TABLE NOTES

¹⁾ WRI: World Resources Institute, Washington D.C., USA

²⁾ DFC: Direction de la Faune et de la Chasse, Gabon

³⁾ ECOFAC : Conservation et utilisation rationnelle des Ecosystèmes Forestiers d’Afrique Centrale, EU

⁴⁾ TREES : Tropical Ecosystem Environment observation by Satellite, Joint Research Centre, European Commission, Ispra, Italy

⁵⁾ UMD: NASA Pathfinder project, University or Maryland, USA

4.2 Field Data Collection

Data base inputs include existing data and information (see above), opportunistic information for the site and systematic surveys managed by MIKE and collaborating monitoring programs.

Field data collection forms and survey protocols for MIKE elephant and LEM surveys are presented in detail in Pilot Project Technical Reports (1 through 3). In addition, the pilot project undertook socio-economic surveys at two sites, Odzala and Ituri. Data collection forms for these exercises are presented in ANNEX 3.

We recommend that systematic data collection incorporate the following:

A rationalized survey design be developed for all data collection (this will include design unbiased and model-unbiased survey design. (See Technical Report 1). All protocols, including socio-economic and other surveys, be harmonized across MIKE sites. Data collection be made compatible with a cybertracker data collection.

4.3 Review of Base Maps

Two base maps were produced by the pilot project and contributions were made to a third one. We use the base map development for the Ituri Forest pilot site (Okapi Wildlife Reserve) to illustrate the process of database and base map development. Prior to the MIKE program, no comprehensive, current map existed for the Ituri region, and so the pilot project provided the first base map ever for this area. ANNEX 4 summarizes this base map process.

Base maps produced by the pilot project for three pilot sites are provided in ANNEX 5. Specific details for each of these maps is given in the text below.

Table 4.3 gives an overview of the sources of the features that were included in the different site base maps.

 

TABLE 4.3 Coverage features included on MIKE site base maps.

Table 4.3: Geographic features of base maps and sources of data

Interaction between the field and the lab is crucial to get good quality data, to interpret the image, to solve punctual problems with field data collection and to avoid collecting redundant or useless information.

In reality, during the pilot project, we had very limited means of communication and it often took many weeks to exchange information between the field and the lab. It would be extremely useful to invest in good communication tools (e.g. satellite phone with email link) whenever and wherever such a mapping exercise is undertaken. Field data files could be uploaded, entered in the GIS and checked in a matter of hours, so that field teams can be guided and timely informed of eventual problems.

Roads, still water and rivers were digitised on screen from the satellite image. Only the bigger rivers and some others not covered by canopy were reliably digitised. Condition of the roads was described in the field in terms of accessibility (by truck, by car, by bicycle). Roads themselves were digitised on screen from the image. Trailheads of important connecting footpaths were localised in the field and plotted on the map. These footpaths will be followed in the future using continuous GPS tracking and then plotted on the map. Navigability of rivers was also recorded in the field, but this information is still incomplete.

Park boundaries were taken from the CARPE (Central African Regional Program for the Environment) GIS database for Central Africa. Boundaries were adjusted to natural limits (rivers and roads) that were visible on the satellite image. When proper marking of the park will be done in the field, the GIS should be updated accordingly.

New, geo-referenced field data was gathered by field teams on the ground. Data input can also take advantage of aerial videography and other remote detection methods. The crucial feature is to ensure adequate geo-referencing. In Ituri, field teams recorded all villages around the park and geographic positions were taken with GPS. They collected socio-economic (presence of schools, clinics, economic activities) and demographic data. Data collection sheets for the Ituri area are shown in ANNEX 3. These forms also permit updating of existing data on these coverages. Whenever another settlement in the forest was found (mining camps, fishing camps) a GPS position was recorded.

Vegetation classification of the image is currently being done using unsupervised and supervised classification algorithms combined with field data of vegetation recorded on the MIKE transects.

The geographic component of the GIS can be further refined and more detailed geographic data (e.g. smaller rivers, forest clearings) can be added as information becomes available or as a more detailed analysis of images and other sources is carried out. Other "base map" information, which would be useful to add when available, includes topography and administrative boundaries.

