Watersheds" Project guide has been an important platform for interaction with stakeholders, testing, and dissemination.

Finally, the Project synthesis and review workshop held in Costa Rica in July 2000 has contributed to establishing bases for new partnerships in ecoregional research, and to the design of future ecoregional research funding opportunities.

Project Design

It was stated that an accepted premise of this Project is that NRM goals and economic solutions related to agriculture often transcend field or farm boundaries, necessitating some form of collective action among landscape users. Another accepted premise is that multiple goals that arise can only be understood and problems corrected through analyses and negotiations that explicitly consider multiple, transboundary effects. This was the rationale for proceeding with a case study of Honduras, one of the region's most impoverished nations. This Trust Fund Project developed, among other outputs, extensive and unique data sets, unique multiple-scale, spatial analysis computer applications, and a forum-based procedural framework for analysis and negotiating diverse interests among stakeholders. With these data sets, tools, and methods now available, opportunities exist to formally test the premise under a variety of circumstances. For example, we have debated amongst ourselves at what point diversity among stakeholders might become so complex that meaningful collective action becomes unattainable. Project tools can now be used to design case studies to examine these situations.

The Project Proposal submitted to the Trust Fund did not explicitly specify hypotheses to be tested in a rigorous, academic sense. However, given that the Project purpose was to "develop principles and procedures for building a scale-consistent database and for performing multi-scale characterization", we hope it is clear that the implicit hypothesis was that choice of spatial scale and landscape for analysis will bias results and interpretations. This statement has implications ranging from diverse stakeholders trying to communicate their respective positions in matters of resource management, to researchers who believe "extension" and "up-scaling" research results are a simple matter of "GIS analysis" of available maps of aggregated thematic variables. Any development project that proposes "up-scaling using GIS" should be examined very carefully for the details!

Noteworthy research acclaimed by the Trust Fund Project External Review Panel resulted in methods (a geographically weighted regression, and self-organizing mapping techniques) and user application "tools" (Accessibility Wizard and Spatial Data Exploration Toolbox) that specifically addressed quantitative, multi-cross-scale spatial analysis. The function of other Project activities and achievements, however, is not dynamic, spatial, cross-scale analysis per se. Rather, those products focus on a class of related issues that are best characterized as "transboundary, on-site/off-site" issues. These are of practical importance when negotiating actual or potential conflicts of interest in resource use, such as would be carried out during an ITDEA planning exercise. An example is scenario analyses using the Spatial Water Budget Model (SWBM). Finally, the Project delivered outputs that in and of themselves have no dynamic multi-cross-scale properties, for example, GIS coverages published in the Mitch Atlas CD-ROM. The role of these outputs was to demonstrate that data are worthless unless they are used, and the more they are used the more valuable they become. The tremendous costs and efforts involved in collecting ecoregional data, such as census and survey data, amount to nothing if they are not accessible for analysis and synthesis. The CD-ROM has played a superlative role for this Project as a tool for motivating acceptance and adoption of the advanced information technology methods developed by this Project.

Opinion varies as to what degree Project focus, strategy, and conceptual design changed over its life. During the Project's course, specific activities were given greater or lesser emphasis as we gained knowledge and reassessed their contributions towards Project outputs and final purpose. The best measure of this can be found in comparing the Gantt Chart included in the original Project proposal with final results shown in the Gantt Chart in this report (Figure 1).

For example, it is clear from the original proposal that Outputs 1 (Database development) and Output 2 (Multi-scale characterization) were weighted much more than Output 3 (Institutional capacity) in terms of time and effort. We feel this balance was appropriate given the general lack of ecoregional research experience of national agricultural research systems (NARS), and emphasis on advanced information technology in this Trust Fund Project.

Within a specific output, for example Output 2, we did make adjustments over the life of the Project relative to the original Gantt chart. Management decisions were governed primarily on perceived trade-offs between pre-engineered, finished analyses (such as the CD-ROM data sets and case studies) versus flexible data sets, tools and methods that can support ad hoc decision making, but at the same time require significant investment in knowledge, time, and resources to use.

Basically, our "design strategy" followed logic such as the following. An activity (2.8) that examined trade-offs between 2-D and 3-D landscape representations was important in that it convinced us of the added value of the expense of digital terrain models (DTMs) versus scanning simpler, available flat maps. That activity caused us to emphasize the remote sensing activity (2.6), which helped us in two ways. One was to "validate/evaluate" agricultural census data. And second, it helped us with Activity 2.5, which called for a comprehensive "transport" map (which we later chose to call "accessibility" mapping).

