This is the fourth African Elephant Status Report (AESR) produced under the aegis of the African Elephant Specialist Group (AfESG) of the IUCN Species Survival Commission (SSC). Like its predecessors, the AESR 2007 is based on data from the African Elephant Database (AED), the most comprehensive database on the conservation status of any single species of mammal in the wild. The AESR 2007 aims to provide the most authoritative, comprehensive and up-to-date source of knowledge on the distribution and abundance of the African elephant at the national, regional and continental levels.
One continuing challenge for the AESR is to interpret apparent trends in elephant numbers, particularly at the continental level. This is a tall order, as large gaps remain in our knowledge of elephant distribution and abundance across their range. Furthermore, guesswork still accounts for a large proportion of the elephant numbers reported in the AED, and an unknown number of elephants remain unaccounted for in the database.
Clearly, comparing guesses to derive population trends is a meaningless exercise. Yet such comparisons of elephant numbers continue to be done by many on a regular basis, despite the AED's existing data categorization systems and repeated warnings in each edition of the AESR. To help ensure that only valid comparisons will be made in the future, several new features have been developed and implemented in this report.
The first of these features is a system for tracking changes in elephant numbers between this and the previous report, at the national, regional and continental levels. Each section now includes a table showing changes in elephant numbers grouped by the attributed causes of any reported change. In effect, the system separates those apparent changes where valid comparisons can be made (repeat surveys) from the rest (e.g. new guesses, different survey techniques, etc.). Where methodologically comparable data account for a large proportion of elephant numbers at the regional level, a statistical analysis of changes since the previous report, as described by Blanc et al. (2005), is also presented.
In order to ensure the correct interpretation of elephant status, it is important to give readers an intuitive feel for the limited quality of elephant data available. To this end, a new Information Quality Index (IQI), calculated from data contained in the AED, has been developed. The IQI assigns a score from zero to one for every country, region and the entire continent, and it should enable readers to understand and compare how data quality varies from one place to another.
Based on the IQI, a system to identify those areas for which population surveys are most needed has also been devised. The Priority for Future Surveys gives a score of one (highest priority) to five (lowest) to every site, country and region, reflecting the quality of data and the need to conduct systematic population surveys. This system is intended to assist managers and donors in prioritizing their elephant population monitoring efforts, an important exercise in view of limited resources for survey work.
Readers will notice that the convention previously used for titling the AESR has been changed for this report. In the past, the title African Elephant Status Report (or African Elephant Database in editions prior to the AESR 2002) was followed by the year to which the most recent information in the report referred. For instance, the African Elephant Status Report 2002 (Blanc et al., 2003) contained data up to the end of 2002, even though the report itself was published in 2003. This led to considerable confusion and incorrect citations in the published literature. The AESR 2007 and future editions will be titled using the year in which the report is published. Thus, the African Elephant Status Report 2007 (this report) contains data gathered up to the end of 2006. We hope that this new convention will provide greater clarity to readers.
The new convention may give the impression that five years have passed between the publication of the AESR 2002 and the AESR 2007, whereas, in fact, it has only been three. Nevertheless, financial constraints continue to make it increasingly difficult to produce the AESR at three or four year intervals. Resources permitting, we anticipate that in future the AESR will be published every five years.
The AED and its status reports have come a long way in the 15 years that they have been under the responsibility of the AfESG. The production of the AESR has faced many conceptual and technical challenges in that period, and the AfESG is keen to share the lessons it has learned. We would like to see similar monitoring systems adopted for other charismatic/iconic species. The AfESG believes that expanding the AED to include other such species, thereby creating a multi-species database, would result in an even more valuable resource, with better prospects of long-term financial sustainability. We would welcome interest from those wishing to take this concept further.
The African Elephant Database is a spatial database used to store, manage, analyze and disseminate information on the distribution and abundance of elephant populations on the African continent. In order to provide a current and accurate picture of the status of African elephants, the database is regularly updated, and African Elephant Status Reports are produced and published periodically.
Information on elephant range and numbers is vital for the effective conservation and management of Africa's elephants. The elephant is a “keystone” species that plays a pivotal role in structuring both plant and animal communities (Dublin, 1995; Owen-Smith, 1988; Shoshani, 1993) and often dominates mammal biomass in the habitats it occupies (White, 1994). While the effect of the African elephant on its habitat is often beneficial (Cochrane, 2003; Magliocca et al., 2003; Nchanji & Plumptre, 2003; Or & Ward, 2003; Ruggiero & Fay, 1994), it can have a detrimental impact on vegetation where high densities build up in confined areas (Craig, 1995; Jachmann & Croes, 1989; Swanepoel, 1993; Tchamba & Mahamat, 1992; Western & Maitumo, 2004).
The potential impact of elephants on their habitats raises important management issues for protected areas. It is pertinent to ask, for instance, how large a protected area needs to be to support a viable elephant population without negatively affecting biological diversity (Armbruster & Lande, 1993). Conversely, and in order to prioritize efforts in elephant conservation, it is necessary to define minimum viable populations within isolated protected areas (Sukumar, 1993).
Elephant distribution, however, is not confined to protected areas. Indeed, the majority of elephant range may still be found in unprotected land. This poses additional challenges for wildlife authorities and wildlife managers (Kangwana, 1995). Levels of human-elephant conflict, for instance, are high in many parts of the continent, and especially where human and agricultural expansion moves into new areas (Hoare, 2000).
In order to meet these challenges, it is essential that management objectives be clearly defined for both protected and unprotected areas of elephant range (Lindeque, 1995; Lindsay, 1993). Information on elephant distribution and abundance must be available in order to set such goals, as well as to monitor the effectiveness of management actions.
In summary, wildlife management authorities need to know the status of their elephant populations, whether they are increasing or decreasing and whether their numbers should be regulated to reduce conflict and to relax the pressure on habitats.
