M.D. GWYNNE* and H. CROZE**
(*) M.D. Gwynne: UNDP/FAO Kenya Habitat Utilization Project, P.O. Box 30218, Nairobi, Kenya.
(**) H. Croze: Department of Zoology, University of Nairobi, P.O. Box 30197, Nairobi, Kenya.
Results and analysis
Methods of large-scale ecological monitoring have been developed and used in East Africa over the past few years. Systematic sampling techniques have proved most promising for producing the spatial and temporal ecological data necessary to manage or develop large stretches of low and erratic rainfall lands.
In this paper we present a review of some methodologies currently in use; the types of result it is possible to obtain; uses to which the result may be put; and a synthesis of strategy that combines ground, aerial and space technologies into an Ecological Monitoring Unit. We argue that such a unit, the prototype of which will shortly be operational in Kenya, is able to provide descriptive and predictive information on the life-support capacity (productivity: actual and potential) of most regions of the earth's surface.
During the past decade the emphasis of ecological research in East Africa has shifted away from single species investigation (e.g. Kruuk, 1972; Schaller, 1972; Watson, 1967) towards ecosystem studies. This upswing in the gynecological approach has occurred largely in response to the request of resource management organisations (Game Departments, National Parks, Range Management Divisions, tourist development bodies and irrigation schemes) that require large-scale ecological data on which to base utilization programmes for extensive tracts of non-urban land. Efficient utilization requires optimization of the balance between life-support attributes of the land and the type of land-use (e.g. livestock range, agriculture, game reserves), both of which depend on its actual and potential productivity. In general, the more arid the land, the smaller the choice of land-use possibilities.
Determination of productivity, both primary and secondary, is a problem for the ecologist. Obtaining such data is obviously an enormous undertaking and can only be accomplished by a single investigator if he concentrates on a relatively small area (e.g. Western, 1973). Larger regions, such as the East African National Parks and their surrounding ecosystems, can only be efficiently investigated by a multidisciplinary team that includes ecologists, botanists soil scientists, geologists and hydrologists. Such an approach was first attempted in East Africa in the Serengeti National Park (Tanzania) in the early 1960's (Watson, 1967) and evolved into the Serengeti Ecological Monitoring Programme under the general direction of Norton-Griffiths (1972). This programme endeavoured to collect long-term data on climate woodlands, fire, and animal and human populations.
The underlying aim of the ecosystem approach is to determine the spatial and temporal pattern of primary and secondary productivity in a particular ecosystem. The first operational question to ask in a gynecological study is whether the area to be studied is an ecosystem. This usually means: Do the boundaries drawn on a map actually encompass a more or less organically and energetically self-sufficient unit ? The gazetted boundaries are almost invariably political rather than ecological. To ecologists, however, the boundaries to the ecosystem are traditionally determined by the limits of the movement of the largest biomass components (Pennycuick, 1974), which of course may change with time. These movements are caused by something - and part of the problem is to determine what. Hence the gynecologist is faced with having to describe the static structure of the ecosystem (topography, drainage, soils, vegetation, plant and animal community components) as well as analyse the dynamics of the system (changes in time and space of climate, productivity, shifts in community structure, etc.). In an area of many thousands of square kilometers this can be a formidable task and can only be accomplished with careful planning and a systematic approach.
The result has been the gradual development within East Africa of the ecological monitoring concept, from its beginning when methods were sought to answer simple questions (e.g. How many animals are there ?), through the intermediate (e.g. Where are the animals located, and when do they move ?), to the complex (e.g. Why are there so many animals, and why do they move in the patterns they do, when they do ?), culminating in the present, where land-use management questions are being asked (e.g. What will happen to the ecology of the area if it is developed for ranching, or is turned into a National Park ?). This demand for information on a large scale has necessitated the development of techniques that are efficient and inexpensive enough to be applied repeatedly over large areas. Considering the difficulties caused by the rangeland ecosystem being in a dynamic state, constantly changing in response to short- and long-term climatic cycles, the methods of data gathering and data handling developed have been most successful.
Many parts of the world's arid and semi-arid areas are over-populated by both humans and livestock in relation to the available resources. Every low rainfall year threatens catastrophe. The prolonged and severe droughts of Sahelian Africa, the desiccation of northern Kenya and the current crises in northern Ethiopia and in Somalia all emphasise the frequency with which such human misery can and will occur (cf. Ehrlich, 1968). These situations can undoubtedly be eased by the judicious supply of relief aid, but such rescue operations, though dramatic, are only short-term and do nothing to remove the ultimate cause of the problem. A more rational approach is to persuade governments to attempt to optimise land-use in the sensitive zones. To do this successfully, however, they must have simple, quick and relatively easily applied methods that will enable them to determine the demands being made on the land now; its capacity for supporting human life; and the future of that land under different forms of management, both year-long and long-term (Croze, Gwynne and Jarman 1973). Most of the methodology required to provide governments with the means to make these determinations already exists, and in Kenya, for example, where much of it was developed, is already being put into practice. An understanding of the practical benefits of the ecological monitoring concept has led the Government of Kenya to establish the Kenya Rangeland Ecological Monitoring Unit (KREMU), which is jointly staffed by the Ministry of Tourism and Wildlife and the Ministry of Agriculture.
Although widely applied in East Africa (Fig. 1) the ecological monitoring methods used are not well understood elsewhere. It is the purpose of this paper, therefore, to provide a brief outline of the East African methodology and the type of results that can be obtained *.
(*) Sufficient references and contact addresses of monitoring personnel to enable interested readers to enter the field with the minimum of difficulty may be obtained upon request from the authors.
I. The nature of ecosystem data
Ecological monitoring uses data from three general categories:
i. Environmental - including data on climate soils, topography, hydrology and floristic dynamics.
ii. Faunal - including data on wildlife and livestock numbers, distribution, population dynamics and habitat utilization.
iii. Economical/political - including data on current land-use forms, projected land demands and national development goals.
This paper will only consider the operational aspects of collecting, analysing and interpreting some of the data from the first two categories. Choice of a collecting or sampling strategy obviously depends on the spatial and temporal distribution of the phenomena being measured. It is convenient to classify ecosystem attributes along a continuum of mutability, viz:
- water holes
- static animal features such as termite mounds
- plant physiognomy (cover, vegetation type, etc.)
- plant community composition
- zoogenic features (wallows, salt licks, etc.)
- distribution of non-migratory large mammal species
- human settlement (villages, roads, farms, ranches)
"Ephemeral or seasonal attributes"
- soil moisture
- plant phonology (greenness)
- plant productivity (biomass, part composition, chemical composition, energy content, etc.)
- distribution of migratory large mammal species
- large mammal productivity (biomass, reproductive state, condition food offtake, etc.)
- large mammal population structure
- surface water
Finally, data may be collected from three operationally separate levels - from the ground, from the air, and from space via satellite (e.g. Earth Resources Technology Satellite). Aerial sampling will be discussed first because the deployment of more traditional ground techniques may be profitably considered as a function of the aerial strategy, and similarly the interpretation of the satellite imagery may be done in the spatial framework of the aerial reconnaissance. Aerial survey is, in any case, the logical first field step in the resource assessment of any new large development area, and the techniques used for aerial monitoring can provide useful quantified quick-look data at low cost (see also Watson, 1969 b).
II. Aerial techniques
A. Systematic Reconnaissance Flights (SRF)
Ecological data collecting from light aircraft is now widely accepted as being both necessary and efficient (Zaphiro, 1959; Swank, Watson, Freeman and Jones, 1969; Pennycuick and Jolly, 1974), as it is possible to collect large quantities of data from extensive areas of land quickly and for small cost. Floral and faunal ecosystem components must be measured periodically since they are usually cyclic or serial (see also Cobb, 1975).
Although the details of an SRF will vary from area to area, depending on terrain, vegetation type, local seasons and an investigator's particular brief, we may construct a typical method currently in use in East Africa. Standardisation of technique is, in any event, desirable because it facilitates comparison between ecosystems.
The first step is to overlay the study area with a reference grid such as the UTM 10 x 10 km grid system. The grid squares can then be used for orientation during flying and for presenting and analysing distribution data.