The "transport map" was meant to be a pre-engineered, hardwired product (which has been done), but experience developing the "map" led us to an interesting (partial) solution to the "modifiable areal unit problem" (MAUP) (which is at the root of the bias in choice of spatial scale). That persuaded us of the greater importance of Activity 2.9, which resulted in the design of the Accessibility Wizard, and later, the more generic set of methods within the Spatial Data Explorer, which helps to generate user-defined, meaningful spatial systems ("health-sheds", "school-sheds", "biodiversity-sheds") and has been acclaimed by the external review panel. The development of flexible tools and methods that can support ad hoc decision making led us to formulate a forum-based procedural framework, ITDEA, that incorporates the methods for analysis and negotiating diverse interests among stakeholders. In our opinion, added value increased with each subsequent activity, but perhaps gave some an impression of change of emphasis and focus.

Who are the intended users and beneficiaries of the outputs from this Trust Fund Project? This question is answered in detail below under "Dissemination, Use of Outputs, and Impacts", but it should be clear that collective action connotes multiple stakeholders with diverse responsibilities, interests, and spatial perspectives. Depending upon issues, stakeholder action groups might include Agency for International Development (AID) Institutions, local farmer cooperatives, and organizations in between.

Our research strategy was also influenced by our sense that multiple-stakeholder decisions involving land and resource management will almost always involve degrees of subjectivity and "deal making". This belief led us to limit the development of products that might be considered "optimizing" or "prescriptive". We adopted the axiom that an approximate solution to the correct question is worth a great deal more than a precise solution to the wrong question. This position influences how we interpret methods of "scenario analysis". We believe productive, long term, resource use planning using "scenario analysis" must include plausible futures that go beyond current process simulation modeling capabilities. Methods developed by this Project (specifically the SWBM and ITDEA) include provisions for scenarios driven by a wide range of subjective system drivers.

In summary, this Trust Fund Project was designed to be synergistic with ongoing CGIAR and CIAT ecoregional projects and global initiatives. Project outputs and achievements fit smoothly into the "Generalized five-point scheme for ecoregional programs" outlined in the Fund's Guidelines for Submitting Proposals. We focused on methodological and procedural bottlenecks limiting collective action by diverse sets of stakeholders working towards development goals. These "diverse sets of stakeholders" are the beneficiaries of our work. The majority of research activities can best be described as "projective", "exploratory", and to a lesser extent "predictive". Our strategy was to develop methodologies that were less prescriptive and exclusive and more inclusive and exploratory.

Program activities

Output 1: Quality controlled, multi-scale spatial databases for Honduras with associated methodology and training workbooks.

The initial identification of sources and gathering of data sets (Activities 1.1 and 1.2) was tedious and a test of patience. Several meetings were held between Government of Honduras officials during which agreement was reached on the protocols for access and dissemination of census data. CIAT was then given access to digital copies of raw national population and agricultural census data. Digital topographic data were purchased from various sources. Published demographic and geographic data were obtained. Satellite imagery was purchased and processed, and weeks of on-the-ground validation of land use interpretation were carried out.

As data sets were collected and created, GIS analysis began validating data points through comparison studies, that is, from different sources and through different procedures (Activities 1.3 and 1.4). Finally, a CD-ROM of ecoregional variables for national coverages was released in agreement with the Government of Honduras (Activity 1.5).

To date, the Project has produced 50 copies of edited GIS data coverages as a CD-ROM entitled "Atlas de Honduras" and 2000 copies of a second version, "Atlas de Honduras: Datos Mitch" (or the "Mitch Atlas").

In addition to the data released on CD-ROM, a "shadow" database exists that the Government of Honduras has not yet decided to generally release to third parties. In collaboration with the DGEC we obtained access, for research purposes, to the 1988 Population and Housing Census and 1993 Agricultural Census at unit level (individual, household, or farm depending on the census), and loaded the entire data set in an Oracle database.

The 1988 Honduras Population and Housing Census is the most recent and complete data set about every single person and household in the country. It gives a panorama of the composition of the Honduran society and of the life conditions of its inhabitants in 1988. It contains answers that the 4,255,105 individuals gave to questions related to education level, profession or vocation, family composition, age, mortality, migration, housing type and construction materials, ownership type, water supply, assets, etc. In total, there are 42 variables for 891,298 households, and 49 for each individual, in addition to 9 variables related to administrative localization of the household.