The status of the African elephant varies considerably across its range, and the long-term survival of national populations is more threatened in some countries than in others. While the desire to conserve elephants is widespread, opinion differs as to how this goal can best be achieved. It is difficult, however, to make objective decisions about elephant management and conservation within and beyond protected areas without the sort of overview that a synthesis of continent-wide information can provide.
Continent-wide information is required because elephants move long distances across protected area boundaries and international borders, and a policy or management decision made in one country can affect elephant populations elsewhere. Changing land-use patterns or different approaches to tourism, such as trophy hunting in border areas, may have impacts beyond sovereign boundaries. Likewise, policies concerned with ivory management and trade, in particular, can transcend political boundaries. Many argue, for instance, that trading by one country could affect poaching or smuggling in another, and that any management action which, directly or indirectly, leads to fluctuations in the price of ivory, could ultimately affect the future of the continent's elephant population (e.g. Bulte et al., 2003; Douglas-Hamilton, 2000). Regardless of whether this view is correct (Kantai, 2000; Stiles, 2004), monitoring at the continental level is necessary.
Civil instability and wars often lead to the mass migration of refugees into previously uninhabited areas of elephant range. Several important Range States are emerging from armed conflict, and have little or no capacity to monitor their elephant populations. These factors all make it difficult to partition elephant management into clear political units.
While regional initiatives, such as the Southern African Elephant Survey and Monitoring Programme (ELESMAP), which involved most Southern African Range States in the 1990s (Craig, 1996a), are necessary to census and manage shared, cross-border populations, a continental perspective is also of utmost importance for identifying conservation priorities at the regional and continental levels.
The AED was initiated by Iain Douglas-Hamilton in 1986. The objective of the project was to develop a comprehensive picture of elephant numbers and distribution throughout Africa. Using data ranging from systematic survey results to guesses collected in questionnaires and interviews, a database of elephant population estimates and distribution was assembled (Burrill & Douglas-Hamilton, 1987) using a Geographical Information System (GIS).
Initially housed at the United Nations Environment Programme (UNEP) headquarters in Nairobi, Kenya, the AED was from its inception until April 1998 a collaborative effort of the Global Environment Monitoring System (GEMS), the Global Resource Information Database (GRID) of UNEP and the IUCN/SSC AfESG. Towards the end of 1992, the AED became the direct responsibility of the AfESG, which had by then become a separate group from the African Rhino Specialist Group (AfRSG). In April 1998 the AED was moved from UNEP to its present location in the AfESG offices in Nairobi.
Since 1992, the structure and management of the AED is overseen by a group of technical experts known as the Data Review Working Group (DRWG). The DRWG meets periodically to review and discuss technical aspects of the AED. The DRWG oversees the selection and categorization of data to be included in the AED, agrees on new features and analyses to be implemented in the AED and reviews the technical content of the AESR. Decisions made by the DRWG are implemented by a full-time database manager.
Prior to the present report, three reports of the AED were published under this framework, namely, the African Elephant Database 1995 (Said et al., 1995), the African Elephant Database 1998 (Barnes et al., 1999) and the African Elephant Status Report 2002 (Blanc et al., 2003). These reports are freely available for download, in PDF format, from the AfESG website (http://iucn.org/afesg). It is a testimony to the success of these reports that, in the three years since it was first posted on the website, the AESR 2002 has been downloaded no fewer than 150,000 times.
At a meeting held in Nairobi in September 2002 under the auspices of the programme for Monitoring the Illegal Killing of Elephants (MIKE) of the Convention on the International Trade in Endangered Species of Fauna and Flora (CITES), African elephant Range States unanimously adopted the AED as the official repository of African elephant population data generated by the MIKE Programme (MIKE, 2002a,b).
The AED contains both spatial and non-spatial (attribute) data, managed using GIS software and a relational database management system (RDBMS). Spatial data layers are currently maintained in ArcGIS 9.1 Geodatabase format within a Microsoft Jet (Access) database, and are stored as polygons or points depending on the geographic feature type. These data sets are combined with base map data derived from the Digital Chart of the World (ESRI, 1992), a widely available global geographical data set.
Survey reports are obtained from wildlife management agencies and other organizations, and questionnaires and maps are distributed to AfESG members and other individuals with possible access to reliable information on elephant distribution and abundance. Data are received in a variety of formats, including paper maps, reports, personal communications and geo-referenced digital data. Information from paper maps is digitized and geo-referenced, while attribute data from reports, communications and questionnaire replies are entered through the keyboard.
New data are conflated with existing data and boundaries are adjusted to rivers, lakes, and political boundaries of the base map. Spatial polygon data are maintained in geographic (unprojected) coordinates in degrees of latitude and longitude. When it is necessary to calculate areas, the data set is projected “on the fly” into the Lambert Azimuthal Equal-Area projection which, when applied to the African continent, results in minimal distortion to shape, distance and direction while retaining true area information. The surface areas of input zones, protected areas and elephant range as calculated by the GIS are aggregated at national, regional and continental levels to determine their total surface area. In addition, the overlay capabilities of GIS are used to determine percentages of both protected and surveyed elephant range.
A number of genetic studies published in recent years have suggested that the previously recognized subspecies of African elephant, namely the savanna elephant Loxodonta africana africana and the forest elephant Loxodonta africana cyclotis, may, in fact, constitute two separate species, namely Loxodonta africana (Blumenbach 1797) and Loxodonta cyclotis (Matschie 1900) respectively (Comstock et al., 2002; Roca & O'Brien, 2005; Roca et al., 2001, 2005). Although many have been quick to adopt the specific distinction between forest and savanna elephants, the above studies have been criticized on the grounds that their sampling was insufficiently extensive. There is still no consensus in the scientific community as to the number of species of elephant currently extant in Africa (Debruyne, 2005; Debruyne et al., 2003). In addition, the existence of a third species, a West African elephant inhabiting both forests and savannas in that region, has also been postulated (Eggert et al., 2002).