Systematic flight lines 10 km apart, e.g. centered on the grid squares, are drawn on a 1:250,000 topographical map of the study area. The flight lines are orientated north-south or east-west as the shape of the area and wind conditions dictate. In general it is best to avoid cross-wind flight lines (problem of pilot navigation), very short flight lines (problem of statistical treatment of the data), and very long (> 100 km) flight lines (problem of observer fatigue).
The flights should not be made during very early or very late hours because of insufficient light and deep shadows. They should also not normally be made during the mid-day period (1000-1500 hours), because this is the time of maximum solar radiation during which many herbivore species retreat into the shade for rest and rumination and are therefore not easily seen from the air; this behaviour pattern is modified, however, by seasonal variations in cloud cover (Gwynne and Robertshaw, in preparation), which permit longer daily flight times.
The SRF's are flown by a crew of four in a single-engine aircraft at a speed of cat 150 kph and a height of, for example, 100 m above ground level. The pilot is responsible for height control, navigation and spatial location, informing the crew of both transect numbers and subdivisions every 10 km flown. All data are recorded by subdivision (i.e. grid square). KREMU aircraft are also equipped with a Very Low Frequency (VLF) navigation system that reduces navigation errors and permits transects to be flown over flat featureless terrain with great accuracy; repetitive flights along the same transect are, therefore, possible provided the starting point of each flight line can be re-located.
Height control is critical (and is best maintained by radar altimeters, although a simple shadow meter can be used: Pennycuick, 1973) because the two rear observers are counting animals by species between two streamers fixed to the wing struts on either side of the aircraft. The streamers are so spaced that at a particular height they subtend a strip of known width (say 250 m) on either side of the aircraft (Pennycuick, 1974). Thus along the flight-lines the crew is counting animals (and features) in transects of known area. By proportionality, estimates of population size and variances may be made (Jolly, 1969a and b; Norton-Griffiths, 1973; Sinclair, 1971). Using portable tape recorders, the rear observers record transect and subdivision numbers together with species and numbers of animals observed. Animal groups too large to count accurately (> 20; e.g. herds of sheep and goats) are photographed using a large magazine, motorised, automatic exposure 35 mm camera (R. Bell et al., 1973; Norton Griffiths, 1973 and 1974).
The front right observer next to the pilot also uses a tape recorder and camera to document habitat information - topography, drainage, vegetation type, cover and other " permanent " features, as well as greenness of vegetation, grass height, intensity of grazing, presence of water and other seasonal features. In a series of SRF's made during a monitoring programme, the "permanent " features need only be recorded once during an initial flight.
The adaptation of high resolution colour videotape recorders (VTR) for use in place of the rear observers is also under development (Gwynne, 1972). The re-usable video tape's virtually frameless format and its instant play-back and stop motion abilities make it particularly suitable for monitoring work. Use of the VTR allows all animal counting to be done on the ground free from flight fatigue and consequent observer bias, while at the same time permitting the collection of quantifiable data on other ecological parameters, e.g. vegetation composition and cover, surface water area, burn area and human settlement activities. A video-tape library of suitably selected flight transects can be built up thus allowing past surveys to be re-examined using updated analytical techniques.
However, the current high capital cost of professional VTR and necessary ancillary ground equipment may reduce the applicability of the VTR method. An alternative system also under development in Kenya involves the use of 35 mm photographic colour film and/or small size cine film, with the camera controlled by an intervalometer and individual frames being exposed at intervals of one every five seconds to one every seven seconds depending on the aircraft ground speed. This method has some of the advantages of the VTR system together with low initial capital cost. The major disadvantages are the processing time required before census and monitoring data can be obtained, and higher operating cost per SRF (in terms of film and processing). The use of image analysing computers may aid in the extraction of data from SRF photographs.
The frequency of SRF's is a function of cost and the rapidity of seasonal changes - every month, every two months, every quarter and every major season have been used. The cost of flying such a survey is currently about US $ 50/1,000 km², not including observer time and cost of films and recording equipment.
The data are all recorded with reference to a sub-unit on a transect, that is, within a particular 10 x 10 km grid square. The data are transcribed from the tape machines directly onto computer coding sheets. At this stage, distribution of animals, greenness, water, vegetation type, etc., may be plotted by hand onto gridded working maps of the study area (see " Results "). The data may also be punched onto computer cards and transferred to magnetic tape for storage and analysis. Programmes are now in use which file the data and produce line-printed distribution maps as well as animal population and biomass estimates (e.g. by Norton Griffiths and Pennycuick, 1973; Cobb and Western, ANPR Programme, personal communications). A computer is not necessary, but it saves much time and labour. The Serengeti Research Institute programme, for example, dealt with 34 pieces of temporal, spatial, faunal and habitat information for each of the 300 grid squares of the Serengeti ecosystem that were overflown monthly over a period of three years. Further, a computer facilitates more complex analysis of the data-species associations, multiple correlations, trend surface analysis, cluster and multiple discriminant analysis, etc.
We have outlined the systematic aerial survey strategy in some detail, for although currently much in use in East Africa (Fig. 1), it is not well known elsewhere. There exist as well, however, numerous other specialised uses of light aircraft for census and monitoring purposes - a few examples of which are given below.
B. Aerial sampling for large mammal population estimates
Ecologists are frequently faced with the problem of determining the size or density of a particular species population. Although such information may be extracted from a series of SRF's (Croze, in preparation), often because of the small sampling fraction (e.g. < 5 %) the confidence limits of the estimates can be quite large. Clearly, to detect a change in population size from, say, year to year, the population estimates should be as precise as possible.
If we know the movement range of a particular population or are able to delineate concentration areas from the distribution data of an SRF, we may then concentrate sampling effort over that particular area. Both systematic flights at a greater intensity (i.e. flight lines spaced 5 km or 2.5 km apart) and stratified random samples have been used (Jolly, 1969 a; Pennycuick, 1969; Watson, 1969 a, 1969 c and 1972; Western, 1973). In general, precision is a function of sampling intensity (Caughley, 1972; Norton-Griffiths, 1973), and only with precise population estimates can we monitor short-term population fluctuations.
C. Large mammal population parameters from low-level aerial photography
Aerial photography, as we have already seen, plays an important role in increasing the accuracy of counts. Sinclair (1969) used low-level oblique photographs of buffalo (Syncerus caffer Sparrman) breeding herds to monitor recruitment through the proportion of first-year animals in known herds. Croze (1972) used measurements of elephant back-lengths from vertical aerial photographs to determine the age structure of elephant populations, and Western (unpublished data) used pre- and post drought colour (Ektachrome) photographs of Maasai cattle herds to assess colour-morph-specific mortality rates.
D. Habitat information from high-level aerial photography
The large-scale (ca. 1: 50,000) aerial photograph is an important tool of geologists, cartographers, soil scientists, landscape ecologists and botanists. Virtually all geological, topographical, soil and vegetation maps are compiled initially from black and white, and more recently, false colour (near infra-red colour) aerial photography (cf. UNDP/FAO Kenya Range Management Project Rangeland Surveys, FAO, 1967-73).
Sequences of aerial photographs taken of the same region at different times can be used to enumerate gross changes in vegetation physiognomy (e.g. the rate of loss of woodland trees: Croze, 1974).
E. Landscape classification
We usually think of mapping a study area according to specialists' needs: soil maps, geological maps, vegetation maps, or very general low-resolution ecological zone maps (e.g. Pratt, Greenway and Gwynne, 1966). A little-known corner of aerial ecological technique is occupied with strategies of landscape classification (e.g. Beckett and Webster, 1965; Scott, Webster and Lawrance 1971; Gerrisheim, 1974). The approach develops ramifying systems by which large land units are subdivided into smaller ones according to the relative homogeneity of morphology, soil and hydrology, and vegetation. The classifier delineates apparently homogenous areas on large-scale aerial photographs and then checks the delineation on the ground. The product standardised, identified units of landscape - come in various sizes and are suitable for different levels of management or research. Land Regions are useful units of organisation for an ecosystem study of management plan; the smaller Land Facet is sufficiently homogenous over its extent that it could be managed uniformly for all but the most intensive kinds of land-use; the Land Elements of which the Facet is composed are only sufficient, for example, to describe the range of one rodent population.