The 1993 Honduras Agricultural Census is also the most recent data set to cover virtually every farm in Honduras (317,187 to be precise)-161 variables covering land ownership, agricultural production, technology, and labor, as well as 6 variables about the farmer, and 8 variables related to administrative localization of the farm. Many people state that agricultural censuses are error-prone, because farmers will avoid giving, to government officials, detailed information that would give the government a chance to invade their privacy.

The census results are compiled at municipal level, in tables distributed within several thick books. This tradition is likely to change soon, because most developed Latin American countries can provide municipal-level census data on line or on CD-ROM. With funding from the World Bank, CIAT is currently implementing a project in six Central American countries to help the governments develop digital data products for public distribution.

Confidentiality was partially ensured by omitting the names and addresses of the individuals. The resulting database for Honduras, the second poorest country in Latin America after Haiti, reaches a level of detail superior to what can currently be obtained for most developed countries. To our knowledge, this is the first successful attempt by a development project to build such a database for a developing country. We have created demonstration data sets of statistical and calculated ecoregional variables in hopes that we can demonstrate compelling reasons why they should be released to at least selected third parties, but most assuredly to local communities and stakeholders in need of information to guide development decisions. Early in 1999, Corporación Hondureña de Desarrollo Forestal (COHDEFOR) made available digital data from topographic map sheets at 1:50,000 scale to CIAT. This includes all the information present on the map, that is, house type, road types, contour lines, rivers, etc. We corrected and assembled all sheets into seamless coverages, and produced a digital elevation model (DEM) at a resolution of 50 m, by far the highest resolution, countrywide, ever achieved.

In addition to baseline data, the Project has built demonstration databases of statistical and calculated ecoregional variables that, in our estimation, are critical for guiding ecoregional development, but are not traditionally released to third parties. This achievement specifically addresses shortcomings in some current ecoregional work where projects have arbitrarily limited the range of information from "outside" sources. This is a survival strategy; it simplifies a confusing, ambiguous, and uncertain world. But it also leads to a narrowly conceived perception of development opportunities based on a homogenous internal mental environment. The shadow database has been fully documented and is accessible through CIAT Intranet (Activity 1.1-1.5).

CIAT scientists have contacted organizations responsible for carrying-out censuses in all Central American countries and sponsored a workshop to study ways in which to add value to the exercise of carrying out government-sponsored censuses (see http://www.procig.org/) (Activity 1.6).

Output 2: Multi-scale characterization of Honduran agro-ecosystems for targeting problems, priority areas, and beneficiaries.

The Project that was accepted for funding included both Drs. Dent and Fresco as important collaborators for outsource research. However, before the starting date of the Project, both Drs. Fresco and Dent left their respective university positions resulting in the reassignment of responsibilities. Specifically, a post-doctoral position at Edinburgh under Dr. Dent was to have researched theoretical aspects and design of tools to process socioeconomic data at multiple scales. After some discussion, this research activity was transferred to CIAT and carried out in collaboration between Dr. S. Openshaw of the University of Leeds (UK) and A. Nelson (CIAT).

Dr. M. Langford, of the University of Leicester (UK) was a visiting scientist to CIAT and collaborated with Dr. W. Bell and Ph.D. candidate, J. Cox, on methodological problems related to information extracted from remote-sensed images of land use under conditions of low thematic accuracy (Activity 2.3). J. Cox and G. Leclerc produced the high-resolution land use maps for 1985 and 1993 included in the Honduras Atlas. The Ph.D. thesis is in process of completion (Julie Cox, personal communication, 1999).

Some explanation of the course of development of the Accessibility Wizard was given above (see Project Strategies) and will be presented in more detail below (under Outputs). A. Nelson and A. Farrow developed a user-friendly ArcView extension, based of the programs originally based on Arc Info GRID. It later evolved into the Accessibility Wizard. A comprehensive manual and tutorial accompany the tool. Applications have been presented at various events (e.g., World Transport Expo, April 1999, Washington DC).

Citations of census data from developing countries can be found in many publications including those of IARCs. One of our first activities was to assess what population and agricultural censuses really describe. For this case study we quantified relationships between data extracted from remote-sensed images and data available from the 1993 Honduran agricultural census. Detailed ground-truth studies were carried out in the region of and surrounding the Tascalapa watershed in the Department of Yoro resulting in maps of supervised classification of land use. Studies were carried out that compared and contrasted several sampling resolutions of the remote images as well as comparisons and contrasts between census data and data derived from the imagery, and technical information and results have been published. Among our discoveries was that the census reported on no more than 35% of the total geographical area of the Department of Yoro. These results are not meant to raise questions about the veracity of either census data or the usefulness of remote imagery. Clearly, in this case, census data characterize some subsystem in an agro-ecological hierarchy-a subsystem pre-conceived by limited historical perceptions and of methodologies employed (Activity 2.6).