In 2003, after carefully reviewing the evidence available, the AfESG agreed that, in view of the lack of consensus among experts in elephant genomics, the premature allocation of African elephants into separate specific taxa would leave hybrids in an uncertain taxonomic and conservation status, and that more research is needed before such an allocation can be made (AfESG2003b). In conformity with this view, the AED and its outputs continue to treat African elephants as a single species.
The AED stores data on two basic variables reflecting the conservation status of African elephants, namely, distribution and abundance. The collection of data on these variables presents a number of challenges related to the availability and quality of information. These challenges, and the ways in which the AED has been designed to handle them to assist in proper interpretation of data, are described in detail in the subsections that follow.
African elephants occur in a wide variety of habitats, from tropical swamp forests to deserts. Elephants often move extensively in search of food, water and minerals or in response to disturbance, and the extent to which they move may depend on a large number of factors. In certain areas, seasonal movements are predictable, while in others, movement patterns are far more difficult to decipher. These factors, together with the scarcity of animals at the edges of range, make elephant range a complex concept to define and pin down. For these reasons, elephant range is broadly defined by the AfESG as the entire area where the species occurs in the wild at any time.
Collecting precise distribution information on such a wide-ranging species as the African elephant presents a number of practical problems, often related to the remoteness and challenges posed by some of the habitats in which elephants are found. As a result, the quality of information varies considerably from one area to another. The range map for a particular country is often updated by a single individual answering a questionnaire, and thus subjective elements inevitably affect the collection of range information. Trying to draw a precise range boundary on maps of varying quality and scale is often an arbitrary exercise. Neat, rounded lines may be indicative of scanty knowledge in comparison to the fragmented, more detailed pictures which emerge from countries where more precise information is available. Elephant range often fits precisely the boundaries of protected areas, because that is where most population surveys are carried out, and elephant movements in and out of protected areas are often ignored.
Frequently, the depiction of range is also delimited by a natural boundary such as a river or a mountain range for convenience rather than accuracy. When range information in one country extends to a national border, it does not always match the range in the neighbouring country. While this is sometimes due to steep gradients in human population density across the border, more often lack of reliable information is the cause of the hard boundaries.
In order to address some of these difficulties, the AED classifies elephant range information into four categories of certainty, as described in Table 1. In addition, range information in the AED is fully documented and referenced to original sources of data. This allows some evaluation of the reliability of range information, based on how and when each record was obtained.
Table 1. Categorization of elephant range data in the AED
|Known||Areas in suitable habitat which, if searched with reasonable intensity, are likely to yield signs of elephant presence. If no information is obtained confirming the presence of elephants for a 10 year period, known range is downgraded to possible range (below).|
|Possible||Areas within historical range and in suitable habitat where there are no negative data to rule out the presence of elephants, including former areas of known range where the source information is more than 10 years old. Areas of possible range are considered to be a priority for studies to establish the presence or absence of elephants.|
|Doubtful||Areas where there are reasons to believe that elephants are no longer present, but which have not been formally surveyed. If further corroborative evidence is obtained, areas of doubtful range are reclassified as non-range. As with possible range, areas of doubtful range are a priority for absence/presence studies.|
|Non-range||Areas which are known to hold no elephants – be it due to habitat modification, local extinction or any other reason.|
|Point records||Sightings of elephants or evidence of their presence outside of known elephant range, shown as crosses on the maps.|
Many areas of possible range shown in previous editions of the AESR had not been updated in many years, and were therefore unreliable. It continues to be difficult to obtain updated information for many parts of elephant range, and yet many factors may have changed in such areas since the data were first collected. Human population density and habitat loss, for instance, are known to be major factors affecting elephant distribution. Hoare and du Toit (1999) found that people and elephants can coexist up to a threshold of human population density of 15.2 persons per km2, beyond which elephants are absent. As human populations have continued to increase throughout much of the continent, an attempt has been made to update and improve the reliability of range data in this report. To this end, human population data was obtained from the Landscan 2002 human population density data set (ORNL/GIST, 2002). This data set models the spatial distribution of human population density by incorporating data from census counts and spatially distributing them based on a number of other factors, such as distance from roads and night lights as seen from space. For this report, Landscan 2002 data were overlaid with elephant range data from the AED, and areas of possible elephant range where human population density is estimated to exceed 15 persons per km2 were categorized as doubtful range. As mentioned in Table 1, it is important to conduct studies in these areas to establish the absence or presence of elephants.
Although a wide variety of methods are available to arrive at estimates of elephant numbers in an area, no single method is perfect. Possible sources of bias include the choice of survey technique, surveyor skill, quality and availability of adequate equipment, financial constraints, climatic conditions and vegetative cover. Ideally, data on elephants in any country should be collected by a wildlife management authority using qualified staff and standardized methods for collecting, recording and analysing data (e.g. Craig, 2004; Hedges & Lawson, 2006). In reality, elephant data are often collected by a multiplicity of agencies and individuals, often without any direct linkage to one another and using a variety of different techniques based on current opinion and available resources. The result is a collection of data of variable quality in most countries, and no data at all on many populations. Very few countries have the means, either financial or in the form of expertise, to conduct systematic surveys on a regular basis. Furthermore, political strife plagues many Range States and precludes monitoring work.
Elephants are often found in unprotected landscapes, where few surveys are undertaken. In some countries, elephants inhabit different types of habitat and it is necessary to combine data from different types of surveys to calculate a national estimate. Seasonal and cross-border movements of elephants are additional factors that can lead to inaccurate national estimates. To date, there have been few cross-border surveys to estimate the size of such populations. Instead, they are treated as separate populations on either side of the border, which may occasionally result in either under- or over-counting.