Specific elements of the three base maps are further discussed under each heading below:

4.3.1 LOPE

A few topographic maps exist for the Lope region. However the more detailed 1:200 000 and 1:50 000 maps are not available for the Lope reserve itself. Rivers and roads were originally digitized from the 1:1 000 000 topographic maps. Serious problems with location accuracy and completeness for major rivers on these maps were discovered after the first draft and it was subsequently dropped. Using GPS, we also found that the existing topographic maps were not always well geo-referenced. Therefore we used radar imagery to digitise the hydrological network. It is very difficult to find good satellite images in the visible light spectrum for Gabon because of permanent cloud cover during most of the year. Radar imagery doesn’t suffer from this as radio microwaves travel easily through cloud cover. Unfortunately we had no time to verify the quality of the ortho rectification of the source image and location accuracy of the digitised map since this was done near the end of the field work.

4.3.2 ODZALA

Rivers and roads were digitized from topographic maps. A shift of up to 250 m was noted in the field between GPS and map locations. Location of human settlements was digitized from the topographic maps and verified with ECOFAC team members for their existence.

A more detailed socio-economic survey of the villages and temporary settlements south of the park was undertaken by a consultant who was hired by the MIKE Pilot project. Besides valuable data, this survey also provided an opportunity to test and develop field data collection protocols and forms for spatially linked socio-economic data. Forms and protocols were harmonized with ECOFAC data protocols to ensure compatibility of datasets with ECOFAC sites. (ECOFAC is one of the biggest conservation projects in Central Africa and manages 8 sites in the region).

An unsupervised classification was carried out by Pascale Janvier for the European TREES project on a landsat 7 image for a big part of the park. Field data collected by MIKE teams include data on vegetation and will be used to ground-truth the image and will lead to a more reliable classification.

4.3.3 ITURI

Refer in this section to Annex 4 for a summary of the Ituri Base Map process. No reliable or accurate maps were available for this area. The regional CARPE dataset was initially used but was found too inaccurate for working at the smaller scale of the Ituri reserve for which this database was not designed. Other sources of information, including digitized aerial photographs, were also explored but none of these were accurate enough and showed particular problems with geo-referencing. We decided that satellite images would give us the best base to build upon. A Landsat 5 image from the southern part of the reserve was geo-referenced with field GPS locations of features that were visible on the image. We collaborated on this with the NASA Pathfinder project, hosted at the University of Maryland at College park. . In the first step a cloud-fee or cloud-poor image was selected and Ground Control Points (GPC) were identified on the image that could be easily and precisely located in the field (e.g. a river confluence, a road crossing a river, etc.). GPS recordings of these points in the field permitted subsequent geo-referencing of the image in the lab. Accurate topographic maps could also be used as a backdrop and make digitising of geographic features much easier (features are clear and distinct), but for most of the region good maps were not available and corrections of these maps were often necessary.

Several expeditions were set up by the MIKE field teams to obtain information on human access (road condition, major river access), human presence (villages, camps), socio-economic data and major footpaths. GPS locations of all human settlements around the park were recorded and overlaid on the satellite image. This permitted a visual check of human occupied and human modified habitat.

Roads were digitised from the satellite image; condition of the road was described in the field and added to the attribute table in the GIS. Main rivers were also digitised from the satellite image.

Fine scale printouts were made and compared with previously digitized rivers from aerial photographs. Important rivers were traced on the printouts with input of expert knowledge. This aided in the digitizing process. GPS points taken in the field also helped identify important rivers.

 

4.4 Use of Base Maps by the MIKE Program

The site base maps constituted a well geo-referenced template for adding a wide range of spatially linked information on wildlife, law enforcement monitoring, human impact, economic and political features, etc.

Three lines of GIS output were particularly useful in the Pilot Project:

Base for developing elephant dung count survey plan Law Enforcement Monitoring Maps Covariate modelling for spatial analysis

4.4.1 Sampling plan

The GIS provided the base map for allocating sampling effort and laying out the sampling design. Although no stratification was done for the pilot sampling plan, we considered stratifying for accessibility, which would not be possible without proper base maps containing information on access. This process is detailed in Technical Report 1.