A study on the accuracy of DEMs generated by interpolation from contours at different scales (Activity 2.7) was completed in 1997 (Byne, 1997) and reported in the second Project Progress Report. As reported above, this activity convinced us of the added value of the expense of DTMs versus scanning simpler, available flat maps. This activity caused us to emphasize the remote sensing activity (2.6).

The significance of spatial structure in hillside agro-ecosystems (Activity 2.9) has been analyzed through case studies of labor productivity and natural resources in Honduras. Results were presented in the Final Review Workshop in Costa Rica, July 2000.

A research partnership was developed involving a Dutch Ph.D. student working under Dr. J. Jones at the University of Florida (Luyten, 1999). Mr. Luyten developed the SWBM as an ArcView extension, which is accompanied by a comprehensive user's manual. The application of the tool focused on a case study of effects of land use changes on the important ecoregional resource, water, for domestic, agricultural and industrial users (Activity 2.10), in the Tascalapa watershed. The shadow database was the main source of data for this research. Applications of the tool have been presented at various conferences (e.g., Systems Analysis in Agricultural Development [SAAD]-III, November 1999, Lima, Peru). This achievement is presented in detail below under Dissemination, Use of Outputs, and Impacts.

Output 3: Institutional capacity to supervise and guide change using multi-scale spatial analysis.

The development of ITDEA (Activity 3.5) was not originally designed into the Project. Preliminary decision support system (DSS) design had been done in collaboration with the University of Georgia Agricultural Engineering Department. The need for this work was recognized during the review workshop organized by ISNAR and held in The Hague on April 20-22, 1998.

The ISNAR workshop made Activity 3.2 redundant so assigned funds were redirected to the final review workshop (Activity 3.3). An important product of the ISNAR workshop was the call to articulate how the "ecoregional process" might be implemented and how the "methodologies" being developed supported the process. The result was that the Project focused two in-country workshops (Activity 3.1) on this new activity. The two workshops were held in Honduras and Nicaragua during development and testing of ITDEA. Participants represented a range of institutions from the international donor community (International Development Research Center [IDRC], Danish International Development Agency [DANIDA] and FAO), to government agencies (Ministries of Agriculture, Health, and Transportation), to universities (e.g., UNA), to NGOs (World Neighbors, Cooperative for American Remittances Everywhere [CARE]), to local village government representatives. The result has been the first version of a participant's manual for managing forums for analysis and negotiation of diverse stakeholders' goals (Nath et al., 1999).

A third workshop was held at the University of Georgia, 22-23 April 1999, during which all research partners met and reviewed progress on ITDEA. The University of Georgia collaborated on elements of the "Goal-driven, multiple-stakeholder planning forum tool" (Activity 3.3).

Following discussions with Dr. Dent and his new colleagues at the RAC, it was agreed that RAC would work on practical tools for disseminating knowledge and products developed by the Project (Activity 3.6), activities originally planned for CIAT. However, comments made over time by the ISAC, research partners, reviewers of publications, and others familiar and not so familiar with the Project addressed an issue that had ramifications related to dissemination and institutional capacity to accept Project outputs. It was determined that the Project could perform an important service by designing an informative, simple, interactive tutorial that illustrated the ecoregional process, together with key knowledge and products developed by the Project. In fact, this was considered a priority above potential traditional workbooks teaching technical tasks. This activity has been carried out by the RAC - CIAT partnership. A Web page has been set up so that all Project scientists could run the tutorial and send feedback as it was developed (http://www.royagcol.ac.uk/ciat-interface/).

The realization of the Project's final synthesis and review workshop (Activity 3.4) was approved in March 2000 and we immediately began designing the workshop and inviting potential participants. Of the 51 people invited, 25 participants attended apart from Project reviewers and Project scientists. After considering various options, we chose to team with ISNAR in Costa Rica to realize the workshop. In May 2000, an agreement was signed with ISNAR, whose Costa Rica office is providing support for the organization. We established contacts with Charles Hall, from the State University of New York, for a joint production of a book compiling:

Note that in August 1999, we asked for an 8-month extension period to the Project (January 1997-December 1999) to develop case studies, realize the final workshop, and tie up loose ends. The extension was approved in March 2000, causing some uncertainty for the management of the Project. The leaving of Project leader, Ron Knapp, at the end of January complicated this situation.