There is no ideal method for counting elephants. Each method has its advantages and disadvantages and is applicable in a different situation. The brief description of some of the most important methods below is not intended to be detailed or comprehensive. For more details, the reader is referred to the specialized texts available on the subject (e.g. Barnes, 1993; Craig, 1993; Craig, 2004; Douglas-Hamilton, 1996; Hedges & Lawson, 2006; Kangwana, 1996; Norton-Griffiths, 1978).
Methods of Estimating Elephant Numbers
Methods for estimating elephant numbers fall into three broad categories: total counts, sample counts and guesses.
Total counts aim to see and record all the elephants in a defined area, either from the air or from the ground.
Aerial total counts are conducted from fixed-wing aircraft or helicopters, and are only suitable for open habitats, where elephants are unlikely to be hidden by forest or thick bush. The speed at which the aircraft is flown also influences the accuracy of the count, with high speeds usually leading to undercounts (Norton-Griffiths, 1978). Aerial total counts are commonly used in savanna habitats, especially in Eastern and Southern Africa.
Total counts of a limited area can also be conducted at ground level by teams in vehicles or on foot. These are uncommon in Africa, but are sometimes carried out in India, where observers ride on domesticated elephants.
In a handful of places, total ground counts have been accomplished by identifying every individual in the population. This is only possible for intensively studied, closed populations where animals can be observed readily. For such individual recognition studies to provide high quality data for the AED, every individual in the population must be registered. Many ongoing studies have so far only covered a fraction of the populations being studied, and cannot therefore provide reliable estimates of entire populations.
Sample counts, in which only part of the area is counted (usually between 3% and 20%), are generally conducted along transects which may be randomly distributed or systematically placed across the study area. The resulting data are used to calculate a population estimate with confidence limits. In contrast with total counts, which tend to produce underestimates of the true population, sample counts have in principle an equal chance of underestimating or overestimating the true population, provided that sampling error is the main source of error. In practice, however, factors such as high aircraft speed or dense vegetation cover will lead to undercounts. Sample counts can be further subdivided into direct sample counts and indirect sample counts.
Direct sample counts are most commonly made from the air, but may also be conducted on the ground, either on foot or from vehicles. Aerial sample counts require considerable technical expertise and coordination, as well as the use of expensive equipment such as radar altimeters. Aerial sample counts are the most commonly employed survey technique in Eastern and Southern Africa.
Indirect sample counts are usually the only way in which to obtain objective estimates of elephant populations in forests, where it is difficult to see any animals. Elephant dung is counted along transects using line transect sampling techniques, and results are combined with estimates of elephant defecation rate and dung decay rate, to provide a population estimate with confidence limits. Dung decay rates vary considerably across sites, and an estimate of decay rate obtained from the study area is crucially important to arrive at an accurate estimate of elephant numbers (Laing et al., 2003). If properly conducted, dungcounting techniques can provide estimates that are at least as accurate as those from direct methods, and more precise than those of aerial sample counts (Barnes, 2001, 2002).
A new indirect sample counting technique was recently applied to the estimation of elephant population size in forests (Eggert et al., 2003). The technique relies on the extraction of genetic material (DNA) from as many dung-piles as possible within a given area, and the use of a DNA fingerprinting protocols to identify the number of unique genotypes (individuals) in the samples collected. The rates of repeat samples obtained can then be used to estimate the population size for the area using the equivalent of a markrecapture census technique (Eggert et al., 2003). This technique is likely to find wide application in sites where other methods are unlikely to give reliable results, for example, areas of mixed habitat and areas with low densities of elephants where other methods would give wide confidence limits.
Guesses are often the only kind of estimate obtainable for many areas. Organizing an elephant survey, whether a total or sample count, from the air or on the ground, requires a considerable investment in manpower, equipment, time and money, and this is often not feasible. It is then that guesses made by people who know the area in question have to be relied upon. If informants provide sufficient data to permit an evaluation of the accuracy of their estimate, such as a survey estimate with little or no details of methodology, then the estimate is considered an informed guess. If no such information accompanies the estimate, or if an estimate is more than 10 years old, then it is considered an ‘other guess’.
Every survey method has its own sources of error and bias, and the choice of method is often not a matter of selecting the best, but of avoiding the worst (Norton-Griffiths, 1978). In addition, pooling individual estimates to arrive at national, regional and continental estimates, presents problems of its own. It is inappropriate, for instance, to obtain a combined estimate for two areas by adding up the results of a sample count in one area to those of a guess for another. Even if similar methodologies were employed in the two areas, adding up the results would be of questionable value if the surveys were conducted at two very distant points in time. For these reasons, it is essential to categorize the information available and to present it in a manner that truly reflects the different degrees of inherent reliability.
Elephant population data in the AED are categorized according to data quality and survey reliability, as described below.
Survey data quality
Data quality is based on survey parameters that may affect the accuracy and precision of the estimate. It gives an indication of the thoroughness with which a survey is conducted, and thus gives a basis with which to compare the quality of surveys of a given area that employ the same methodology. A score of 1 to 3 (best to worst) is given to each survey estimate, as described below.