Without doubt the technique as practiced now is subjective and requires " an expert's eye " to identify unit boundaries; and conceivably not everyone can acquire the knack. The problem of individual variations may be solved by using numerical classification and descriptive techniques such as cluster analysis and trend surface analysis.
III. Ground sampling strategies
Ground sampling techniques are, of course, those most frequently used by ecologists, and although we can collect much data from the air, we must eventually come back to earth. Geology, cartography, hydrology, soil science and descriptive botany are all specialised disciplines, sophisticated and well documented. These data are clearly vital to the complete understanding of the structure and dynamics of an ecosystem. For our purposes they provide more or less static descriptions of what we have chosen to call the " permanent " features of an ecosystem. Over these features we lay our reference grid described in the previous section and proceed to monitor biotic dynamics. It is important to note that productivity can be monitored without the "permanent " data, but investigations of causation are incomplete without them.
Productivity depends temporally and proximally on water - surface, sub-surface and precipitation - taxed by transpiration and evaporation. In the arid and semi-arid tropics it is the availability of this water that becomes important rather than the supply, as often these areas receive a higher annual rainfall than more temperate regions (Gwynne, 1966). Meteorological methods and equipment for estimating rainfall are well known (Monteith, 1972; Munn, 1970; HMSO, 1956; WMO, 1965). Ideally, meteorological stations ranging from simple monthly storage rain gauges to complete hydromet installations and automatic weather stations (Leese, Strangeways and Templeman, 1973) should be spaced evenly over the ecosystem. Rain gauge distribution at a density of 1: 1,000 km² is desirable, but some workers recommend even greater densities, e.g. 1:500 km² (Grimsdell, 1975). Consideration of site accessibility and cost, however, render the attainment of this goal unusual, particularly if very large areas are involved, as in a nationwide monitoring programme. Fortunately, there are useful statistical techniques such as trend surface analysis (Norton-Griffiths, Herlocker and Pennycuick, 1975) that allow extrapolation and isohyet delineation from irregularly placed stations.
B. Soil Moisture
The amount of water available to plants is a major factor affecting their productivity. As the majority of this water is that which is present in the soil within the plant root range, after the physical requirements of the soil have been met, it is important in a monitoring programme to follow seasonal changes in soil moisture throughout the soil profile, preferably at sites close to the rain gauges. These sites should also be chosen with regard to the prevailing soil type (which will influence soil permeability and the physical retention of water in the soil) and position in the soil vegetation catena (e.g. hill summit, hill slope, valley sump).
Water availability patterns can be obtained from the logarithm of the resistance in ohms of plasterof-Paris blocks buried in the soil at different depths (Pereira, 1951; Pereira and Hosegood, 1962). A more useful and rapid method for obtaining quantifiable data, however, is to use a neutron probe. The probe is lowered into a permanently installed aluminium tube in the soil and readings taken at appropriate depth intervals (usually 15 cm or 30 cm for a total depth of at least 2 m). In this method fast neutrons are emitted from the radioactive probe source and are scattered and slowed by collisions with soil element atomic nuclei, maim): the hydrogen of soil water. These slow neutrons are detected by the probe and converted into pulses, which are displayed as the count rate (the wetter the soil, the higher the count rate). The count rate is thus related to the hydrogen content of the soil and changes in moisture content (J. Bell, 1973). The instrument has to be calibrated for each soil type.
C. Systematic Ground Survey (SOS)
The strategy for SGS is essentially similar to that of SRF except that more detailed data are collected for a disproportionate increase in cost. Cost aside, however, there is as yet no aerial substitute for SGS at the finer levels of resolution (Jarman, 1971; Western, 1973; Rodgers, 1974). The survey lines are seldom as regular as in the SRF because the observer's ground mobility is restricted by the nature of the terrain and the density of the vegetation cover (Rainy, 1969). SGS sample site lines, therefore, tend in many areas to parallel existing game trails, paths and roads. The observer uses the time-saving advantage afforded by the tracks and places the recording sites in the undisturbed habitat to one side of the path. Site intervals of 0.5 and 1 km have been used in East Africa for SGS monitoring transects. Data recorded at sample points include, for example, grass phonology, height, species, species density and cover; woody plant species, growth stage, density and cover (and intensity of browsing or grazing by large mammals). Samples are collected for chemical composition and energy content determinations. The Point-Centered Quarter (PCQ) technique, a plotless sampling strategy, provides a very rapid method of determining plant species density provided the species being measured are not too contiguously distributed (Cooper, 1963; Dix, 1961; Heyting, 1968; Croze, 1974).
Ground sampling by vehicle and on foot has been done by Rodgers in the Selous Game Reserve, Tanzania; by Western in Amboseli National Park (formerly Amboseli Game Reserve), Kenya; by Rainy in Samburu District, Kenya; and by Gwynne in South Turkana and Lamu District, Kenya.
If the area monitored is very large and widely spaced sampling sites such as the 250 ground truth stations of the KREMU (see below) are used instead of transects, then aircraft are used to transport the field teams wherever possible.
Data collected by SGS must always be related to existing cartographic, geological, soil, vegetation and land system maps, climatic data acquisition sites or mean annual rainfall, rainfall probability and potential evaporation data (e.g. Rijks and Owen, 1965; Woodhead, 1968 a and b).
D. Primary production biomass and use
Various methods for estimating above-ground primary production of grasslands are given by Milner and Hughes (1968) and for forests and woodlands by Newbould (1967), but these methods are more successful for grassland than for woodland and forest. Traditional methods involve destructive sampling by clipping quadrats of appropriate size, sorting into species and plant part (e.g. leaf, sheath, stem, inflorescence, etc.), followed by dry weight determination and chemical analysis. This is both time consuming and tedious, so that other techniques have been sought. One of the most useful has proved to be the canopy intercept method (Goodhall, 1952; Warren-Wilson, 1960 and 1963) normally used to calculate cover and leaf area index. Suitably calibrated it will give biomass data by species on green leaf, green stem, dry leaf, dry stem and any other plant component needed. It is most applicable to grassland and has been used very successfully by MacNaughton in the Serengeti (unpublished data). Similarly in the Amboseli area of Kenya, Western (unpublished data) has developed a relationship between biomass density, grass height (culm or leaf) and plant density; this method has also been used by Cobb in the Tsavo National Park, Kenya. Both of these methods are rapid to use and nondestructive, but neither gives data on chemical content, for which separate samples must be clipped. They do permit routine determinations of grass standing crop at frequent intervals, both temporally and spatially- (e.g. monthly and every 0.5 km), along lengthy monitoring transects.
A new monitoring technique for green biomass determinations based on canopy spectroreflectance (Pearson and Miller, 1973; Tucker, Miller and Pearson, 1974) is at present in use in East Africa, having been introduced at the Serengeti Research Institute by MacNaughton. The ratio of chlorophyll reflectance (8,000 A) to absorption (6,750 A) has, in the grass layer at least, a linear relationship to green biomass density. The relationship for woodland species is more complex and remains to be elucidated as it is affected, for example, by the depth of canopy and the amount of chlorophyll contained in the bark. The use of the ratio is necessary to reduce the effect on the reflectance values of the light bands by factors such as cloud cover, time of day and sun angle. The instrument (a digital spectral photometer with a sensing probe for each wave band) must always be calibrated against actual harvest determinations prior to use in any new ecosystem. It can be used on the ground for sample green biomass estimates, and from aircraft. MacNaughton (unpublished data) has used green biomass determinations made from the air to construct seasonal green biomass density maps for the Serengeti National Park (Figs. 12 a and b).