Outputs

The Project numbers among its contributions nine major achievements related to the planned outputs as follows:

Output 1: Quality controlled, multi-scale spatial databases for Honduras with associated methodology and training workbooks.

The main achievements were:

(1) A publicly available, user-friendly digital atlas on CD-ROM of maps of ecoregional variables for Honduras, the Honduras Atlas. The unique databases developed by the Project played an important role in the post Hurricane Mitch reconstruction efforts and have been recognized by the international press.

The Project has produced 50 copies of edited GIS data coverages as a CD-ROM entitled "Atlas de Honduras" and 2000 copies of a second version, "Atlas de Honduras: Datos Mitch" (or the "Mitch Atlas").

The CD-ROM opens with a friendly, menu-driven interface to a large collection of coverages including FAO and Simmons soils, geology, rivers, watersheds at several levels, roads, contour lines, slope, hypsometry, climate (monthly rainfall, minimum and maximum temperature), recommended land use, land use capacity, forestry, high-resolution land use for 1986 and 1994, mangroves, wetlands, and ecofloristic zones. They also include Olson ecosystems, protected areas, population and density for 1986 and 1988, education, poverty, annual and permanent crops by area and production, farming systems, the 1988 Population and Housing Census, and 1993 Agricultural Census at unit level (individual, household, or farm depending on the census). We constructed a friendly, menu-driven interface where users can select and combine coverages, query the databases, and apply some advanced GIS functions such as spatial joins or buffers. A tutorial is included in the distribution CD-ROM.

(2) A "shadow" database consisting of unit-level census data and large-scale information, readily available on CIAT servers.

The 1988 Population and Housing Census and 1993 Agricultural Census at unit level (individual, household, or farm depending on the census) constitute the main component of the shadow database. This census database for Honduras, the second poorest country of Latin America, reaches a level of detail superior to what can be obtained currently for developed countries. It is accessible either through ArcView or through a custom interface via ODBC or JDBC.

Another component of the shadow database consists of seamless countrywide digital data from 1:50,000 topographic map sheets made available to CIAT. This includes all the information present on the map, that is, house type, road types, contour lines, rivers, and a DEM at a resolution of 50 m.

In addition to countrywide baseline data at 1:50,000, the shadow database consists of statistical and calculated ecoregional variables that, in our estimation, are critical for guiding ecoregional development, but are not traditionally released to third parties. This output specifically addresses shortcomings in some current ecoregional work where projects have arbitrarily limited the range of information from "outside" sources. This is a survival strategy; it simplifies a confusing, ambiguous, and uncertain world. But it also leads to a narrowly conceived perception of development opportunities based on a homogenous internal mental environment.

Output 2: Multi-scale characterization of Honduran agro-ecosystems for targeting problems, priority areas, and beneficiaries.

The main activities and achievements related to this output are development of:

(3) The Spatial Data Exploration Toolbox, which includes flexible-scale, spatial data-screening computer applications and procedures to systematically "screen" samples of population statistics, for example, for high-risk health factors. Our case studies using the Honduras databases have been able to screen dozens of socioeconomic ecoregional variables (from community to regional levels), and "map" clusters of communities for most efficiently targeting development assistance.

This output specifically addresses a problem that was brought to our attention in the popular book, "How to Lie with Statistics" (Huff, 1954). A more recent book, "How to Lie with Maps" (Monmonier, 1991) is as relevant, and hopefully will become just as well known. The underlying issue is simply that manipulating the statistical sample can legitimize nearly any point of view. These tools offer protection against the unchallenged use of biased statistics, scenarios, and maps.

From the literature it would seem that the scale issue is not well understood nor thoroughly researched. When this Project was initially funded in 1997, few if any recognized texts or papers dealt with the issues of complexity and scale, and those that did were not spatially explicit. Over the last 4 years, books such as "Ecological Scale" (Peterson and Parker, 1998) and "Scale in Remote Sensing and GIS" (Quattrochi and Goodchild, 1997) have appeared, but even in books such as these, the predominant theme seems to involve defining scale and talking about scale, rather than actually doing anything about it.