Ground survey data
ground total counts (gt), including individual registration (ir) studies, are given a data quality score of 1. It is worth noting, however, that the categorization of IR studies is currently under review, as many such studies do not always cover entire populations, and hence do not necessarily merit a high quality rating.
ground sample counts (gs) are rated according to sampling intensity, or sample fraction, which is defined as the proportion of the input zone covered in the survey. The following quality scores are based on percentage sampling intensity of a given area:
Greater than 20%
5% to 20%
Less than 5% or not reported
Aerial survey data
aerial total counts (at) are categorized in terms of search rate, or the area covered per hour, as follows:
Less than 100 km2/hr
100 – 200 km2/hr
More than 200 km2/hr or not reported
aerial sample counts (as) are categorized using sampling intensity. In stratified aerial sample counts, effective sample intensity, defined as the proportion of animals actually seen to the population estimate, is used instead. In both cases, the following quality categories are based on percentage sampling intensity:
Greater than 20%
5% to 20%
Less than 5% or not reported
Dung counts (dc)
Percentage Relative Precision (PRP)1 for mean elephant density less than 30% and one of the following:
Dung decay rate measured on site for 50 dung-piles or more
Defecation rate measured on site
PRP for dung density estimate ≤ 20%
Sampling done in both dry and wet seasons
OR any three of the above four conditions in (a)–(d).
PRP for elephant density of 50% or less
OR any two of the following three conditions:
Decay rate measured on site for 30 dung-piles or more
Defecation rate measured on site
PRP for dung density ≤ 30%
When the conditions for (1) and (2) are not met.
Genetic dung counts (gd)
Effective sampling intensity, defined as the number of unique genotypes identified expressed as a percentage of the estimate, is used as the measure of quality for genetic dung counts, as follows:
Greater than 40%
20% to 40%
Less than 20% or not reported.
Both informed guesses (ig) and other guesses (og) are given a data quality rating of 3.
Population estimate data entered into the AED vary in quality from the identification of individual animals to plain guesswork. The addition of population estimates of varying quality into national, regional and continental totals is, from a statistical viewpoint, strictly invalid and produces misleading results. On the other hand, discarding low-quality estimates would produce equally misleading estimates, as high-quality survey estimates are not available for most areas in which elephants are found.
In order to solve this problem, the AED incorporates a system to accommodate all types of estimates by categorizing them according to their type and allocating them into non-overlapping categories. Thus, while it is still impossible to produce a single continental estimate, it is at least possible to obtain totals for a number of categories of differing degrees of reliability.
Four categories are used, each associated with a different level of uncertainty. The categories are definite, probable, possible and speculative. In order to place estimates into the appropriate categories, population estimates are classified according to survey type along a scale of survey reliability ranging from a (highest) to e (lowest). Each data reliability category contributes to the four categories as detailed in Table 2.
In addition to determining the breakdown of population estimates into definite, probable, possible and speculative numbers of elephants, survey reliability gives an indication of the level of certainty that can be placed on a given estimate, as determined by the type of method employed. Survey reliability gives a basis with which to compare surveys of a given input zone that employ different methodologies.
The categorization system described above is implemented in the AED through a series of algorithms. When executed, these algorithms categorize each population estimate in terms of data quality and survey reliability. The categorized records are then used to produce national, regional and continental totals.
In order to produce national, regional and continental totals, the variances of sample counts are added together in order to produce a 95% confidence interval for the sum of the estimates (Norton-Griffiths, 1978) before allocation of the pooled estimates to the four groups, definite, probable, possible and speculative. This is the reason why the regional totals in the definite, probable and possible groups are not always the sum of the corresponding national group subtotals. Likewise, the continental total numbers of elephants in these three groups do not match the simple sum of the regional subtotals.
Table 2. Categorization of elephant population estimates according to survey type and contribution of each to the four categories of elephant numbers.
a. Rounded to the nearest whole number if necessary.
b. If the lower confidence limit of the estimate is a negative figure, the estimate will be zero or, if reported, the actual number of elephants seen in the survey.
c. For dung counts it is assumed that there are no elephants unless any are observed directly (which is seldom the case). This is because, unlike with aerial surveys, where the estimate is almost invariably lower than the true population size, dung counts may underestimate or overestimate the population size, depending on the choice of parameters used (such as forest area, decay rate, or the mathematical model used). For genetic dung counts (gd) the number of distinct genotypes identified is regarded as the number of elephants actually “seen”.
|Survey Reliability||Survey type(s)||Categorization of estimates|
|a||• individual registrations (ir)
• aerial total counts (at)• ground total counts (gt)
|definite = the population estimate.
probable = none.
possible = none.speculative = none.
|b||• aerial sample counts (as)or ground sample counts (gs) with 95% confidence limits
• dung counts (dc) with 95% confidence limits and an estimate of dung decay rate obtained on site
|definite = the lower 95% confidence limit of the population estimate (there are at least this number of elephants) or the number actually seen, whichever is greater.
probable = the differencea between the estimate and the lower confidence limit, or between the estimate and the actual number seen or between the estimate and zero, if the lower confidence limit is negativeb.
possible = the difference between the upper confidence limit and the estimate.speculative = none.
|c||• dung counts (dc) with 95% confidence limits but no on-site measurement of dung decay rate
• genetic dung counts (gd)
|definite = none, or the number actually seen, if givenc.
probable = the population estimate.
possible = the difference between the upper confidence limit and the estimate.
speculative = none.
|d||• aerial sample counts (as), ground sample counts (gs) and dung counts (dc) without 95% confidence limits
• informed guesses (ig)
|definite = the number actually seen, if given.
probable = none.
possible = the population estimate or the lower estimate if a range is given, minus the actual number seen, if given.
speculative = the difference between upper and lower estimates, if given.
|e||• other guesses (og)
• Any of the above survey types in which the estimate is over 10 years old
|definite = the number actually seen, if given.
probable = none.
possible = none.
speculative = the estimate, or the mean of the upper and lower limit, minus the actual number seen, if given.
At all levels of addition (national, regional and continental), estimates in the definite (Df), probable (Pr), possible (Ps) and speculative (Sp) totals are non-overlapping. In other words, a possible estimate does not include definite or probable estimates. Thus, for a country, a region or the entire continent there are, simply speaking, “definitely” Df elephants, “probably” Df + Pr elephants, “possibly” Df + Pr + Ps elephants and “speculatively” Df + Pr + Ps + Sp elephants.