In the case of Odzala and Lope, a grid was generated, overlaid on the base map and sampling units (recce-transects) were located in the centre of systematically sampled grid cells. Coordinates of start and end points of recce-transects were stored in a table.

Map, table and GPS were taken into the field to locate sampling locations. The maps and GIS were also used to determine the most efficient route between sampling locations on a particular monitoring trip.

Sampling design and placement of transects has recently been automated in DISTANCE 4.0 software using Arcview compatible files. With small modifications for the proposed MIKE sampling design this program could be used in the future to generate sampling designs.

4.4.2 Law Enforcement Monitoring

Spatially linked law enforcement data can easily be entered in the GIS and plotted on the base maps. Spatio-temporal data on law enforcement effort (patrol routes, patrol frequency etc.) and law enforcement results (indicators of illegal killing of wildlife, poachers arrests, etc.) are among the most important correlates of elephant distribution, and illegal elephant killing. A GIS-produced map provides a strong visual and analytical aid for this assessment and can help identify priority areas, resource needs, management needs etc. Law enforcement monitoring data is essential for understanding the relationship between protection, elephant killing and elephant abundance. Protection is an important covariate in spatial modelling explained below.

Technical Report 3 presents the results of the pilot project law enforcement monitoring program.

5. DATA BASE UPDATES AND ARCHIVES

Data bases serve little use unless the information is updated on a rational basis, and properly archived. Data archives must include information on survey design, and sets must be archived. The same schedule for data updates need not apply for all data fields. The important point is that surveys and data analysis anticipate and respond to expressed information needs, and that information be captured in an adequate data storage system. Protocols for both updating and data archiving should be agreed to at the start of the monitoring program.

5.1 Database Updates

Data updates can be done on in a variety of ways, but systematic surveys should include timing for resurveys, and inputs into a data base should be done on a scheduled and controlled basis to ensure quality control. This latter recommendation would be modified in the case of a data base being linked to a cyber tracker, in which case updates would be linked to the cyber program.

Although the pilot project did not deal with this directly, the pilot experience nevertheless has provided some insight into the process and qualified recommendations for an updating program. We provide a range of recommended update periods. Factors affecting this include the rate of change of the phenomena in question, available funds for surveys, logistical constraints, and the availability of survey teams. Schedules for data updating will have to remain flexible, as highly dynamic situations will require more frequent surveys.

Elephant surveys: We suggest at forested sites systematic ground surveys using MIKE methods occur on a 24 to 36 month cycle, with surveys covering 9 month to 12 month periods, separated by 12 to 24 months. LEM Monitoring should occur in conjunction with anti-poaching programs, and should provide a continuous monitoring of anti-poaching activities at some level. Monitoring reports should be adapted and scheduled to provide pertinent inputs to guide anti-poaching programs. LEM summaries for MIKE should be reported in conjunction with site report forms (see Section 6 below). For the pilot project we recommended site reports on a 4 month basis (three times/year). However, monthly, 6-month or annual reporting schedules are also possible. Human access and settlement. Updates every two to five years. The choice of timing depends upon the rate of change in the survey region. Collaboration with national administrations, churches, private sector or other partners may be possible for these censuses. Socio-economic co-variates. Updates every two to five years, as above. Habitat. Updates every two to ten years, as above. Updates for these parameters should be based on integration of updated imagery. Availability of these images will determine the schedule of updates..

5.2 Data Archives.

During the pilot project we explored options of data archives with ECOFAC.

Of the pilot sites, only Odzala had a system of data archives, but this system was not evaluated during the pilot project. Within the Central African sub-region, data archives are rare and generally not well managed. Where archives occured they were usually limited to a single site, with negligible backup at the national or regional level. This is certainly an areas that will require significant input in the next phase of MIKE.

Decisions such as who and how data can be accessed, data ownership etc. must be well-understood and agreements reached before proceeding with the full-scale program. At present, MIKE data is a national product, and managed through the national MIKE structure.