The Project aims to contribute to a series of techniques and tools that allow spatial data sets to be constructed and de-constructed in a generalized yet context sensitive manner. The outputs from such techniques can be explored and described through various user-defined levels, thus revealing spatial patterns and processes that are arguably more useful than raw data or standard representations. Potentially, complex hypotheses and models can be developed based on the improved understanding that such mapping techniques provide. Additionally, the opportunity to re-express the data at different levels-levels appropriate to different decision makers-enables conflicts to be rapidly highlighted and the effects of a decision at one level to be visualized at other levels of organization. This is an essential step to address what Rhoades (1998) calls "scale wars" in project planning and execution, when each actor tries to get activities tied to the level with which he is comfortable and have comparative advantage.

The identification of appropriate scales for analysis and prediction is a challenging problem. Although the factors producing scale-dependent patterns may not be clearly understood, we have been able to create accurate and reliable descriptions of scale-dependent patterns and processes to design data sampling procedures and test the accuracy and reliability of methods of prediction. There is clearly some way to go before scale effects can be fully understood and accommodated, but this research has aimed to be the "next step" in that vital process.

The Spatial Data Exploration Toolbox includes tools to productively manipulate statistical samples in an exploratory "screening" of populations. Figure 3 illustrates an analysis resulting from the unique ability to delineate spatial "hotspots" at different scales using the Geographic Analysis Machine, one of the tools from the University of Leeds that is part of the Data Exploration Toolbox. The figures are composite images of several analyses across scale, highlighting the problem areas in small communities (small spots) through to regions (larger patches).

Probably the greatest methodological challenge this Project has tried to address has been known for over 50 years, and amongst geographers is called the MAUP. Symptoms of the problem are that the choice of both spatial scale and region of analysis will bias results and interpretation. There is no known cure. What is surprising is that so many books, papers, and projects dedicated to issues of multi- and across-scale analysis fail to address the issue and, in most cases, fail to acknowledge its existence. This Project has gone beyond researching effects of the MAUP; rather we have adapted available procedures to use the problem to advantage, by investigating alternative spatial representations of key variables, and by allowing census data to be re-expressed at new ecoregional levels (Output 4).

(4) The Accessibility Wizard, a computer application that evaluates and maps "accessibility" and allows the definition of a new, flexible, ecoregional areal unit - the econoshed.

"Accessibility" is arguably the single most important ecoregional factor driving land use change and development in general. It is surprising, therefore, that initial work on multi-scale analysis by the research team that developed the CLUE model that inspired this Project, did not include "accessibility" or even mention it as a potential driver for land use change. We are in agreement with Geertman and van Eck (1995) who said, "The concept of accessibility is used as a means in rural development policy, an indicator of rural deprivation, and as a variable in location analysis".

Deichmann (1997), in his report on software development for assessing accessibility, describes four "accessibility indicators" that the development community is using for assessing "service provision", for example, in the health sector. Although this use is undeniably important, the Project wished to explore other uses for which the methodology would be valuable. To mention a few examples:

The analytical model is a cost-distance, spatial, cell-based approach that defines the ease each cell (i.e., location) can be accessed from other locations, in this case villages, by generating something called a friction or cost surface.

Figure 4 shows accessibility maps (in terms of travel time) that are a variation on the theme "potential accessibility indicator". Combining data for roads before October 1998 and 3730 villages (called aldeas in Honduras) created the map in Figure 4a. Villages represent a set of sources and targets. From this basic network, least-cost transport routes were calculated taking into consideration geographical factors impeding movement, such as slope and land cover. It is important to keep in mind that this map represents only one of several possible stakeholder perspectives of the concept of accessibility. Our goal was to create a tool flexible enough to incorporate a wide range of perspectives and indicators.

In October 1998, Hurricane Mitch devastated Honduras. Virtually overnight, the accessibility map shown in Figure 4a was made obsolete. An updated map (Figure 4b) was developed within hours of the hurricane and made available to and used by relief agencies to help target disaster relief. This use of the Accessibility Analysis Wizard and the ecoregional database on CD-ROM may have been serendipitous, but it is a powerful example supporting our initial hypothesis-that narrowly conceived perceptions of development opportunities arbitrarily limit the range of valuable "outside" information. Recognition of this achievement is reported under "Dissemination, Use of Outputs, and Impacts" below.

Generating alternative ecoregions - the Econoshed.

Modern agricultural location theory stems from the seminal contributions of David Ricardo and Johan Heinrich von Thunen nearly 200 years ago. Today most of the spatial variation in agricultural land use depends on only two factors:

We propose that the concept of accessibility, based on the spatial structure of markets, transport networks, and topography, could provide a common framework in which to attempt such a synthesis. Surfaces of accessibility define "catchment" areas for each market. These catchments ("econosheds") are a new set of user-defined areal units ideal for describing ecoregions, and as such are a powerful extension to the Von Thunen model. Figure 5 illustrates one possible set of these new ecoregions that are neither political units nor watersheds.