It is important to note that the totals presented for each country and region are minimum estimates, based on the estimates for the areas that have been surveyed or for which guesses are available in that country or region. In many countries, and in all regions, there are large areas of elephant range where elephant numbers have not been estimated. No extrapolations have been performed for these areas in the AED, and they are therefore not included in the totals. If all of the elephant range is listed, then the totals are national estimates. If, on the other hand, estimates are only given for a fraction of the elephant range in the country, the total cannot be considered a total national estimate. For this reason, the estimates given for the regions and for the continent cannot be interpreted as complete regional and continental estimates respectively.
One of the questions that most interests decision makers involved in African elephant conservation and management is whether elephant populations increase or decrease over time at the continental level. Many authors have in the past incorrectly compared estimates from different AESRs to derive continental population trends (e.g. Government of Kenya & Government of India, 2002). This is invalid and misleading for a number of reasons, as described in Blanc et al. (2005), from which the text below is adapted.
Many of the continent's elephant populations have never been systematically surveyed. Most elephant surveys tend to concentrate in and around protected areas, although nearly 70% of elephant range may lie outside these (see this report). Any changes reported in the AESRs are only derived from a subset of all elephant populations, and may therefore not reflect overall changes in numbers. The extent of unsurveyed range across the continent amounts to nearly half of the total elephant range in Africa (this report), but even this estimate is subject to considerable uncertainty. As noted above, elephant distribution data for the AED are obtained from questionnaire replies and other potentially unreliable sources, which can quickly become outdated, and knowledge on the actual extent of elephant range remains unreliable.
Many important populations are not surveyed frequently and several have only been surveyed once. In consequence, any one AESR repeats a number of estimates from the previous report because these are still the most up-to-date available. This makes using total numbers invalid as a measure of change, as constancy of numbers at some sites reflects only the same information carried forward from one status report to the next. The totals in the definite and probable categories may decline where an out-of-date estimate has been degraded to the speculative category and no more recent information is available. Conversely, where a population is surveyed for the first time, the resultant increase in the total is due not to population increase, but to the inclusion of new information. False increases (or decreases) may also happen when the boundary of the study area changes between surveys, although the site name remains the same. When only parts of the ranges of elephant populations are included in the surveyed area, changes in estimates may be caused by elephant movements rather than real changes in population size.
Even where two successive surveys of the same area are available, misleading changes may be observed when different methods, liable to different levels of accuracy or bias, are used in the two consecutive surveys. Variation in survey conditions – like the time of the year or even the use of different survey crews – may result in changes in numbers of elephants seen, thus contributing to differences recorded over time. In addition, many estimates come from sample surveys, and are therefore subject to statistical sampling error. As a result, differences between successive estimates could be due purely to chance, but can still make a large contribution to the differences between totals.
In order to disentangle these confounding factors from real changes in elephant numbers at the national, regional and continental levels, a new system to track changes in estimates and their ostensible reasons has been devised and implemented in the AED. The system relies on linking estimates contained in the current version of the AED to the corresponding estimates contained in the version of the AED used to produce the previous report, and assigning a “cause of change” to each pair of estimates, as described in Table 3. Changes in estimates are then grouped by cause of change, and overall differences in the definite, probable, possible and speculative categories are calculated at the national, regional and continental levels.
While the repeat survey group (see Table 3) contains all those sites where surveys have been repeated using comparable methods, not all pairs of estimates in that group are necessarily statistically comparable. Other factors described above but not captured by the tracking system, such as a different season or different survey crew, could still render comparisons meaningless. Where it is suspected that such factors may be responsible for the difference in the estimates, the survey pair is deemed not to be comparable and is marked rs'.
Where the more recent surveys in methodologically comparable survey pairs account for a large proportion of the definite plus probable estimate for a given region, a statistical comparison of elephant numbers over time, albeit restricted to a segment of the population, can be performed (see Blanc et al., 2005 for details). A list of methodologically comparable surveys featured in this and the previous report is provided in Appendix II.
The status of African elephants varies considerably across their range, with elephants occurring in large, dense populations in some parts of the continent but only surviving in small, fragmented populations in others. In a very similar way, the quality and extent of knowledge on elephant status varies widely across the continent. Some populations have never been surveyed, or are only surveyed rarely, while others are counted annually. The objectives of the AED include promoting the use of standardized, reliable survey techniques throughout elephant range, as well as facilitating the task of donors, wildlife authorities and decision-makers in prioritizing their efforts to monitor elephant populations.
In order to assist in meeting these objectives, a simple index has been developed to measure the quality of elephant population data available at the national, regional and continental levels. Based on this index, a system to identify and prioritize the areas, countries and regions where systematic surveys are most needed has also been developed and implemented in the AED. These new measures, both of which are calculated from data contained in the AED, are described in turn below.
Information Quality Index (IQI)
If all elephant populations on the continent were systematically surveyed, and unbiased estimates with measured precision produced, the sum of the definite and probable categories in the AED would be an accurate statement of true elephant numbers. Even with incomplete data, the sum of these two categories provides the “best estimate” of elephant numbers from systematic surveys (i.e. surveys in data reliability categories a–c).