The use of primary productivity by livestock or wildlife species can be determined a) by clipping and weighing grass within and without an animal-proof enclosure (e.g. Bell, 1971; Western, 1973); b) by inspecting vegetation for signs of grazing (e.g. Vesey Fitzgerald, 1969 and 1974) or browsing (e.g. Vesey Fitzgerald, 1973; Croze, 1974); c) by direct observation of the animals feeding (e.g. Goddard, 1970; Leuthold, 1970; Field, 1971; Jarman, 1971); or d) by the use of direct sampling of ingested food items by means of oesophageal fistulation (e.g. Van Dyne and Torell, 1964; McKay and Frandsen, 1969; Duncan, 1974; Kreulen, 1975), rumen fistulation (Kreulen, 1975) and more recently, repetitive trocar sampling (Fellis and Spillett, 1972). It has proved difficult to quantify categories b) and c) in terms of absolute biomass used - but the increased use of remote sensing techniques should soon eliminate this problem.
E. Ground acquisition of large mammal data
Ground surveys to establish animal population parameters are not normally necessary if SRF data are available for the species concerned, unless the vegetation structure is such that animal concealment leads to major bias (Watson, 1969; 1972). Population estimates of some animal species not readily seen from the air, such as those at the small end of the size spectrum (e.g. oribi, Ourebia ourebi) and those that spend the day largely in concealment (e.g. greater kudu, Tragelaphus strepsiceros) can best be made from the ground (e.g. Lamprey, 1963), as can checks on herd structure and composition. Data on age structure, growth rate and reproductive history and recruitment can be most reliably gathered from collected specimens (Laws, 1966, 1969; Grimsdell, 1973) and ground observations (Sinclair, 1973 a and b).
Small samples of slaughtered animals will also produce information of food offtake derived from rumen content analyses (Gwynne and Bell, 1968; Field, 1972; Jarman and Gwynne, 1976) and faecal analysis (Stewart, 1967), although results obtained by these methods must be treated with caution in view of differential digestion effects (Thornton and Minson, 1973; Duncan, 1974). Such results can be related to ground observation of feeding behaviours and quantitative determinations of vegetation on offer (Jarman and Gwynne, 1976). Estimates of body condition can be made using muscle-to-bone ratios, kidney fat and bone marrow condition (Sinclair and Duncan, 1972).
F. Ground checks on SRF data
Although SRF surveys can operate independently, they are of more value for resource measurement and ecological monitoring if they are closely linked to ground truth data. Interpretation of SRF information becomes easier when data from ground studies are available (cf. III-A-E), and many investigators wish at the end of the first 2-3 years that they had spent more time devising adequate ground sampling techniques, particularly with regard to vegetation production, growth stage and animal herd composition.
IV. Remote sensing from satellites
A. Satellite Imagery
The earliest remote sensing device was a camera attached to a balloon, which flew in 1858 (Rabchevsky, 1970). Since then a series of sensing platforms have been launched (Colvocoresses, 1974) to record or monitor events on or near the earth's surface, using photography or electromagnetic spectral sensors. Now weather; crop condition; forest diseases; air, sea and fresh water pollution (Colwell, 1971); and more recently primary production (Carneggie and De Gloria, 1974 and personal communication) are all being monitored, mostly on an experimental basis, via satellite (see also Bale et al., 1974).
Since 23 July 1972 the first Earth Resources Technology Satellite (EATS-1) has been orbiting the earth at a height of 900 km, passing over the same place on the surface every 18 days. On this vehicle is a multispectral scanner that senses the wavelength and intensity of light reflected from the earth's surface features in units with a resolution of about 80 X 80 m. The information in four spectral bands is coded into magnetic tape on the ERTS and transmitted to earth receiving stations in digital form (Rabchevsky, 1970; Colwell, 1971; CENTRO, 1971).
For the ecologist the ERTS output has two important formats:
1. picture-like images (false colour composites; black and white positive or negative in each of four spectral bands) that appear as aerial photographs, each depicting a 185 X 185 km area of the earth's surface.
2. digital output of the intensity and wavelength of light reflected from any 80 X 80 m portion of the earth's surface.
The images show clearly all major land features - drainage lines, mountains and faults. Spectral band 7 (0.8 - 1.1 m) emphasises soil boundaries, while spectral band S (0.6 - 0.7 m) enhances green vegetation (which appears red on false colour images).
Essentially the same data, only with a finer possible resolution and accuracy of interpretation are contained in the digital output - indeed the digital information from the multispectral scanning device is integrated by a computer to produce the image. For example, a false colour image will show an area of green grassland as bright red and an area of drying grassland as bright pink. The digital analogs are reflectance intensities in band S. the red portion of the electromagnetic spectrum, for the two areas. Clearly, these numbers can be calibrated to primary production biomass density (Carneggie, 1974 and personal communication); Rouse and Riter, 1973). Thus subsequent estimates of regional primary production may be made using data collected from a height of 900 km. Since the satellite passes overhead every 18 days, it is possible to monitor the ebb and flow of primary production over an entire ecosystem. Animal biomass distribution data from Systematic Reconnaissance Flights may then be correlated directly to absolute estimates of primary production.
In this section we will review a selection of the types of results it is possible to obtain using the methodologies outlined in the previous section. The data are taken from studies recently completed or currently in progress in East Africa. We will present them in this order - ground, air and space - keeping in mind that we ultimately want to integrate the three types of results into a synoptic view of the ecosystem. No model-constructs or simulation results will be presented, for the simple reason that none are yet available for East Africa (an exception on a small scale is Norton-Griffiths, in preparation, a construction of ecological regions using a cluster and multiple discriminant analysis). We will, however, discuss a few of the analytical techniques.
I. Results from ground monitoring
A. Rainfall isohyets using Trend Surface Analysis
Norton-Griffiths et al. (1975, in press) constructed rainfall isohyet maps for the 25,000 km² Serengeti ecosystem (Figs. 2 a and b), using data from 62 irregularly spaced monthly storage rain gauges. He subjected the data to trend surface analysis (e.g. Gittins, 1968), a multiple regression technique that finds the polynomial equation describing the " surface of best fit" for a continuous variable in two-dimensional space. Regular contours of the surface are, in this case, rainfall isohyets. The advantage of the method is that it presents a rainfall map, which, unlike the products of more subjective techniques, has been tested with an analysis of variance. One can also use data on rainfall variability, evapotranspiration and climatic indices.
B. The response of large grazing mammals to pasture growth stage
Each large mammal herbivore species has a specific feeding mechanism, a combination of anatomical and behavioural characteristics, which enables it to feed off a particular physiognomic vegetative type that is optimal for the species in terms of ensuring the maximum food intake for the minimum expenditure of energy (Gwynne, 1971). In the sequential grass species growth that follows the onset of rain (Cobb and Gwynne, unpublished data) grazers move from grass species to grass species as each approaches the physiognomic state optimal to that animal species. This utilization by grazing animals is closely related to changes with age in grass leaf tensile strength (Gwynne, in press).- An indication of the effect of physiognomic vegetation stage on feeding efficiency can be seen in Fig. 3, which shows that among sheep an increase in grass height results in an increase in time/bite. Goats feeding in the same area at the same time showed the same relationship, but consistently took longer per bite than the sheep; in neither case was bite size quantified (Gwynne, 1974). Monitoring of vegetation physiognomic state is essential in understanding large mammal feeding habitat utilization.
C. Elephant use of wooded-grassland tree species
In a study of elephant-woodland interaction, Croze (1974) measured woodland composition and elephant-browsed trees using an extremely efficient ground sampling technique known as the Point-Centered Quarter (PCQ) method (e.g. Heyting, 1968). A comparison of tree species used with tree species availability (Fig. 4) showed that one out of six species was taken preferentially, one was avoided, and the rest were taken as expected by proportionality. One can only talk about preferential use of pasture components if the relative availability of the components is known.
II. Results from Systematic - Reconnaissance Flights
A. SRF: large mamma/distribution by season and vegetation type
As has been outlined under "Methods", large mammal distribution as recorded on an SRF can be ranked into density classes (Norton-Griffiths, 1972) and plotted on maps according to the sub-unit of observation. Plotting can either be done manually or with one of the various computer programmes available (e.g. ANPR of Cobb and Western, cf. "Methods") where distribution is given in the form of a density-ranked digital printout map. Figs. 5 a and b illustrate the actual dry season distribution of two closely related Alcelaphine antelopes, the topi (Damaliscus korrigum) and the little-known Hunter's antelope (Damaliscus hunteri) in the Lamu-Garissa area of Kenya. These distribution maps clearly indicate that at this season the two species occupy contiguous areas. Comparison with an SRF-generated broad scale vegetation map (Fig. 6; cf. II-C) shows that the Hunter's antelope is more or less restricted to a low rainfall area of Acacia dry thorn bushland, with the topi occupying an ecotone zone of woodland with grassland glades. Neither species occurs in forest or in areas of dense woodland (Duncan, Gwynne and Jarman, in press). Further analysis suggests that the dry season distribution of the topi is closely related to the occurrence of permanent surface water.
Similar seasonal distribution data from SRF's can be subjected to trend surface analysis (Norton-Griffiths and Pennycuick, 1973) to show the basic underlying distribution patterns. Figs. 7 a and b give wet and dry season surfaces of best fit (p. < 0.01) for the Hunters antelope. These show that in the wet sea-son the species is in two local concentrations, both occurring in the Acacia dry thorn bushland, with outliers extending well into the higher rainfall wooded grassland ecotone also occupied by the topi. Thus the two species can overlap at this time. In the dry season the eastern population disappears, presumably by retreat into Somalia, while the western group moves to the northwest, closer to the river Tana. Similar seasonal distribution data for cattle (Figs. 8 a and b) show that there is little change in the overall occurrence pattern with season, the animals mainly being found on the woodland-grassland ecotone complex between the dry thorn bushland and the forest and dense woodland. The higher concentrations in the dry thorn bushland of the northwest, near the Tana river, are due to livestock moving through the area along stock routes (cf. Fig. 9). The increase in wet season cattle numbers in the centre of the map is closely correlated with green grass availability as shown by SRF greenness estimates (cf. Figs. 17 and 18). The wet season absence of stock from zones in the northeast and southwest is due to apparent avoidance of dense forest and floodlands, respectively (Duncan, Gwynne and Jarman, in press).
These few examples from one area of Kenya illustrate some of the types of large mammal distribution information that can be obtained from SRF's quickly and at relatively little cost; that which is presented here represents only a small portion of the total gathered. Similar data are available from each of the areas in East Africa being monitored at present. Such data can be used to construct habitat utilization probability maps showing the likelihood of any particular area being used as feeding grounds for a particular species or complex of animal species (e.g. domestic livestock, wildebeest, etc.) in a manner directly analogous to the better known rainfall probability maps (e.g. Glover, Robinson and Henderson, 1954).
B. Patterns of use by large mammals
Large mammals repeatedly walking over the same route to and from water or grazing grounds soon abrade the vegetation, leaving obvious trails, often eroded deep into the soil. From the air most rangelands are seen as a maze of inter-connected minor game and livestock paths that sometimes merge to form large, very well defined routes for long distances. The causal species can often be inferred from the characteristics of the route alone e.g. the broad, smooth, dung-littered, often deeply worn highways of the elephant, and the wide, parallel-ribbed tracks of herds of driven domestic cattle (Fig. 10). These major routes can be mapped during SRF by having each observer note the position of the track when it is crossed by the flight line and record the approximate direction of the track according to cardinal and half-cardinal compass lines (e.g. E-W, NE-SW, N-S). The route net becomes apparent when these are plotted on the SRF grid map. The direction of trail use can be ascertained by actual observation of animals using the trail or inferred from other data. Fig. 9 shows a cattle stock route map for the Lamu District, Southern Garissa District area of Kenya, constructed from data collected by this method during the course of a routine SRF in 1973. The dotted line indicates a very well defined trail now overgrown and no longer in use; its course follows closely that of a track already marked on the 1:250,000 map sheets of the area (Gwynne, unpublished data). The fine arrows indicate the general direction of minor trails too numerous to show in detail.
C. SRF: Preliminary vegetation mapping
If the rear observers during an SRF are requested to record vegetation type by structure (cf. Pratt, Greenway and Gwynne, 1966), then a preliminary vegetation map of the survey area may be compiled very quickly. Fig. 11 is such a map for the Ilkisongo region of Southern Kajiado District in Kenya. Four broad vegetation types are distinguishable - grassland/swamps, bushed grassland, mixed bushed/ wooded grassland, and forest. Intrusion of cultivation into the area is also mapped. A knowledge of the local species allows species lists for zones to be compiled from the air. An SRF-generated vegetation map can serve as a useful temporary working basis before detailed aerial photographic and ground work. Fig. 11 involved 15 hours of flying and represents only a fraction of the information gathered during the flight.
D. Green biomass estimation
Figs. 12a and b show seasonal green biomass density isopleth (gms/m²) maps prepared by MacNaughton (unpublished data) for the Serengeti National Park in 1974, using the digital spectrometer " Biometer " (cf. " Methods ", D) carried in a light aircraft. Flight lines were approximately north-south, with the meter readings being taken at approximately regular intervals. When values showed that the green biomass gradient was rising, the interval between readings was shortened and adjacent flight lines run to delimit the steep gradient boundary. Map isopleth lines were fitted by eye (McNaughton, unpublished data). This powerful new tool represents a major advance in monitoring technology, even though at present it can only be satisfactorily used over grassland and wooded grassland (green biomass values for forest and woodland are minimum values; however, as such they are still useful). The development potential for this method is very great.
E. Trend surface analysis of non-systematic distribution data
It should be obvious by now that we are in general advocating systematic methods of data collection. Occasions may arise, however, when one must deal with unsystematic data. In some instances, they may be analysed in systematic ways. Croze (1973) surveyed parts of Northern Kenya to determine the instantaneous dry season elephant distribution. Time was limited, so it was necessary to ignore stretches of completely dry country and restrict the survey to green areas where elephants were expected. 57 % of the total 368 10 X 10 km UTM grid squares in a portion of Samburu District were overflown. Data from this partial coverage were analysed with trend surface analysis (see above). Surface equations were generated, and from the analysis of variance the quartic surface (Fig. 13) was chosen as the surface of best fit (p < 0.01). Such a technique could well be applied to unsystematic historical distribution data to fill in gaps from before the initiation of an ecological monitoring programme.
F. Population parameters from aerial photography
Croze (1972) illustrated a method for assessing the age structure of African elephant populations using low-level vertical aerial photography (Fig. 14). The method depends on a growth-curve for elephants, which has been obtained from post-mortem (i.e. ground) studies. The age structure may be used in the usual ways - for determining the biomass density of a population the size of which has been estimated by aerial sampling ("Methods ", above), for assessing the " health " of a population, and as a basis for deductions about recruitment and population growth.
G. Habitat dynamics from aerial photography
The change of the structure and abundance of woody vegetation may be monitored using sequential low-level aerial photography. Fig. 15 shows the striking reduction in thicket vegetation in the north of the Serengeti National Park (after Gerresheim in Norton-Griffiths, 1972). Relative changes in wooded grassland may be determined by means of a grid sampling method (e.g. Norton-Griffiths, gunning and Kurji, 1973) or by direct count in sample plots on the photographs (e.g. Croze, 1974). The results, of course, are descriptive. If the results are combined with concurrent monitoring of modifying factors (animals and fire) and controlling factors (soils and climate), then causal links may be identified.
III. Results from satellite remote sensing
A. Soil mapping from ERTS imagery
ERTS images in the form of false colour composites and monochromes of spectral band 7 (0.8 - 1.1 m) have proved most useful in making exploratory soil surveys of large areas of semi-arid rangeland quickly and at low cost. the drier regions are most suitable, as the sparseness of the vegetation allows the spectral signature of the soil to come through relatively unimpeded. the ground truth necessary to categorise the various soil types can be obtained during ground monitoring programmes, while areas of uncertainty can be checked for soil type by light aircraft reconnaissance flights. fig. 16a shows a portion of such a map delimiting the main soil types of the Lamu-Garissa region of Kenya, prepared directly from ERTS images (fig. 16b) in this way and related to existing geological data (Gwynne, in preparation). there is good agreement between this map and one of part of the same area prepared at a later date by others using more conventional methods.