(5) A Spatial Water Budget Model (SWBM), a scenario-building tool to support local decision making on water security issues.

The SWBM is intended to support local decision making and for teaching local stakeholders about basic functions of multiple-community watershed components, such as relationships between land and water resources, effects of land use and demographic changes on future water accessibility, and upstream-downstream relationships (Output 2).

The SWBM addresses a classic NRM issue - water security - and is meant to facilitate scenario analysis during multi-stakeholder planning forums. As mentioned above under "Program Strategies", the SWBM is not an application to simulate or characterize dynamic, spatial, multi-cross-scale, system drivers. Rather, it is used to articulate a subset of issues known as "on-site/off-site" relationships. The SWBM has restricted "scales" of use, meaning that model driving variables are "pre-engineered" and may not simulate processes important at highly detailed spatial and temporal scales.

The SWBM has been developed using hillside watersheds in Latin America and the Caribbean of up to about 50,000 ha and for which limited biophysical data are available. Before the SWBM was developed, a thorough evaluation of many available water-based simulation models was made including AGNPS, ANSWERS93, SWRRB, SWAT, PRMS, ARC/INFO GRID, ARCVIEW Spatial Analyst, BASINS, and OWLS. None met all our criteria. In general, they are too complex, too data demanding, or incapable of simulating and manipulating stream flows or simulating flow control structures, for example, dams. The SWBM fills the gap between the current needs and resources of rural communities throughout developing countries, and data demands of established, more complex, multi-faceted models designed for resources and applications in developed countries. The model is not designed for engineering specific hydrological projects or for describing the movement of water and soil based on detailed physical processes. This simulation model was developed around the belief that an approximate answer to the right question is worth a great deal more than a precise answer to the wrong question.

Point 1 of the generalized five-point scheme for ecoregional programs outlined in the Fund's Guidelines for Submitting Proposals is "problem definition and delineation of the ecoregion (or resource management domain)" (Figure 2). There are, of course, popular methods for delineating ecoregions. One example is the use of watersheds as an organizing unit for focusing ecoregional resource management. This is a logical line of research for a project focusing on integrating data across geographic scales because "watersheds" are naturally organized into biophysical, and in some cases socioeconomic, hierarchical ecosystems. Few will deny the role water plays in all ecosystems, but especially those incorporating agricultural-based rural communities. As related above, however, a major achievement of the Project was the refinement of procedures to define a new ecoregional analysis unit, the econoshed, which represents accessibility domains and as such is subjected to the definition of poles of interaction, access routes, topography, and land use.

Point 3 of the generalized five-point scheme for ecoregional programs outlined in the Fund's Guidelines for Submitting Proposals calls for "dynamic representation(s) of ecoregional processes that permit evaluation of alternative land use options and management". A case study demonstrating the value of the SWBM was carried out for a benchmark watershed, Tascalapa, in Central Honduras. Results are being made available on the Worldwide Web as part of the Project's tutorial.

Based on the results, one can make statements such as, "In the second half of April, about 20% of the residents in the watershed have to walk more than 1 km to access any stream, and 30% have to walk over 5 km before they reach a point where stream flow is greater than 50l/s" (Figure 4). Not surprisingly, the poorest families and those settling the "frontier" of the watershed are often those without piped water and therefore may rely on stream water for domestic needs. The analysis also has implications for gender studies.

Based on the above analysis, decision makers could, for example, set goals for minimum flows at the watershed outlet during the driest seasons so that all farm families have access to water within a set distance from their homes. Such flow requirements have further implications: Are complexes of small dams and reservoirs necessary? What are acceptable targets for water extraction and what governance mechanisms are required.

A second type of analysis (Figure 5) carried out with the SWBM demonstrates the effects of different land use patterns on water yields. Current land use varies significantly between zones within the watershed, which in turn affects relative water yields and temporal variability. Using the SWBM, scenario analyses were carried out describing the effects of plausible development paths on future water yields and patterns. Yet another analysis looked at equity issues, that is, which zones in the watershed were net suppliers or net consumers of accessible water.

Output 3: Institutional capacity to supervise and guide change using multi-scale spatial analysis.