Data archives are a major product of all long term monitoring programs. Cooke et al (2001) lay out basic principles and recommended practices for data archives that should be incorporated into the MIKE program:

Assign descriptive file names. File names should reflect file content includes location, survey period, investigator, file type. Use consistent and stable file formats. Use ASCII formats, and retain the same file structure throughout. The first row should contain identification information and companion file listings. Define parameters and all data fields. Be sure that the data field titles are spelled out, and that the units of measure are indicated. Use standard, unambiguous notation for dates and time and spatial coordinates. Specify codes for missing data, and unknown values. It is critical to be able to distinguish an observed "zero" from no observation made.. Use a consistent data organization. Standard practice is to use columns as data fields, and each observation in its own row. Perform regular quality control. Check for data entry mistakes by doing sort and extract exercises and investigate all outlier data values. Conduct statistical summaries of results (frequency distributions, means and variances). In some cases double data entry may be used to check for data entry errors.. Provide documentation. The documentation accompanying a data file should be written in a separate file. Do not use acronyms. Plan to have this document read up to 20 years into the future. Provide the following information for each file (or set of linked files):

Data set name Why data collected What data were collected What instruments used Who was responsible for data collection, and key participants/team members. How each parameter was measured. Summary statistics generated.

6. REPORTING

6.1 Pilot Project Reporting Forms

The MIKE site summary reporting form developed during the pilot project is presented in ANNEX 6. This form provides data on elephant survey and LEM effort, and results. Data on protected status, and basic socio-economic parameters and indicators are also provided. The site report form is to be submitted with attached elephant survey (recce transect) and LEM spread sheets. The two spread sheets are presented and discussed in detail in Technical Reports 2 and 3

The site report form is designed to be submitted on a regular basis to CITES. Originally it was proposed that the reports be made on a monthly basis by national elephant officers to their national steering committee representative. During the pilot project, we learned that logistical and financial constraints and communication problems would make it unlikely that national officers could visit all the MIKE survey sites in their country on a monthly basis. In addition, many sites had low levels of activity so that monthly visits would produce very little information. By the same token, if site reports are filed at too infrequent a basis, then information will likely be lost and the data quality degraded.

We propose a compromise timing of trimestrial report, with each report and accompanying spreadsheets covering a four month period.

Not all national officers provided final summary reports following the pilot project. This was due to a number of factors, including turnover of national officers at the end of the pilot project, communication problems, lack of computer capacity, and lack of familiarity with the data. The three pilot site officers (field team leaders), in contrast, all completed and submitted forms for their specific site, and in some cases undertook analyses as well. We recommend that the MIKE programme consider direct reporting of by site officers to the MIKE Sub-regional Coordinating Officer in some cases. Further details for reporting structure are provided in Technical Report 6 which evaluates national elephant programs.

The draft reporting form included in this report is designed to completed by each of the Field Team Leaders in a national MIKE program, in collaboration with the national officer. Copies should be retained by the Field Team Leader, collaborating NGO or other site partners.

The geographic area covered by a given site report is the area covered by the designated field team. Normally, these teams would be based in protected areas, and specifically in designated MIKE sites. Surveys, however would cover a much wider area, and the site report would summarize information over this wider area as well. Recommendations for developing a strategic approach to a national survey program are presented in Report 6.

6.2 Metadatabase

Site data can be managed in a metadatabase. This has potential to be developed to a national MIKE data management system. A simple metadatabase for spatial information was developed to keep track of incoming spatial information and spatial processing of GIS layers. A very simple user friendly interface was developed in Filemaker Pro 4 for Windows (figure 4.3) to handle data entry and retrieval.

We had often difficulties to track down the source and data of much of the original source information and maps. Besides satellite images, for which a good reference was always available, we depended on existing maps and other data in digital format but very rarely this information was accompanied by a description of the source and the quality and uncertainty of the information. It was therefore often difficult to gauge the quality of the source material and this became sometimes only apparent in the field. It is critical to ensure proper documentation of source data for site data bases in the next phase of MIKE.

The database was also useful to systematically record the updating and processing of GIS layers and spatial data. Non-spatial data such as field datasheets, reports etc could also be entered in the database.

We want to emphasize here that a system to keep track of incoming spatial information is essential for the production of a reliable product. Local data managers should be encouraged to use a system, which allows them to organise multiple files and sources of spatial information and to keep track of the important process of updating.

Figure 4.3: main window interface of spatial metadatabase for the Central African MIKE Pilot project

 

7. CAPACITY BUILDING FOR DATA MANAGEMENT

Essentially all sites will need capacity to maintain and utilize data bases, update data files, print and modify maps, and ensure some level of data analysis. Some of this capacity, in particular that which requires more advanced training, will be shared between sites. At least one person at each site should have computer capacity.

MIKE national officers should have the same basic computer capacity as the Site Officers (Field Team Leaders). In addtiona, they should have training in GIS and Access data bas management.

An evaluation of the pilot training program and recommendations for strengthening capacity are provided in Report 5..

8. SPATIAL MODELLING OF ELEPHANT DISTRIBUTION AND ABUNDANCE

Certain environmental and human factors are known to have an important effect on elephant distribution and abundance. These factors can be used in a model to explain variation in elephant densities across the study area. In this model elephant density is dependent on a set of independent variables, covariates with their respective constants that describe the relationship between the independent variable (e.g. distance from roads) and the dependent variable (elephant abundance).

An important covariate in the model that initiated the MIKE project is the CITES decision to down list the elephant from appendix I to appendix II for three Southern African countries. The original goal of the MIKE project was to show if there was a link between this decision and increased illegal killing of elephants and consequent decline of elephant abundance. However in our understanding it was generally accepted in the course of the Central African Pilot project that this causal link would be extremely difficult or even impossible to establish. It was agreed that the wider objective of MIKE was to inform the CITES debate by documenting trends in illegal killing and abundance of elephants and the factors that drive these trends.

Many variables that have an influence on elephants are spatially explicit.

For example in previous studies a strong correlation was found between distance from roads and elephant densities in Gabon (e.g. Barnes et al. 1991, Barnes et al. 1995). Areas close to roads are easily accessible by poachers and suffer from heavy hunting, while areas further away have higher elephant densities. Condition and accessibility of the road, distance to the nearest town etc. could also play a role.

Spatial variables that may have an impact on elephants in the Central African forests include:

Human access: distance from roads, navigable rivers and major footpaths Human presence: villages, temporary camps Human activity and exploitation of natural resources: forestry, mining, agriculture. Protection: distance from the boundary of a protected area, distance from protection bases, patrolled areas and patrol routes Distance from an international boundary Habitat, vegetation types Topography

Non-spatial variables include:

The CITES downlisting of elephants Socio-economic status of the country Political status of the country: peaceful areas versus war zones

A spatial model that includes spatial covariates can be fitted to the transect or recce-transect data (Hedley, Buckland and Borchers,1999) using generalized additive models.

The benefits of spatial modelling are:

an improvement in precision of encounter or density estimates and thus our ability to detect significant trends in elephant densities. an assessment of which variables are related to animal densities a density map for the study area showing projected densities of elephants (or elephant dung)

The model will help us understand the complex relationships between poaching and elephants. When temporal data become available a spatial model could be fitted to each year’s data to obtain a density estimate for any location for which a trend estimate is required (Buckland and Underwood, 2000). Trends in elephant densities can then be interpreted in relation to spatial factors affecting these.

An example analysis (annex 7) was carried out on the Odzala elephant data, modelling elephant dung encounter rates as a function of distance from roads and villages, and distance from protection infrastructure and patrol routes. The preliminary results of this analysis show an important improvement in precision (coefficient of variation was almost halved) and a projected density map of elephant dung in the surveyed area.

 

9. CONCLUSIONS AND LESSONS LEARNED

The following points summarize the major conclusions of the pilot project experience with data management and reporting, and provide recommendations for reporting for the next phase of MIKE.

1) Monitoring programs must be site based and developed at a landscape scale

Monitoring must inform site-based management. Monitoring has to provide information important to the site manager and collaborating NGOs, as well as providing information for CITES/MIKE.

Interpretation of MIKE survey and LEM results will depend upon information of the broader landscape context of the survey area.

This includes the larger spatial matrix of the surrounding human occupied areas. Issues pertinent to MIKE include:

Impact of forestry and mining activities on elephant distribution and abundance, and on illegal killing. Relationship of illegal killing to settlement, international frontiers, military operations. Evidence of trans boundary elephant movements.

2) Site base maps and site data bases are essential

Maps and their GIS linked databases have proven to be the most effective means to integrate and present a broad and diverse information base for each site. It is important to maintain the databases at the site itself where the information will be readily available. Data archives, if not at the site, must also be accessible.

A high quality geo-referenced satellite image has proven to be the best base image for a GIS platform. Good images are readily verified in the field, are the best platform on which to incorporate new data fields and new imagery.

3) Information management and transmission has not developed at same pace as data generation

Much biological and sociological information exists at a number of sites. More attention to sampling design and analysis is needed in many cases. Many surveys and wildlife inventories were conducted as one-time efforts with little consideration for further sampling. Repeated surveys over time, however, will be the only way to evaluate trends in the key features of a protected area. If the design of data collection programs is adequately simple these can be repeated over time despite changing staff and resources. Simple, standardized data collection constitutes the most valuable long-term program.

There was limited capacity to analyse, manage and transmit information at essentially all MIKE pilot sites. As a result, much information that is properly collected is inadequately utilized, and sometimes lost. There is a poor cost benefit ratio of information cost to information utility.

Computer capacity must be maintained at each site. During the course of the pilot year, three of the five field laptops for the national officers had significant problems that constrained information management and transmission. Monitoring program must budget for computer and software maintenance, and plan for data back up.

 

4) Develop monitoring programs incrementally:

Start with base map and simple databases that can be built upon and developed in steps. Focus on information that is needed immediately: This includes LEM, key faunal surveys, selected socio-economic information.

5) Train continuously and evaluate capacity frequently.

The focus should include data base management and capacity to produce and communicate specified output (maps, spread sheets, site reports)

6) Integrate MIKE with other monitoring programs from the start.

Monitoring programs must be harmonized if they are to succeed. Information needs of diverse monitoring programs will likely overlap, and this should be incorporated into collaborative design. The MIKE program must make data management and information transmission its first priority.

10. BIGLIOGRAPHY

Barnes, R.F.W., Barnes, K.L., Alers, M.P.T and A. Blom 1991. Man determines the distribution of elephants in the rain forests of northeastern Gabon. African Journal of Ecology, Vol. 29, pp. 54-65

Barnes-R-F-W; Blom-A; Alers-M-P-T; Barnes-K-L 1995. An estimate of the numbers of forest elephants in Gabon.. Journal of Tropical Ecology 11(1): 27-37

Buckland, S.T. and F.M. Underwood. 2000. Analysis of data and survey design for the mike central African pilot project: first interim report

Cook, R.B., R.J. Olson, P Kanciruk and L Hook. 2001. Best Practices for preparing ecological data sets to share and archive. Bulletin of the Ecological Society of America 82 (2): 138-141.

Hedley, S.L., Buckland, S.T. and Borchers, D.L. 1999. Spatial modelling from line transect data. Journal of Cetacean Research and Management 1, 255-64.

Hedley, S.L., Buckland, S.T. and Borchers, D.L. in prep. Spatial models for line transect sampling.

 

11. ANNEXES

ANNEX 1: Evaluation of available hardware and software (in French).

ANNEX 2: Evaluation of SE Cameroon monitoring program and database for the MIKE program.

ANNEX 3: Field data collection forms for socio-economic surveys of MIKE sites used in the Pilot Project.

ANNEX 4: Summary of Base Map development process (Example Okapi Wildlife Reserve)

ANNEX 5: Pilot Project base maps

Ituri/ Okapi Wildlife Reserve

Odzala National Park

Central Gabon / Lope Reserve

ANNEX 6: Site reporting form for MIKE (Intensive survey sties).

ANNEX 7: Spatial Modelling of Elephant Distribution and Abundance