Table 3. Codes and descriptions of causes of change as implemented in the AED.
|Code||Cause of Change||Description|
|rs||repeat survey||Both surveys were conducted using comparable methodologies.|
|da||different area||Both surveys were conducted using the same methodology, but the extent of the areas covered differ by 10% or more.|
|dt||different technique||The most recent survey uses a different survey methodology, or replaces a guess.|
|np||new population||A new entry into the AED, i.e. no previous survey or guess to compare with.|
|pl||population lost||The population is known to have disappeared from the site, be it through translocation or local extinction.|
|ng||new guess||A guess replaces an older guess or a survey estimate that has been downgraded to the category of other guesses for being more than 10 years old.|
|na||new analysis||Data from previous report has been re-analyzed or re-interpreted in the light of new information other than a new estimate.|
|dd||data degraded||The estimate in the previous report has been downgraded to the category of other guesses for being more than 10 years old.|
|—||no change||The estimate has been retained unchanged from previous report.|
A good indication of the overall quality of available survey data is given by the ratio of good-quality population data to total population data (i.e. the sum of the definite, probable, possible and speculative categories). This ratio, or ‘probable fraction’ (PF), is therefore defined as
In order to make it a more meaningful indicator of the quality of information, however, the PF needs to be combined with a measure of the completeness of estimate coverage. Such a measure can be obtained simply from the proportion of total range for which population estimates (of any quality) are available. Thus, the Assessed Range Fraction (ARF), is calculated as
The product of the above two factors gives an unbiased, normalized and scaleable index of the overall quality of information on elephant population estimates. Thus, the IQI is defined as
The IQI ranges from zero (no reliable information) to one (perfect information). Thus a country or region where few reliable surveys have been conducted, and which cover only a small portion of its total range, will have a score closer to zero. A country or region where high-quality data are available for most of its elephant range, on the other hand, will have a score close to one. Note that at the infra-national (i.e. site) level, the ARF is constant, and hence the IQI is simply equal to the PF.
The IQI can further be used to derive an unbiased system for setting priorities as to the areas that are in most need of systematic surveys. For a truly accurate continental picture of elephant abundance to emerge, reliable estimates would have to be available for all elephant range. Thus, countries that account for a large proportion of total continental range should be prioritized more highly. It is therefore important to include in the prioritization system the Continental Range Fraction (CRF) accounted for by each country:
To yield a score of Priority for Future Surveys (PFS), the IQI and CRF are combined as follows:
When calculated for each country or region, the result of the above equation, rounded to the nearest integer, gives a convenient measure, ranging from 1 to 5, of the countries/regions where population surveys are most needed. At the infra-national (site) level, the CRF is replaced by the proportion of national range accounted for by the site in question.
Note that the logarithmic nature of the priority scores means that the difference between two successive priority scores is of an order of magnitude. All areas of elephant range that have never been surveyed, i.e. those for which estimates are currently unavailable, are automatically assigned a priority of 1. Systematic surveys should be conducted in areas of unsurveyed known range. In areas of doubtful range and unsurveyed possible range, elephant presence/absence should be established prior to conducting systematic population surveys.
It is important to stress that neither the IQI nor the PFS are measures of the health of elephant populations, or of overall elephant conservation priorities, but rather of the quality of elephant population data and of the need to conduct systematic surveys in future. For instance, range loss in a country will often result in a decline in the proportion of unassessed range, thus causing the IQI to increase and the priority ranking to decline. A list of all African elephant Range States with their IQI and PFS scores is shown in Appendix I.
While it is hoped that the PFS system will prove useful for prioritizing elephant population monitoring efforts, the system is not intended to be prescriptive. Individual Range States may have good reasons to use different criteria and different systems for prioritizing elephant population surveys.
Information in this report is presented at the continental, regional and national levels. The continental section is followed by regional sections, each of which contains the relevant individual country sections, in alphabetical order. Each section follows the format described below.
Each section begins with a brief overview intended to supplement the information provided by the maps and tables that follow. The overviews are not intended to provide the reader with exhaustive information on each country, but simply to describe the current situation and to highlight any factors that may have contributed to it. This report no longer contains Historical Background sections featured in the previous report; readers interested in the history of elephant populations can consult the AESR 2002 (Blanc et al., 2003).
The overview contains the following sub-sections:
General Statistics. This section provides summary statistics of country area, protected area coverage, area of elephant range, amount of elephant range in protected areas, and the amount of range which has been surveyed or has elephant population estimates, IQI, CITES Appendix and year of CITES listing. Only protected areas that fall within the IUCN protected area management categories I through IV have been included for these calculations. While many important management areas for elephants belong in categories V (Protected Landscapes) and VI (Managed Resource Areas), their conservation importance and effective protection is far from uniform across the continent.
Current Issues. Any issues that may, directly or indirectly, affect elephant populations and their conservation and management. These may include poaching, political conflict, refugee crises, land use and wildlife management policies.
Range Data. Summary information on how elephant range was determined and categorized, as well as any changes made to the map since the last report.
Population Data. Description of the areas that have been surveyed and the methods employed, and how the data have been interpreted and categorized. Any changes between individual and pooled estimates are described and explained here.
Cross-border Movements. Information on movements of elephants across international borders. This section is ommited from the Continental Overview.
The text overviews are followed by three tables that summarize elephant population estimates, changes in estimates since the previous report, and the areas of range covered by each type of estimate. These tables are described in turn below.
Summary Totals Table
The summary totals tables present pooled estimates at the national, regional and continental levels, separated into four groups, definite, probable, possible and speculative numbers of elephants, based on the survey reliability categories (a–e) described in the Data Types and Categorization section. It is worth repeating that the totals presented for each country are not necessarily complete national estimates, and depend on the amount of range that is covered by estimates (see below). Totals from the previous report are also shown on the table.
Interpretation of Changes in Elephant Estimates since the Previous Report
This table shows the breakdown and net changes in the four categories of elephant estimates, grouped by the ostensible reason for change, as described in Table 3. Due to the method of pooling variances to calculate totals in the four categories, the calculated changes would not necessarily add up to the net changes between the estimates presented in this report and the AESR 2002. Thus, and in order to make the rows of the table add up to the net, each component figure is adjusted by dividing it by the net difference between the two reports and multiplying that by the total change calculated through pooling variances. In a few cases, however, the discrepancy between the net and calculated changes is such that the sign of the components is reversed. In such cases, the change is proportional to the magnitude of change, but not to its direction.