The main achievements under this output were:

(6) The Intelligent Team Decision Assistant (ITDEA), a goal-driven, multiple-stakeholder planning forum tool, that guides stakeholders through a process. ITDEA takes generic concerns and develops them into concrete statements of values and goals in a way that allows all stakeholders to see their significance enabling them to negotiate collective action within a collaborative environment. The method was specifically tailored to facilitate the ecoregional process outlined in the proceedings of the ISNAR workshop held in The Hague in April 1998.An interactive, computer-assisted tutorial, which "walks" users through the "ecoregional process", has been designed to demonstrate how the key concepts addressed by the Project can work together harmoniously.

ITDEA: A Goal Driven Multiple Stakeholder Planning Forum Tool

This output addresses shortcomings in ecoregional projects that are built from an initial problem-oriented analysis. A most obvious deficiency is that these ecoregional research and development projects will invariably overlook prevention of future risks to currently desirable, non-problematic conditions. Classic examples are the sudden collapse of the health of common-pool resources such as water for domestic, agricultural, and industrial needs. This methodology focuses on the future. It is specifically tailored to facilitate the ecoregional "process" outlined in the Fund's Guidelines for Submitting Proposals (Figure 2). It asks stakeholders to think about how their world could be different, and it presupposes that "breakthrough" thinking greatly benefits from outside participation in the form of collective action.

This activity can trace its beginning to work carried out at CIAT starting in 1992. Six tasks were identified as crucial for the success of consortia in managing development projects that require collective action (Knapp et al., 1999). The Forum Tool (Figure 8), now called ITDEA (Intelligent Team Decision Assistant) includes the following nine-steps:

An interactive, computer-assisted tutorial that "walks" users through the "Ecoregional Process".
The collaborators from RAC plan to follow up work initiated by this Trust Fund Project by developing a user-friendly, decision-support software application. Any DSS is complex in its use. The assumptions underlying models may be opaque at best. The way a user works through a hierarchy of choices and decisions can be anything but intuitive. The output of a query can seem illogical without explanation. At the same time, it can be distracting to the point of uselessness to have explanatory messages, prompt boxes, or confirmation requests popping up every time you use an information system in real working situations. To overcome this, a tutorial on using the DSS is being developed alongside the DSS itself. This tutorial mimics the DSS and is designed to look like the full working system. However, the processing is simulated, the range of localities and scenarios are restricted, and the output is pre-formed. Each stage is accompanied by text that explains what is happening at that stage, what the consequences of any choices will be, and the significance of the output generated. There are also feedback forms to allow the users to communicate directly with the development team so that they can modify the tutorial in line with user comment. Depending on CIAT research priorities and funding, the tutorial will ultimately be packaged with the full DSS so that it can give context-sensitive tuition at any stage of use of the DSS itself. During this phase of the Project, it is being developed before the full DSS, as a proof of concept for potential users of Project outputs. In light of this proof, the DSS would be refined, the working interface designed, and then the tutorial and the DSS integrated. The tutorial (in construction phase) can be accessed at http://www.royagcol.ac.uk/ciat-interface/. Figure 6 outlines the architecture of the tutorial.

(7) The Project review and synthesis workshop, "Scale effects in decision making for ecoregional development", is a major achievement that will result in an integrated view of the ecoregional process and on the role of information technology by a multidisciplinary panel of international and regional experts. The individual contributions as well as the result of the roundtable discussions will be assembled in a book that we hope will represent the state-of-the-art on the subject. We are confident that we can generate a primer that a major publisher would be interested in producing.

The Project review and synthesis workshop was a critical component of this Project. The workshop went beyond the scope of this particular Project. Participants included regional and international partners as well as experts working on related research, with one thing in common: the need to work across scales, with multiple stakeholders, and with information technology. The role we expected from the outside presenters was to present their work so that together we might assess our capacity to contribute to humanitarian development efforts through an ecoregional perspective, which is to say: to comment on the quality of the science and research management achieved by this Project and to call to our attention any potentially bad habits or pitfalls that might jeopardize our future efforts and those of our colleagues involved in other ecoregional projects.

The 3-day workshop took place in San Jose, Costa Rica, 4-6 July 2000. The official language was English. During the first 2 days, the "experts" presented their work in 30-minute talks to establish the research environment for cross-scale applications. The morning of the third day focused on tools, where participants had opportunities to experiment with software tools developed by the Project. A dinner marked the closure of the workshop. The outline of the upcoming book will be announced. During and after the workshop, Project personnel worked with the Project External Review Team to facilitate the review process.

Project researchers presented the following case study analyses: