Introduction
Methodology: survey procedures and data collection
Results: resources inventory and land stratification
Discussion
References.
Les enquêtes à basse altitude dans la recherche sur les systèmes pastoraux
Summary of discussion session 2.
Résumé des débats de la deuxième séance
Kevin Milligan¹ and Peter de Leeuw²
¹Air Survey Coordinator, ILCA, Nigeria
²Ecologist, Arid Zones (Eastern and Southern Africa) Programme, ILCA Kenya
The recent development of the science of low-altitude aerial survey has provided a rapid, reliable and cost effective method for assessing the numbers and distribution of livestock and people over large tracts of land.
In the past, descriptions of pastoral systems had to rely upon indirect methods, such as vaccination counts or tax returns, for obtaining information about livestock numbers. Such methods have well-publicised errors and biases. However, it is rarely pointed out that these indirect methods are usually based upon counts taken only once a year and thus seasonal distributions are unknown. Also, because the animals are generally congregated into camps for these counts, their actual distribution in relation to resources during their daily grazing orbit is not known.
The methods of low altitude aerial survey were initially developed by ecologists in East Africa wishing to map the distribution of wild animals in national parks. Specific strategies suitable for assessing livestock and people are still being perfected and ILCA has taken a leading role in exploring the possibilities. An area of particular interest has been the inclusion of recordings of rangeland resources. This information, especially on vegetation and water availability, can be compared to the observed livestock distribution and thus an understanding of resource utilisation can be included in the descriptive phase of pastoral systems research. Such comparisons often allow identification of specific constraints to livestock distribution and thus a diagnosis of possible interventions to improve the system.
This paper describes the basic methods now being used for low altitude aerial survey and illustrates the types of results that can be obtained and how these results are being presented. The possibilities of linkage and integration between such aerial surveys and other methods of remote sensing are covered in another paper at this workshop (de Leeuw and Milligan, 1983), and thus this paper will discuss how low level aerial survey can be used to diagnose constraints within pastoral systems and the relative cost effectiveness of such operations.
Details of ILCA's methodology are given in previous papers (Grimsdell et al, 1979; 1979a; Okali and Milligan, 1980) and reports (Milligan et al, 1979; 1982; Milligan, 1980; 1981, 1983). What makes the general method different from other remote sensing techniques is that the aircraft is flown at a very low altitude, so that a team of observers on board can make direct visual counts of animals, people and range resources that pass by the aircraft.
Flight and sample procedures
ILCA usually carries out its aerial surveys using a systematic, unstratified flight pattern (see Fig. 1), so that every part of the study area is covered evenly and uniformly and results can immediately be presented as distribution maps. It also provides a data base for post-survey stratification.
Spacing between the systematic parallel flight lines is usually decided by a balance between time and cost constraints and the required levels of sampling and precision of results. Flight lines based upon the 10 km UTM grid projection provide a useful system for global reference. However, sometimes a flight pattern that exactly corresponds to existing maps will be more useful when ground truth work is carried out or planned in the study area.
Each flight line is divided into intervals down its length. These intervals, together with the parallel flight lines, allow the entire study area to be divided into a grid pattern of fixed dimensions. All information collected can thus be related to the individudal grids and the resulting maps show simple and clear distribution patterns.
Fig 1. Flight pattern, grid system and sampling.
The area actually sampled is restricted to a fixed band on either side of the aircraft (Fig. 2). This band is determined by the projection, from the observer's eye to the ground, of two parallel rods attached to each wing strut. The choice of a suitable flying altitude is important. The higher the altitude, the greater the sample area observed and thus the greater the expected precision of results; however, animals and people become increasingly difficult to see and count at higher altitudes. The selected altitude is usually between 400 and 1 000 feet above ground level. Depending upon survey design, overall sample cover is usually between 5 to 20%.
The strip sampled (w) on either side of the aircraft is determined by the projection, from the observers' eye to the ground, of the two rods (a and b) fixed on the wing struts. Strip width can be varied by altering these rods and is directly proportional to the aircraft height above the ground (h). Typical settings for cattle surveys are h - 1000 feet, w = 500 metres.
Fig 2. Schematic representation of aircraft during observation flight.
Information collection and analysis
In the aircraft, there are two back seat observers who count and photograph the numbers of animals and dwellings seen in the sample band in each grid. Individual observer bias, which varies according to certain factors such as the effects of vegetation structure and time of day, requires careful calibration and correction. A front seat observer, seated beside the pilot, is responsible for navigation and recording ecological conditions. Depending upon survey requirements, ecological zones and seasons, these would normally include:
- vegetation physiognomy;
- tree density per hectare, dead trees and indicator tree species;
- grass cover, height and greenness and extent of burning;
- water sources and extent;
- farm areas, extent and major canopy crops.
Ecological recordings are supplemented by hand-held side-looking photography and occasional vertical photography.
The primary objective of an animal and human survey is usually to answer the fundamental questions:
|
How many ? |
- this can be worked out from simple statistics upon the recordings down each individual sample flight line; |
|
Where ? |
- this can be readily seen from the grid distribution maps; |
|
Why ? |
- relationships between animals and people and their environment can be tested by a series of multivariate analyses. Although simple correlations can be confounded by partial relation ships, step-wise regressions or more preferably analyses of variance or factor analyses can display all the variables in terms of their individual relationships to each other. |
The opportunities for such analyses can be envisaged from Fig. 3, which represents-some data sets that could be collected from a typical survey. Assuming that one variable such as cattle herds, is of particular interest, this can be separated from the rest of the data file. The general distribution of cattle herds can be seen by examining the grid entries on this data element. The relations of cattle herds to any of the data files can be examined statistically and, in the case of fig. 3, cultivation, grass cover and distance to water are all likely to influence cattle herd distribution. The final analysis, which is particularly important to the identification of actual constraints to cattle herd distribution, is to give a priority rating to each of the data files in terms of their sequental ability to explain variation in the cattle herd distribution.
The total number of data files for analysis is not restrict to information collected during flight, but can include other information from existing maps. In this way, the importance of parameters such as rainfall, geology or administrative boundaries could also be assessed.
BY CODING ALL ORIGINAL INFORMATION AT A COMMON FORMAT AND SCALE, ANALYSIS (SUPERIMPOSING DATA, IDENTIFYING PROXIMITY OF RESOURCES AND MATHEMATICAL WEIGHTING) ARE POSSIBLE WITH RESULTS IN A 'SPATIAL CONTEXT'
Fig 3. Typical data elements used in the computer data handling technique.
To illustrate some of the results from low-altitude aerial survey and the way these results are being analysed and presented, examples are taken from ILCA's work in Nigeria, Mali, Niger and Ethiopia.
Animal and human populations: abundance and distribution
An initial analysis of data provides overall population totals for the study area, including herd or camp sizes, and densities. Table 1 shows results from the entire 81,555 km USAID/NRL project area in Niger, during two seasons.
Table 1. Seasonal livestock populations (+% SE) in the NRL project area, Niger.
|
Animal type |
Herds |
Herds |
||||||
|
Total (Nos) |
|
Stocking (Ha/Hd) |
Total (Nos) |
Mean Size |
||||
|
Dry |
Wet |
Dry |
Wet |
Dry |
Wet |
Dry |
Wet |
|
|
Cattle |
288,653(12) |
376,533(16) |
28 |
22 |
8,706(10) |
10,043(12) |
33 |
38 |
|
Bororo |
152,814(15) |
246,900(17) |
54 |
33 |
4,025(13) |
5,444(16) |
38 |
45 |
|
Azawak |
135,839(12) |
129,632(17) |
60 |
63 |
4,681(10) |
4,599(10) |
29 |
28 |
|
Sheep/goats |
780,289(8) |
1,147,914(11) |
10 |
7 |
16,964(8) |
16,513(10) |
46 |
70 |
|
Camels |
70,162(9) |
155,708(32) |
116 |
52 |
12,917(9) |
7,829(13) |
5 |
20 |
|
Donkeys |
13,540(14) |
23,375(12) |
588 |
345 |
1,890(14) |
4,086(9) |
7 |
11 |
Such population and herd sized totals give information about likely production levels and economy in the project area. A comparison between the seasons suggests changing herding strategies. For example, the population of the Bororo cattle was 62% greater during the wet season than during the dry season, while the Azawak cattle breed numbers remained stable between seasons; this suggests that the Bororo cattle are more mobile and may be part of a larger-scale transhumant pattern. Detailed distribution analysis detected: areas of seasonal herd splitting; areas of total herd immigration; areas of cattle influx linked to immigration of primarily sheep and goat owners; areas of influx associated with livestock ownership by settled cultivators; and a substantial wet-season immigration of camels.
Similar tables for the total pastoral population of the NRL project area identified five distinct pastoral groups: three Tuareg; one WoDaaBe Fulani; and one Arab. An immediately interesting result was that, while the initial project preparation document for the area estimated 50,000 people, the actual -number of pastoralists alone was about 175,000.
Although information about the total number of animals or people in an area is valuable because it indicates the size of the target group and thus the possible total regional benefits that might result from successful intervention, information about distribution is often more important, in that it indicates the regional impact of animals upon their resources, and guides follow-up research and development activities into areas of concentration For example, an [examination of the livestock and pastoralist distribution in this: NRL project area showed that the Bororo and Azawak cattle breeds had very different areas of concentration and these corresponded to the distribution of WoDaaBe Fulani and tented Tuareg respectively. These differences in distribution immediately indicated that detailed. information about either breed or tribe would only be relevant to certain parts of the zone, while the opportunities for research or development contact with. the two. tribal peoples would be different. The different transhumance patterns between the two groups suggest different flexibilities in management strategies which are likely to affect their long-term development paths and choice of interventions.
ILCA is at present using four different methods to represent the results of distribution data (Fig, 4). As the original data is collected on a grid basis, each grid may be shaded to represent the density it supports (choropleta symbolism). Alternatively, the individual grid information can be represented as squares or circles (proportional point symbolism), whose sizes correspond to the observed densities. Such representation of actual recordings, on a grid by grid basis, can often be more easily interpreted if variations between adjacent grids are mapped as contours which can also be represented as a three surface view.
Dry-season grass distribution in a central region of the pastoral zone of Niger - May 1981
Figures shown are percentage grass cover
Fig 4. Methods of representing aerial survey distribution data. - Choropleth symbolism
Fig 4. Methods of representing aerial survey distribution data. - Porportional point symbolism
Fig 4. Methods of representing aerial survey distribution data.- 2 - dimensional surface
Fig 4. Methods of representing aerial survey distribution data.-3 - dimensional surface
A common feature of pastoral systems is the changes in seasonal distribution patterns. of livestock and people. Although it is almost impossible, by aerial survey, to link these movements to individual specific herds (without radio collaring or herd marking) the patterns of herd concentrations can be easily understood from repeated seasonal surveys. If these distribution patterns are investigated on the ground, by social interview, and movement histories are established, regional mobility strategies can be mapped and quantified.
As stated in the introduction, most of the ground census data relates to one period of the year. for example, ground census figures for the Jos Plateau of Nigeria, based upon vaccination and tax returns, suggested a population of about 300,000 cattle, while aerial surveys (Milligan, 1980) showed a dry-season population of 140,000 and a wet-season population of 400,000. Often ground studies have provided general indications of movements; however, the value: of an actual census can be illustrated from work by ILCA in the Nigerian subhumid zone which was generally considered to be an area of dry-season grazing, with animals returning north early in the wet season when surface water becomes again available and the southern tsetse belts expand. Table 2 shows that two of ILCA's four ease-study areas had high wet-season concentrations: an exactly converse pattern to that expected.
Table 2. Cattle densities in four case-study areas in the Nigerian subhumid zone.
| Area (km²) | Density (no. km²) | Mean size | |||
| dry | wet | dry | wet | ||
|
Kurmin Biri |
2,500 |
17.3 |
4.2 |
51 |
54 |
|
Abet |
2,475 |
37.4 |
22.7 |
48 |
61 |
|
Mariga |
2,750 |
6.6 |
23.5 |
25 |
53 |
|
Lafia |
3,500 |
12.7 |
37.7 |
59 |
86 |
The Niger delta region in Mali is another area of marked seasonal movements of livestock. This seasonally flooded region within the Sahel supports, during the period of greatest animal concentration, about 1,200,000 cattle and an additional 500,000 sheep and goats. Maps of the seasonal distributions of cattle are illustrated in Fig. 5.
Fig 5. Seasonal cattle distribution in the Niger delta region of Mali. - October 1980
Fig 5. Seasonal cattle distribution in the Niger delta region of Mali. - March 1981
Fig 5. Seasonal cattle distribution in the Niger delta region of Mali. - June 1982
These maps show, during October, high cattle densities in the transition zone surrounding the delta and low populations where water is lacking and also in the central basin which is inaccessible due to flooding or because it is protected by river barriers. The March pattern showed a concentration of cattle in the floodplain which supported 72% of the total population. Two distinct sub-populations were detected; one south of the River Niger in the Djenne region; the other north, near Lake Debo. During June, cattle were leaving the delta, and distinct movements westwards and eastwards from Lake Debo were recorded.
Environmental conditions
All environmental variables recorded during a survey can individually be illustrated as distribution maps based upon the aerial survey grid pattern. Alternatively, average or total figures can be computed for particular regions.
An example of these resource maps for the Mali delta region shows that only about half of the survey areas can be termed 'upland Sahel with little or no flooding' (Fig. 6). Farming is of high density in the Mema Jura to the east of Mopti and in the Sanari plains. The rest of the area consists-mostly of flooded plains and backswamps (34%) with considerable land devoted to floating rice cultivation. The highest levels of rice cultivation were recorded in the Macina plains and those south of Mopti; these plains are broken by higher level terraces, levees and point-bar systems, which are much less'. inundated and show a mixture of rainfed and rice farming.
Figures shown are percentages
Fig 6. Levels of flooding, rice and cereal cultivation in the Niger delta region of Mali. - Level of flooding
Fig 6. Levels of flooding, rice and cereal cultivation in the Niger delta region of Mali. - Rice cultivation
Fig 6. Levels of flooding, rice and cereal cultivation in the Niger delta region of Mali. - Cereal cultivation
As for livestock, seasonal changes in environmental conditions can also be illustrated by a series of seasonal maps and graphs. An example of a graphic application for the drier areas surrounding the Mali delta is given in Fig. 7, for grass cover during March and June. Such frequency distributions of grass cover levels are often related to the levels of grazing: in general, curves with a rightward skew indicate understocking, while a leftward skew indicates overstocking, except where there are other causes for dry matter disappearance.
Fig 7. Frequency distribution of seasonal grass cover levels in the upland Sahel region bordering the Mali delta.
The actual environmental variables collected during a survey will depend upon the objectives of the survey and the capabilities of the survey team. Questions about specific environmental conditions can often be answered; such as possible changes in degradation which can be inferred from soil and erosion patterns, from grass cover, or from the distribution of dead or indicator plant species. In Niger, the distribution of a small shrub, Calotropis procera, was recorded as well as the distribution of dead trees (Fig. 8); the former was in response to USAID requests based upon observations that this species had increased substantially in their area during the past few years and the suggestion that it may indicate deteriorating conditions.
Fig 8. Distribution of Calotropis procera and dead trees in the NRL project area in Niger. - Dead trees
Fig 8. Distribution of Calotropis procera and dead trees in the NRL project area in Niger. -Calotropis procera
Interrelations between animals and environment
The interrelations between animals or people and the various measured environmental parameters can provide information that will directly suggest appropriate management interventions into the pastoral system. For example, if animals are usually concentrated into cultivated areas, opportunities for better interactions with farmers may exist; whereas, if cattle are usually away from cultivated areas and are concentrated in the natural rangelands and savannas, grazing reserves or improvement of range management practices may be feasible. Similarly, direct linkages to specific variables can indicate their relative importance and thus whether they are likely to be a constraint to the system. For example, if animals mostly graze near water sources; can be investigated.
While any set of two variables can be examined, for example grass cover to distance from water, it is often more valuable to study the system as a whole, considering the combined and cumulative effects of many variables upon another independent variable. The Objective may be to identify, for example, which out of all the variables recorded most closely corresponds to the distribution of cattle, on the assumption that such a variable may thus be a limiting or controlling factor to cattle distribution.
This particular stage in the analysis is perhaps the most valuable to pastoral systems research, as it moves away from the basic descriptive phase of the system into the diagnostic phase, hypothesising specific constraints that can be validated and tested by ground research teams. It is important to stress that results from such analysis are necessarily "predictive" rather than "causal". For example, results could indicate that animals are usually concentrated near water, i.e. that a knowledge of water distribution allows one to predict animal distribution. Water may, in fact, not be the cause of the observed animal distribution, which could be due to factors not even recorded during the survey, such as tribal or border conflicts, or that a particularly desirable plant occurred near these water sources. Clearly, a ground research team, equipped with the predictive hypotheses generated from an aerial survey, can quickly clarify their validity.
In the USAID/NRL study area, the overall cattle population distribution during May could be best predicted from a knowledge of cultivated areas, grass cover and distance to various water sources. During October, cultivation and grass cover continued to be most important but predictions could also be made from the general vegetation structure and tree density maps. However, the two cattle breeds were conspicuously different. Bororo distribution could nearly always be predicted from grass cover, further refined by vegetation and tree density distribution. Azawak were most concentrated in cultivated areas and near water sources. Thus there was a clear division between the two breeds in terms of their relationship to natural vegetation versus man-induced conditions. Such observations indicated that the basic options between rangeland development or water development may have a different relative value for the two breeds and the pastoral group associated with them.