Area of Range Covered by Each Data Category
These tables depict the contribution of each survey type to the total area (in km2) for which estimates are available. In addition, areas of unassessed known and possible range are also shown on the tables. Pooled estimates of elephant numbers for countries or regions where large areas of range remain unassessed are likely to be underestimates. Large areas of unassessed possible range, however, could simply reflect inadequate information on current elephant distribution.
A map is shown for each country, region and the entire continent, showing elephant distribution, input zones, protected areas, national and/or regional boundaries, major towns, rivers and lakes. Neighbouring countries and regions are shown to highlight important cross-border populations, as well as the spatial relationships between elephant populations in different countries. A thumbnail locator map is shown at the bottom of each map to easily identify the location of the country in the continental context.
Input zones are shown with a grey hatched pattern. The reliability of the associated population estimate is reflected by the spacing of the hatching, with reliable surveys being depicted with a more closely spaced hatching than guesses.
Elephant range is displayed according to the elephant range categories described in the Data Types and Categorization section above. Known range is shown in dark green, possible range in light green, and doubtful range is displayed in a dotted pattern, while non-range is transparent (white). Point sightings are shown as crosses on the map.
Protected area boundaries are shown in khaki, and are individually labelled on national maps. The official designations of protected areas are abbreviated on the labels; a list of these abbreviations can be found in Appendix IV. An alphabetical list of protected areas within elephant range, including details on surface area, year of establishment, IUCN Category, and the country in which they are found is shown in Appendix III. Note that the IUCN Category of any given protected area is not necessarily an indication of the effectiveness of protection.
Each country map is preceded by a national table of estimates showing an alphabetical listing of input zones. By providing the location of the centroid of each input zone in decimal geographic coordinates, the table also serves as a key to the input zones shown on the map. In addition, national tables of estimates present details on estimates, their quality and other metadata, as described in Table 4. The columns shown in the regional and continental tables are different from those shown in the country tables, and they are described in Table 5.
Table 4. Details and survey parameters provided in national tables of estimates in this report
|input zone||Name of the input zone followed by its legal designation (if any), e.g. Kruger National Park.|
|cause of change||Attributed reason for the change in the estimate with respect to the previous report. As described in the Data Types and Categorization section, causes of change are coded da (different area), dd (data degraded), dt (different technique), na (new analysis), ng (new guess), np (new population), pl (population lost) and rs (repeat survey). Where an estimate has been retained from the previous report, a dash (—) is shown to indicate that the estimate has not changed. Where a new systematic survey has been conducted (i.e. rs, da, dt and np), the cause of change code is shown in bold type.|
|survey type||Type of survey conducted and its assigned quality score (1, 2 or 3), as described under the Data Types and Categorization section.|
|survey reliability (reliab.)||Category (a, b, c, d or e) into which the elephant population estimate falls. Survey reliability is dependent on survey type and additional criteria, as described in the Data Types and Categorization section.|
|survey year||Year in which the survey was conducted, or in case of guesswork, the year to which the guess applies.|
|number of elephants||Elephant population estimate from the survey or guess reported.|
|95% c.l.||The 95% confidence limit for the estimate or, in the case of informed guesses, the upper range of the guess marked with an asterisk. This cell is blank for surveys in which there are no confidence limits (e.g. total counts and unreliable dung counts), as well as for other guesses.|
|source||Author(s) and year of the report, questionnaire reply, personal communication or published source from which the estimate was obtained. All sources appear in the list of references at the back of this report.|
|priority for future surveys (pfs)||Based on the precision of estimates and the proportion of national range accounted for by the input zone, the pfs is a measure of the importance and urgency of the need for future systematic surveys. Priorities range from 1 (highest) to 5 (lowest). All areas of unassessed range (i.e. un-hatched areas on the map) are considered to be of the highest priority (1). For full details on the how the PFS is calculated, please refer to the Data Types and Categorization section.|
|area||Size of the input zone in square kilometres (km2). Where available, the area given is as reported by the reference source. If unreported, the area is either derived from the size of the protected area(s) to which the estimate refers, or is calculated using the GIS in the Lambert Azimuthal Equal area projection.|
|map location||Longitude (lon) and latitude (lat) of the centroid of the input zone, given in decimal degrees with one decimal.|
Table 5. Details of the columns shown in the regional and continental tables of estimates of this report.
|country / region||Name of the country or region.|
|elephant numbers||Elephant numbers in the four categories (definite, probable, possible and speculative).|
|range area (km2)||Estimated total elephant range area (known + possible range) in the country or region.|
|% of regional range||Percentage of the regional/continental range accounted for by the country/region in question. Rounded to the nearest integer.|
|% of range assessed||Percentage of elephant range in the country or region for which elephant estimates are available.|
|information quality index (iqi)||In the regional tables, the IQI is shown for each country, and for the entire region in the totals row. In the continental table, the IQI is shown for each region and for the continent in the totals row. Please refer to the Data Types and Categorization section for details on how the IQI is calculated. A complete list of all Range States with their IQI scores can be found in Appendix I.|
|priority for future surveys (pfs)||In the regional tables, the PFS is shown for each country, and for the entire region in the totals row. In the continental table, the PFS is shown for each region. No priority is shown for the continent as a whole. Please refer to the Data Types and Categorization section for details on how the PFS score is calculated. A complete list of all Range States with their PFS scores can be found in Appendix I.|
1. PRP is the 95% confidence limit expressed as a percentage of the estimate. Thus an estimate of 30 with confidence limits of ±15 has a PRP of 50%
< previous section < index > next section >