In the southern Ethiopian rangelands, 19 variables, including cultivation, grass cover, vegetation physiognomy, growing period gradients and distance to three water source types, were included in an analysis of livestock distribution. The distance to ponds was ', nearly always the best predictor of livestock distribution. From such a knowledge, the relationship between cattle and ponds was more closely examined. Figs 9 and 10 show two ways that this information was presented.
Fig 9. Cattle distribution in relation to water sources in the SORDU project area of southern Ethiopia.
Fig 10. Number of cattle and the distance to ponds in the SORDU project area, southern Ethiopia.
Land stratification
Over large tracts of land, ecological conditions usually vary considerably and consequently so do opportunities for intervention. Thus an effective diagnosis of the principal constraints within a pastoral system will often require an initial stratification of land into major vegetation or ecological units; followed by an assessment of probable constraints within each strata; followed by an evaluation of whether particular strata are likely to respond to the proposed interventions.
In Niger, geomorphology and landscape were used to divide the NRL project area into six landtypes. Further subdivision into management units depended on the supposition that, at the present state of pastoral development, practical interventions require units that are geographically distinct, with suitable operational bases. The project area in Niger was finally divided into four major strata with 10 units. Specific research and development possibilities could be identified for each unit. For example, one of the units, the Ighazer plains, was essentially a clay soil floodplain while another was a sand dune grassland. While the development of surface water in pools and ponds was likely to persist in the former, ground seepage in the latter would necessitate construction of deep boreholes. However, controlled animal stocking was likely to be of little value in the floodplains where there was a low grass cover of scattered perennial tufts, compared to the dunes which had a good cover of annual grasses.
A similar approach to land stratification was adopted for the Mali delta-region, with breakdown into four strata subdivided into 19 units, for each of which aspects of the animal and human populations as well as the environmental conditions could be described (Table 3).
Table 3. Cattle and selected environmental conditions in four [and management strata in the Mali delta region.
|
Land management strata |
Area (km²) |
Cattle population |
Environment |
||||
|
Oct. |
March |
June |
Flood |
Rice |
Cereal |
||
|
('000) |
('000) |
('000) |
(%) |
(%) |
(%) |
||
|
Upland Sahel |
13,197 |
226 |
63 |
116 |
0 |
0 |
19 |
|
Transition zone |
5,811 |
237 |
91 |
102 |
10 |
7 |
14 |
|
Elevated plains |
5,818 |
230 |
189 |
181 |
21 |
12 |
9 |
|
Inundated plains |
11,117 |
114 |
871 |
410 |
72 |
24 |
3 |
The general iterative sequence of' pastoral systems. research can be seen to have four essential phases:
Figure
Low-altitude aerial survey is mainly valuable at the descriptive and diagnostic phases. The descriptive phase provides an inventory of the resources, together with an evaluation of their structure and functioning within the production system. This descriptive phase needs to be open ended, as comprehensive as possible and,. most importantly, should be both rapid and cost-effective.
Thereafter, the diagnostic phase must first identify the main constraints, devise ways of tackling them and then evaluate the probability of the success of any proposed interventions. The remote sensing nature of low level aerial surveys means that the constraints to the system can only be proposed or hypothesised, and ground validation is essential. Results of the work provide predictive statements, for example that animals tend to be found near wells, but the actual reasons and mechanism for this may be more complicated: animals may be near water for reasons other than water, while developing alternative water sources may be either technically or socially unfeasible.
Fig. 11 indicates how the essential flows from description towards diagnosis can operate. The survey immediately answers the resource questions: How many? Where? and When? while the interrelationships provide the hypotheses: Why? These hypotheses about possible causation and limiting factors to the system are posed within the multi-variate animal, human, environment context.
Fig 11. low altitude aerial survey.
Although low-altitude aerial- surveys record a large number of parameters relating to the animal and human populations as well as their environment, numerous parameters cannot be recorded. For example, while the animal population size and distribution are recorded, age and sex structures are not. Similarly, baseline information about human demography, household and labour budgets and decision making processes are not recorded, neither are environmental factors, such as geology and rainfall, though these often exist in published literature. While detailed range resource maps have been produced from intensive photo-interpretation and ground truthing, the general vegetation descriptions made during flight will often be sufficient to indicate whether range resources are an important constraint to animal distribution and thus whether further, more detailed, research is in fact necessary.
A particularly useful advantage of low aerial survey is its ability to guide aspects of the sampling design of ground investigations. (Okali and Milligan, 1980). If detailed animal production studies require a stratified sample design based upon herd size, this can be directed from the results 'of: the aerial surveys. Subsequently, aerial and ground data can be linked. together and detailed results from the ground can be extrapolated to the entire study area. Similarly, if household studies are to be based upon wealth strata and if the wealth characterization can be linked to factors visible from the air such as camp size, number of granneries, or cattle holdings? the basic sample frame and its locations can be determined before the ground work begins.
Outside its advantage as a multiple resource inventory technique, low altitude aerial surveys are both rapid and cheap .Although survey time will depend upon sample procedures and various logistic considerations,: an area of about 100,000 km could be covered in three ' to four weeks end preliminary reports from the survey could be made available to ground teams one month later. Costs are variable, The actual flying operations themselves are usually substantially less expensive than man-time costs:, and thus analysis and reporting time become important considerations. Normal costs, from planning to reporting, are about US$ 1.00 per km , although costs would be higher for particularly intensive investigations. This amount usually represents as little as 1% of the total research: end development costs within e: pastoral project.
De Leeuw, P. and Milligan, K. 1983 The integration of remote sensing techniques for resource evaluation in pastoral systems research. Invited paper, Workshop on Pastoral Systems Research in-Sub-Saharan Africa, ILCA, Addis Ababa.
Grimsdell, J.J.R., Bille, J.C. and Milligan, K. 1979:. Alternative methods for aerial livestock census. Invited paper, International Workshop on Aerial Survey Methods, Nairobi.
Grimsdell, J.J.R., Sihm, P. and Milligan, K. 1979a. The role of aerial surveys in livestock development projects. Invited paper, International Workshop on Aerial Survey Methods, Nairobi.
Milligan K. 1980. Abundance and distribution of cattle on the Jos Plateau, Nigeria. Report to the Federal Livestock Department, Lagos.
Milligan, K. 1981. Aerial survey of human, livestock and environmental | conditions in a central region of the pastoral zone of Niger. Report to USAID, Niamey.
Milligan, K. 1983. An aerial reconnaissance of livestock and human populations in relation to land use and ecological conditions in the SORDU project area of southern Ethiopia. Report to RDP, Addis Ababa.
Milligan, K., Bourne, D. and Chachu, R. 1979. Dry and wet season patterns of cattle and land use in four regions of Nigerian subhumid zone. Report to ILCA, Kaduna.
Milligan, K., Keita, M. and De Leeuw P. 1982. Recensement aerien saisonnier du cheptel et types de paysage du delta central du Niger au Mali. Report to ODEM/IBRD, Severe.
Okali, C. and Milligan, K. 1980. Ground/air linkage and the role of socio-economic survey in interdisciplinary research for intervention testing. Invited paper, Workshop on the Role of Anthropologists and other Social Scientists in inter-disciplinary Teams Developing Improved Food Production Technology, IRRI/UNEP, Manilla.
Résumé
La technique des enquêtes à basse altitude qui connaît depuis peu un certain essor s'est avérée une méthode rapide, fiable et peu coûteuse pour évaluer le nombre et la distribution des populations animales et humaines sur de vastes superficies. Le document décrit les méthodes de base utilisées actuellement pour effectuer les enquêtes à basse altitude, y compris les procédures de vol et d'énchantillonnage et la collecte et l'analyse de l'information
Des exemples tirés des travaux du CIPEA au Nigéria, au Mali, au Niger et en Ethiopie sont utilisés pour illustrer certains des résultats qui peuvent être obtenus avec les enquêtes à basse altitude et la manière dont ceux-ci peuvent être analysés et présentés. Ces exemples mettent l'accent sur l'importance et la distribution des populations animales et humaines, sur les conditions mésologiques, sur les interactions entre les animaux et l'environnement et enfin sur la stratification territoriale.
Les enquêtes à basse altitude sont surtout importantes dans les phases de la description et du diagnostic de la recherche sur les systèmes pastoraux. Le recours que font les enquêtes à basse altitude à la télédétection signifie que les contraintes aux systèmes identifiés par le biais de telles enquêtes ne constituent que des hypothèses et que la confirmation par l'observation au sol est indispensable,
L'un des avantages les plus importants des enquêtes à basse altitude réside dans le fait qu'elles permettent d'orienter certains aspects du plan d'échantillonnage des enquêtes au sol. Si les études détaillées sur la production animale font appel à un plan d'échantillonnage stratifié basé sur la taille des troupeaux, il est possible d'obtenir un tel plan grâce aux résultats produits par les enquêtes aériennes. Par la suite, il est possible d'établir la corrélation entre les données issues des enquêtes aériennes et des observations au sol; ainsi des informations détaillées recueillies au sol peuvent être extrapolées pour l'ensemble de la zone d'étude. De même, si les études sur les ménages doivent être basées sur la stratification de la richesse et si la détermination de la richesse peut être lice à des facteurs tels que la taille des campements, le nombre de greniers ou de bovins visibles à partir d'un appareil, le cadre d'échantillonnage de base et ses éléments peuvent être déterminés avant le début des travaux au sol.
Les enquêtes à basse altitude représentent une technique multiple; en outre, elles sont à la fois rapides et peu coûteuses. Une superficie d'environ 100 000 km² peut être couverte en trois à quatre semaines et les rapports préliminaires relatifs à l'enquête pourraient être mis à la disposition des équipes au sol un mois plus tard. Normalement, les coûts (de la phase de la planification à celle de l'élaboration du rapport) sont d'environ 1 dollar E.-U. par km², quoique pour les enquêtes particulièrement intensives, ce chiffre puisse être dépassé. Les sommes consacrées aux enquêtes à basse altitude représentent en général 1% seulement des coûts totaux de recherche - développement des projets pastoraux.
Chairman: Dr Cees de Haan (ILCA)
Discussion led by Dr Noumou Diakite (Mali)
Dr Rhissa suggested that finance be sought for aerial surveys of all the countries in the Sahel zone from international organisations such as OAU, World Bank, UNDP. Dr Abel warned that in using remote sensing there was one danger in particular that resulted from the detachment of the observer from the object of the survey. Objects that were visible on photographs and images may not be relevant to the local users of the land. Land classification from remote sensing may, because it ignored local use of particular resources, result in development plans which adversely affected local land users. There was also an unnecessary waste of existing information in the form of local knowledge of the land and its use which had been tried and tested over a long period. It was not, in Dr Abel's view, sufficient to use the social scientists of an inter-disciplinary team to counteract the inadequacies and dangers of remote sensing. He suggested that if local land classification systems were described and used from the outset, some of the dangers of inappropriate classification would be removed. If different ethnic groups practiced a variety of land uses on the same land, each of their classification systems should be incorporated into an integrated scheme, otherwise certain groups may be placed at a disadvantage in the competition for land which commonly accompanied planned changes in land use.
In commenting on the paper by Drs Milligan and de Leeuw, Prof. Saka Nuru said that the remote sensing technique would be a useful preliminary aid in the conceptual framework of LSR in terms of its time-saving and cost benefit effect before more serious work was done by scientists at a particular location. But he said that it was only useful when used as a complement to ground surveys which were probably more reliable in the identification of parameters of interest. There were certain constraints to the technique, such as the difficulty in identifying goats due to their small colour and size. Dr Milligan pointed out that sheep and goats were not normally a problem because of an aspect of their behaviour - they usually ran when an aircraft approached and thus could readily be distinguished from cattle. It was difficult to separate sheep from goats, and thus they were normally classified as 'shoals'. Prof. Saka Nuru asked what was the relative efficiency/sensitivity of the remote sensing technique vis-à-vis the ground survey. Dr Milligan replied that counting animals from the ground was time-consuming and there were numerous practical and theoretical problems. Dr Milligan agreed that ground truthing was important, depending on the level of information that was required.
Items estimated from the air could be checked and some biases corrected from ground survey calibration. Also a knowledge of the ground conditions helped one to make correct observations from the aircraft. Prof. Saka Nuru strongly supported the idea that ILCA should be involved in a country-wide use of the remote sensing technique in order to get a better picture of the distribution of natural resources as an aid to future planning for livestock development.
Dr Zulberti emphasised the need for a conceptual framework before any information was collected to avoid gathering a large amount of data, some of which might not then be used. Dr Chema also stressed the need for a specific purpose to be clearly defined at the beginning of the survey. Dr Milligan agreed, and stated that specific objectives were indeed identified before each flight. The hypotheses and questions that were stated at the start formed the basis for the kind of data that was collected and the design of the sampling strategies.
Dr Hiernaux asked Dr de Leeuw whether or not he thought that the use of satellite remote sensing methods and aerial surveys could result in increased costs due to the many field observations that had to be made to support such methods. Dr de Leeuw said that this highlighted the need for researchers to be more clear about what data they really needed from such surveys.
Président: M. Cees de Haan (CIPEA)
Débats dirigés par le Dr Noumou Diakité (Mali)
Le Dr Rhissa a suggéré de rechercher le financement d'enquêtes aériennes pour tous les pays de la zone sahélienne auprès d'organisations internationales telles que l'OUA, la Banque mondiale et le PNUD. Le Dr Abel a déclaré que l'utilisation de la télédétection comportait un aléa particulier qui résultait de l'éloignement de l'observateur de l'objet de l'enquête. Les objets qui étaient visibles sur les photos et sur les images peuvent ne rien signifier pour les utilisateurs locaux de la terre. La classification des ressources territoriales par télédétection pourrait se traduire par des plans de développement néfastes pour les utilisateurs locaux, notamment parce qu'elle ne tient pas compte des utilisations locales de ressources particulières. En outre, on ne tirait pas parti d'informations disponibles sous forme de connaissance locale de la terre et de ses utilisations et éprouvée par le temps. Aux yeux du Dr Abel, il n'était pas suffisant d'utiliser les spécialistes de sciences sociales d'une équipe interdisciplinaire pour pallier les insuffisances et les dangers de la télédétection. Il a déclaré que si les systèmes locaux de classification des terres étaient décrits et utilisés dés le départ, certains des risques de classification inadéquate pourraient être écartés. Si divers groupes ethniques ont adopté diverses formes d'utilisation des terres sur la même superficie, chacun de leurs systèmes de classification devrait être incorporé dans un schéma intégré, autrement, certains groupes pourraient être placés dans une situation désavantageuse dans la compétition pour les terres qui accompagne en général les changements planifiés de l'utilisation des terres.
Dans son commentaire sur le document de MM. Milligan et de Leeuw, le Prof. Saka Nuru a déclaré que la technique de la télédétection constituait un outil préliminaire dans la conception du cadre de la recherche sur l'élevage en raison de l'économie de temps qu'elle permet et de son effet coûts/bénéfices avant que des travaux plus approfondis ne soient entrepris par des scientifiques sur un site déterminé. Mais il a déclaré qu'elle n'était utile que lorsqu'elle était utilisée comme complément d'enquêtes au sol qui étaient probablement plus fiables dans l'identification des paramètres importants. Cette technique comportait certaines contraintes telles que la difficulté d'identifier les chèvres en raison de leur couleur et de leur petite taille. M. Milligan a souligne que les moutons et les chèvres ne constituaient pas normalement de problèmes à cause d'un aspect de leur comportement: il s'enfuient généralement lorsqu'un avion approche et peuvent ainsi être facilement distingués des bovins. Il est difficile de distinguer les moutons des chèvres et ils ont donc été normalement classés en "cavins". Le Prof. Saka Nuru a demandé quelle était la fiabilité/efficacité relative de la technique de télédétection par rapport aux enquêtes au sol. M. Milligan a répondu que le dénombrement des animaux à partir du sol prenait beaucoup de temps et qu'il posait de nombreux problèmes théoriques et pratiques. M. Milligan a reconnu que la confirmation au sol était importante, compte tenu du niveau d'informations requis. Les estimations faites à partir de l'appareil peuvent être vérifiées et certaines distorsions corrigées par calibrage basé sur les enquêtes au sol. La connaissance de la situation au sol a également permis de faire des observations correctes à partir de l'appareil. Le Prof. Saka Nuru s'est déclaré entièrement en faveur de l'idée selon laquelle le CIPEA devrait participer à l'utilisation à l'échelle des pays des techniques de télédétection pour avoir une image plus claire de la distribution des ressources naturelles en vue d'une assistance à la planification future du développement de l'élevage.
Le Dr Zulberti a mis l'accent sur la nécessité d'un cadre conceptuel avant la collecte des données, pour éviter le rassemblement d'informations abondantes dont certaines pourraient ne pas être utilisées. Le Dr Chema a également souligné la nécessité de définir clairement un objectif précis au commencement de l'enquête. M. Milligan a accepté ce point de vue et a déclaré que des objectifs spécifiques étaient en fait identifiés avant chaque vol. Les hypothèses et les questions formulées au départ constituaient la base à partir de laquelle les divers types de données étaient collectées et la conception des stratégies d'échantillonnage mise au point.
M. Hiernaux a demandé à M. de Leeuw si oui ou non il estimait que l'utilisation des méthodes de télédétection par satellite et par enquête aérienne pourrait résulter en un accroissement des coûts en raison de la multiplicité des observations qui devaient être faites sur le terrain pour appuyer de telles méthodes. M. de Leeuw a déclaré que cela soulignait la nécessité pour les chercheurs de définir de manière très claire les types de données dont ils ont besoin dans de telles enquêtes.