B. Primary productivity, monitoring using ERTS imagery
A comparison of (a) grass-greenness data collected during an SRF in Southern Kajiado District of Kenya (Croze and Western, in preparation) and (b) the subjective colour range of spectral band S from an ERTS image of the same area two weeks later, shows very good agreement (Wilcoxon test, p < 0.6; Figs. 17 a and b). The match was not perfect because of a small time lag and because the ERTS scanning device records the greenness of both woody and herbaceous vegetation. The woody vegetation greenness component could be removed by subtracting a dry season (green trees but no grass) standard. Seasonal grass cover reflectance as recorded by ERTS could be correlated with primary productivity on the ground. Then seasonal primary productivity could be monitored for very large areas directly from the satellite imagery and quantified, using the digital output of the ERTS data. For monitoring productivity, productivity model building, and predictions of productivity failure (drought and famine conditions), satellite imagery is becoming invaluable. Fig. 18, showing the distribution of cattle in the Amboseli area at the time that the SRF greenness distribution was determined, should be compared with Figs. 17a and b.
I. Synthesis of methodologies
Long-term gynecological monitoring probably first began in 1947, four years after Oxford University received Marley Wood on the Wytham Estate as a gift, and Charles Elton began to lead generations of researchers and students through the Wytham ecosystem. After a quarter of a century of repeated measurements in time and space and the development of a central data storage system, all coordinated by a research committee, Marley Wood, all 4 km² of it, must be the best understood ecological system in the world today. In America the systematic, integrated, large-scale team approach was initiated on a grand scale by Van Dyne and his co-workers in investigating temperate grasslands (Van Dyne, 1969, 1972). Elton and Van Dyne have much in common with respect to the conceptual basis of ecosystem research - the differences are in scale of operation and analytical techniques. The approach of the Grassland Biome group is essentially a highly organised network of detailed ground studies aimed at establishing and modelling the cause-effect chains of temperate grasslands.
For many developing countries (with vast stretches of semi-arid or at best dry sub-humid climatic zones) we recommend, initially at least, an approach that combines the three operational levels of ground, air and space. There are at least three things to consider:
1) Few countries, even with international aid, can afford a national commitment of a team of 90 scientists and an annual budget of US $ 2.0 million to be spent on ecological research (Van Dyne, 1972).
2) Land resource management agencies require practical answers quickly, because of the exponentially increasing demand for land-use plans. Investigations into causation usually take time. If productivity descriptions, correlations and at least preliminary predictions can be made quickly, then so can sensible management plans. Detailed causation studies may be postponed until time and funds allow. The fact that large areas must be surveyed (whole countries) and that large, far-ranging animals (pastoral stock and wildlife) must be accounted for make aerial reconnaissance indispensable.
3) Finally, one might argue that there exists a greater need in xeric tropical regions for baseline data and the monitoring of gross climatic and vegetation dynamics. For one thing, tropical ecology has only been seriously studied in the last decade: few background data are available. For another, we are beginning to suspect that a fundamental process of temperate habitats - that of a seral progression to an equilibrium state - may be the exception rather than the rule in semi-arid to dry sub-humid tropical ecosystems. The type of small-scale study that might give much insight into cause-effect links in a temperate " old field " could lead to false predictions in the tropics, if we fail to recognize that we are in a phase of a dramatic vegetation cycle.
Croze, Gwynne and Jarman (1973) outlined an international Ecological Monitoring Programme that would design, implement and coordinate Ecological Monitoring Units (EMU) in various countries. An initial EMU is about to become operational in Kenya as a joint Kenya Government-Canadian International Development Agency venture. To monitor approximately four-fifths of the country the total cost, including all operating expenses, capital outlay and staff salaries, amounts to just under US $1.0 million per year. If only operating and maintenance costs are considered, however, the outlay reduces to US $ 200,000 per year, or US $ 400/ 1,000 km² of country surveyed. The SRF component alone without the associated ground programme will cost about US $ 50/1,000 km² of country surveyed.
Since this first EMU is conceptually and operationally a synthesis of the methodologies already discussed, we shall in the following section present an outline of its structure.
II. An integrated research programme: the Kenya Rangeland Ecological Monitoring Unit (KREMU)
The land area of Kenya is about 570,000 km² and the country has a present population of some 12 million concentrated in three main areas - western Kenya, central Kenya, and along the coastal strip. Currently the population has a geometric growth rate of 3.3 % per annum. The population concentrations roughly coincide with the more fertile and well-watered areas of the country, which account for about 18 % of the surface area of Kenya, leaving over 80 % of the land with a rainfall of less than 750 mm per annum.
This lack of rainfall acts as one of the greatest constraints on the development of huge areas of Kenya, and a major aim of Government is to improve the productivity of these arid and semi-arid areas through a programme of scientific range management. This will involve the phased replacement of traditional nomadism by commercial stock ranching. In addition to supporting pastoralists and their stock, this region also contains very large numbers of herbivorous wildlife species, ranging from the elephant (Loxodonta africana) to the very small dik-dik (Rhynochotragus spp.), which are distributed according to topography and availability of water and food. The Government is aware of the possible conflict of interests between these two major resources and the consequent need for sound long-term management practices. This has arisen with the change in attitude towards wildlife from the purely protective to one in which wildlife is to be utilized as a resource, either alone or in combination with domestic livestock.
Government has, therefore, decided to establish a Kenya Rangeland Ecological Monitoring Unit to determine the numbers, distribution and seasonal movements of domestic livestock (cattle, sheep, goats, camels, donkeys) and the major wildlife herbivores (about 15 species). This unit will also monitor climatic parameters, habitat features and changes in human land-use. The incoming quantitative data will be used to construct models for the various ecosystems and self-contained land units involved, and these in turn will be used to generate management policy and plans.
To establish the KREMU, Government has entered into a bilateral aid programme with the Canadian Government for the initiation and staffing of the KREMU for an initial 4-year period, after which it will be run entirely by Kenyan personnel. Additional expertise is being supplied by the united Nations Food and Agriculture Organisation.
The KREMU will be inter-ministry in composition and outlook and will make its findings available to all Government agencies. Pre-project activities (organisation, ground-unit delineation, etc.) have started; it is expected that routine sampling will commence early in 1976.
The KREMU will use methodology developed in East Africa during the last decade for small-scale (under 25,000 km²) habitat monitoring programmes and already outlined in this paper. This involves a systems approach to data collection and handling and the use of both field teams and regular systematic reconnaissance flights.
The KREMU will thus be amassing a large amount of ground truth data. Such data will also be used to develop methods for quantifying multi-spectral satellite images such that the resultant data can be used in the regular updating of the ecosystems models. If this proves successful it will mean a great saving in manpower and will allow the extension of the EMU concept to other areas that have an even more severe shortage of trained personnel than Kenya, e.g. parts of Sahelian Africa. In fact, with suitable technical modifications, any world ecosystem - marine or terrestrial - may be monitored using the EMU strategy.
The function of the KREMU is to describe in quantifiable terms what animals and what human land-uses occur on a specified area and how they are distributed in relation to each other and to measured and described parameters of the environment such as vegetation, surface water, rainfall, etc. Extending the descriptions over time will provide information on the dynamic aspects of the relationship. This is therefore, basically a systems approach to synecological community analysis. If properly established the KREMU will be able both to forecast and to detect the responses of the ecosystem to changes resulting from natural climatic extremes and from deliberate human-induced changes, and will enable practical land-use solutions to be generated quickly.
To obtain and process these data and produce the findings in a form useful to land management, the KREMU will take the following operational steps:
1. Identification of the sampling areas, wherever possible on ecological rather than administrative grounds, so that ideally they represent ecosystems of self-contained land-units.
2. Intensive collection of base-line and non-variant data within the sampling areas; in some cases this would lead to further stratification for sampling purposes.
3. Collection of variant data on environmental parameters, land-uses, and human, livestock and wildlife populations, using a 3-tier approach and techniques:a) Ground level, using field teams, ground situations, plots, transects, etc.
b) Underflight, using specially equipped and instrumented light aircraft.
c) Overflight, using remote sensing from ERTS-type satellites.
4. Analyses and interpretation of data to reveal the current status of the relationships between the variables and non-variables.
5. Development of a strategy of data storage, processing and recall to ensure rapid analysis and promulgation of monitoring data for prompt and efficient implementation of management plans.
6. Dissemination of these findings (4 & 5) to concerned operant agencies within Government.
It will be possible to relate incoming KREMU information to ERTS data in such a way that some of the required quantitative data can be generated directly from the images. The most important of these are:
1. Plant dry matter production on a seasonal basisa. Grass and fortes (the most important)
b. Woody plants (> 1 m tall)
3. Surface water
4. Stored soil moisture
These four are interrelated and must be worked on together. Other useful quantifiable variables to be obtained from ERTS images include:
5. Cultivation in pastoral areas
6. Human settlement distribution and areas
7. Soil and erosion patterns.
Pastoral Kenya will be divided into a number of ecosystems or self-contained land units, e.g. Amboseli, Samburu, Lamu-Garissa. The ecological zone maps of Kenya (Pratt, Greenway & Gwynne, 1966; Pratt and Gwynne, 1975, in press) will serve as a basis with sub-division on other existent base-line and research data. The first 4 months of the KREMU will include an underflight survey in light aircraft (ca. 625 furs) of the whole area to refine sub-division boundaries and to determine major-unit characteristics, and thus to allow determination of underflights and field team sampling frequencies.
1. Ground truth
Ground truth sampling plots (up to 250) on a 40 x 40 km grid will be located, marked and photographed from cat 10,000 feet. Plot size will be about 2 hectares, but the exact size and shape will depend on the length of the local catena; it is proposed to sub-sample the catena summit, mid-point, and sump at each plot site. lt is not envisaged that the catena length will extend beyond 7 km. Final number of installed plots will depend on the accessibility of the sites, the ecological zone, and the ability of the KREMU resources to handle the incoming data.
At each ground truth station there will be a basic site description (description of topography, land system/facet, and drainage including erosion); the establishment of soil pits and a description of the soil profile, physical properties and the catena; and a description of the initial plant cover and plant species composition. While this is being done plot markers, neutron probe tubes and storage rain gauges will be installed, and access tracks and air strips (remote areas only: each 400 m long) cleared.
Once a plot is established, monitoring of climatic and botanical parameters will commence. Plant composition will be measured periodically with respect to herbaceous/grass cover; standing crop, leaf table height; physiognomy and phonology; and plant part composition.
In addition, within each ecosystem or self-contained land unit the following will be obtained on a periodic basis, the frequency of which will be dependent on prevailing local conditions. Small samples of domestic stock and major wildlife herbivores will be slaughtered to provide data on body condition, reproductive state and diet. Ground estimates will be made on the population structure of both wild and domestic herbivores with respect, for example, to age, sex and herd composition.
Prior to starting routine survey, the following will take place:
a) Tests to measure: correction factors for bank and turbulence, correction factors for crab, best horizontal angle, effect of fixed versus free transect streamers, and best methods for low-level flight determination of plant greenness and cover.
b) Factor experiment to test variation in method by an analysis of variance involving the following factors: height, speed, time of day - lighting, time of day - turbulence, strip width, transect length, habitat, number of species counted, observer experience, and observer fatigue.
c) Observer screening programme with respect to visual acuity and propensity towards air sickness (cf. Savidge, 1973), and a training period involving method practice (briefing and inflight), estimating animal numbers, and an accuracy check with a known number of animals or models and/or against experienced observers.
Past experience has shown an air speed of 150 km/in, a height of cat 100 m and a strip width of 200-250 m to be the most suitable, but final choice for any one area will depend on the outcome of the pre-survey tests, the nature of the habitat, etc. Systematic sampling with transects 10 km or 5 km apart will be used initially, with accuracy of flight course assured by the use of long-wave radio navigation and altitude by radar altimeter.
In addition to obtaining precise and accurate estimates of animal numbers and densities, the underflights will record for each flight line sub-division (5 or 10 km lengths) the data categories already listed in " Methods ".
3. ERTS Data
We should emphasise at this point that the value of an EMU is not dependent on remote sensing data, but it is undoubtedly enhanced by them. The fact that the operational programmes we have reviewed have produced results without satellite imagery should make this clear. The point of ecological monitoring techniques is that they are relatively simple, inexpensive and productive. Therefore access to remote sensing data should not be a deciding factor in the question of establishment of an EMU. However, since ERTS coverage is global, there is no reason why any country should not take advantage of it.
a. Data receiving facility
The success of the ERTS segment of the Kenya rangeland monitoring activity depends on the ability of the research team to receive near-complete coverage of the pastoral areas of the country on a regular short interval basis; intervals longer than 18 days would reduce sensitivity in detecting short-term changes in vegetation.
To facilitate the Kenya ERTS programme the Kenya Government will probably implement one or both of the following courses of action:
a) Modify the existing receiving facility of the Italian San Marco Project satellite launching station at Malindi on the Kenya coast. To enable this station to receive transmission from ERTS additional specialised equipment will be needed, the specifications of which will depend upon whether the incoming data are to be processed by NASA in the normal way, by the Italians at their Rome space centre, or by international agencies at some other space centre to be established in Africa (e.g. at Kinshasa). This will result in an ERTS station capable of receiving spectral data relating to an area of eastern Africa of a radius of 1,200 km centred on Malindi;
b) To establish a new international ERTS data receiving and processing centre at Nairobi which will cover, in addition to Kenya, an area of Africa of a radius of 2,775 km centred on Nairobi.
b. Data handling plan
Incoming 70 mm black and white positive transparencies or "chips" will be examined visually to elucidate both permanent and transient features. For this purpose correlative ground truth will be available from the regular underflight and field team activity of the KREMU.
Suitable cloud-free imagery of good definition will be used to develop permanent feature maps for pastoral Kenya at a scale of 1: 500,000 or 1: 1 million, showing land system boundaries, drainage systems and surface soil complexes.
Chips of appropriate spectral bands will be examined regularly to monitor short-term changes in such habitat variables of economic importance as vegetation growth flush, large surface water pools and local flooding, and grass and woodland burning. A synoptic view of these features will be of value to agriculture, rangeland utilization and tourism. Data in the form of distribution maps wild, therefore, be constructed. For example, in these remote areas of low and erratic rainfall, a sudden relatively localised vegetation flush usually indicates local rainfall; such information passed promptly to the Ministry of Agriculture will allow its Livestock Marketing Division to alter its stock routes accordingly, with consequent financial savings.
Imagery will also be examined to determine long-term changes within the pastoral areas, such as encroachment of agriculture activities, irrigation scheme development, burning/removal of forest stands, development of forestry plantations, and any other detectable major change in land-use.
Finally, imagery will be used to determine visually the coordinates of particular test areas in regions of interest, such as localised growth flushes, so that they can be more quickly identified on the computer compatible tapes.
ii. Digital data
Visual interpretation of ERTS colour composites will be used to determine areas of active plant growth on or near one or more of the test sites of the Kenya Rangeland Ecological Monitoring Unit. Routine ground truth acquisition from these sites will supply quantitative plant production data on a phenological and a plant composition basis (dry weight, part, crude protein, crude fibre, energy).
Spectral reflectance measurements of the chosen test areas (size to be determined in the field) will be made from the ground throughout the development period and these values graphed against time. Elsewhere it has been found that the ratio of spectral reflectance values associated with wavelength bands 8,000 and 6,750 Å shows a high degree of correlation with changes in the phenology and condition (i.e. greenness/dryness). Attempts will be made, therefore, to see whether similar relationships hold for tropical semi-arid rangelands and whether the plant productivity curves are similarly correlated. If they are, tests will be made to see whether the irradiance spectral curves can be used to estimate plant biomass (standing crop) as total vegetation and (with suitable ground sampling) in terms of plant parts (leaves, stems, etc.).
ERTS irradiance data will then be extracted from the computer compatible tapes for the test sites, the spectral values being obtained for each individual picture element (80 X 80 m on the ground) in each of the four spectral bands. These values, plotted against time, will be compared with phenology, condition and standing crop values, and with the ground-obtained irradiation curves.
If the results are satisfactory the possibility of using ERTS irradiation data as a means of determining plant standing crop directly will be further tested, and a computer programme developed for determining the standing crop values directly from the NASA-EATS computer compatible tape. These output data will then be supplied directly to the data centre of the Kenya Rangeland Ecological Monitoring Unit for routine incorporation in its ecosystem models.
Initial programmes for locating specific areas on the computer compatible data tapes and extracting the individual picture element spectral values for each spectral band have already been developed (e.g. at the Space Science Laboratory of the University of California; Remote Sensing Centre, Texas A and M University; Oregon State University). Further computer programme development will be done with the cooperation of the Computing Centre of the University of Nairobi, using both University and Kenya Government computers and with possible aid from overseas facilities.
Again we would emphasise that it is possible to produce the above without access to satellite-collected data - it would cost more and take considerably more time. If, however, the ground productivity/ERTS spectral value correlates can be made, the productivity of an entire country could be monitored at a resolution and frequency never before imagined.
D. Data handling by computer
1. Storage and analysis
The vast amounts of data collected by KREMU can only be processed with the help of a digital computer. There are several in operation in Nairobi (KREMU headquarters), varying in size from an ICL 1902 A with 16 K capacity to an ICL 1905 F with 112 K capacity. Data filing programmes will have to be written using standard file packages as a basis. At least two programmes are locally available for initial treatment (plotting numbers and density estimates) of SRF data (e.g. ANPR of Cobb and Western, unpublished). Finally, there are standard statistical packages for various multivariate analytical techniques. In short, if one has access to the hardware, the basic components of the software already exist. They need, of course to be tailored into a working package for the KREMU's specific purposes.
2. Predictive models
From recent East African research, there are already enough inductive data to construct a first-generation semi-arid ecosystem simulation model (cf. Jeffers, 1972, especially pages 249-344). A model is simply a set of reasonable rules for complex processes, such as the workings of an ecosystem, which are expressed mathematically so that work can be done by computer. The computer is primed with the rules of the model, variables and constants, all consistent with the data already collected, and asked to process them according to the rules. The state at the end of the process is a prediction, which may then be compared to the real-world ecosystem that we are monitoring. If agreement is good, then it is likely that the model simulates the processes of the real ecosystem. Models are rarely good ones the first time round and must be continually updated with new monitoring data and intelligent guesses. Once we have a model that gives predictions of maximum likelihood, we may then begin to make management predictions. If, for example, the model predicts increased and sustained productivity as a result of a particular management action (equivalent to an experimental change in the variable input to the model), then that course may be worth pursuing. If, however, the model predicts an ecological disaster, we may think again and try out an alternative.
III. Practical applications of ecological monitoring
So far we have stressed the potential uses of systematic ecological monitoring; we will now give examples of how ecological data have been put to use by management agencies in tropical ecosystems. The examples are admittedly few, if only because the art is new.
A. Management plans for national parks
Ecological research organisations frequently find a niche in a national park, viz: the Uganda Institute of Ecology in the Ruwenzori National Park, Uganda (formerly the Nuffield Unit of Tropical Animal Ecology in Queen Elizabeth National Park, Uganda); the Serengeti Research Institute in Serengeti National Park, Tanzania; and the Tsavo Research Project in Tsavo National Park East, Kenya. National parks are able to afford such institutions because they are relatively well funded from overseas donations and tourist spending. Moreover, there is a recognized need for ecological data from which to plan management of the parks (cf. Huxley, 1962, Starker-Leopold, 1970).
As a result of ecological research, the Serengeti Research Institute drafted management proposals for the Serengeti National Park. The topics covered included: ecological zone development (Kruuk, 1970), road development (Braun, 1970), fire control (Norton-Griffiths, 1970), and elephant problems (Croze, 1970). These reports are complemented with statements from management personnel on poaching (Turner, 1970) and tourist lodge development (Owen, 1970). Thus within one year of the inception of the Serengeti Ecological Monitoring Programme, operational management suggestions were produced.
More recently, as a consequence of the monitoring and research activities of Western (1973), comprehensive policies for management and development have been implemented by the Kenya National Parks for the Amboseli National Park area (Western and Thresher, 1973; Western, 1974). Similarly, the boundaries for the proposed new County Council Game Reserves now being considered in the Lamu-Garissa area of Kenya are based on data gained during SRF and monitoring activities (Gwynne and Smith, 1974 a-c). The more data that are collected, the finer will be the detail of the management plans and the more precisely will National Parks be able to predict the outcome of any policy they may eventually adopt. Similar relationships hold for other development enterprises such as ranching.
As we have shown, ecological monitoring is a powerful tool for development and resource assessment purposes. It is not, however, a universal panacea for all the development problems of rangelands and must be used with care and common sense, not followed blindly. In monitoring it is often, for example, as bad to collect too much data as not to collect enough (Cobb, 1975). Monitoring leans heavily on mathematics in the treatment of data but the mathematics can only be as good as the reliability of the data collected and the ability of the EMU to handle them. It is worth remembering the words of the eminent statistician G.E. Yule (1920):
" If you get on the wrong track with the mathematics for your guide, the only result is that you get to the Valley of Mare's Nests much quicker; get there so smoothly and easily that you do not realize where you are and it may be hard to unbeguile you. Logic and mathematics are only of service, then, once you have found the right track; and to find the right track you must exercise faculties quite other than the logical. Observation and Fancy, and Imagination: accurate observation, riotous fancy, and detailed and precise imagination."
Note: We are indebted to Vincent Quin, Librarian of Balliol College, Oxford for verifying some bibliographic details, and to Mrs. Elma Sayer for her care in typing the manusicript
1. BALE, J.B., CONTE, D., GOEHRING, D. and SIMONETT, D.S. - Remote sensing applications to resource management problems in the Sahel 1974. Prepared for USAID. Washington, D.C.: Earth Satellite Corporation.
2. BECKETT, P.H.T. and WEBSTER, R. - A classification system for terrain. 1965. Milit. Engng Expl. Establ, 872: 1-29. Christchurch, Hampshire: Milit. Engng Expl. Establ.
3. BELL, J.P. - Neutron probe practice. 1972. Rep. Inst. Hydrol. U.K., 19: 1-63.
4. BELL, R.H.V. - A grazing ecosystem in the Serengeti. 1971. Scient. Am., 224 (1): 86-93.
5. BELL, R.H.V., GRIMSDELL, J.J.R., VAN LAVIEREN, L.P. and SAYER, J.A. - Census of the Kafue lechwe by aerial stratified sampling. 1973. E. Afr. Wildl. J., 11: 55-74.
6. BRAWN, H.M.H. - Draft statement on the management policy to be adopted in the Serengeti National Park with regard to roads. 1970. Typescript pp. 5. Arusha, Tanzania: Serengeti Research Institute.
7. CARNEGGIE, D.M. and DE GLORIA, S.D. - Determining range condition and forage production potential in California from ERTS - 1 imagery. 1974. Proc. int. Symp. remote Sens. Environ. 9. Ann Arbor: University of Michigan.
8. CAUGHLEY, G. - Aerial survey techniques appropriate to estimating cropping quotas. 1972. KEN: SF/FAD 26 Work. Pap. 2: 1-13. Nairobi: Food and Agriculture Organization of the United Nations.
9. CENTRO. - Seminar on the application of remote sensors in the determination of natural resources 10-13 November 1971. 1971, Ankara: CENTRO.
10. COBB, S. - Preliminary results of the aerial monitoring programme in the Tsavo region. 1975. Cyclostyled pp. 23. Oxford: Animal Ecology Research Group.
11. COLVOCORESSES, A.P. - Remote sensing platforms. 1974. Circ. U.S. geol. Surv., 693: 1-75.
12. COLWELL, R.N. (ed.). - Monitoring earth resources from aircraft and spacecraft. 1971. NASA SP-275, pp. 170. Washington, D.C.: Scientific and Technical Information Office, NASA.
13. COOPER, C.F. - An evaluation of variable plot in shrub and herbaceous vegetation. 1963. Ecology, 44: 565-569.
14. CROZE, H. - Draft statement on the management policy to be adopted in Serengeti National Park with regard to elephants. 1970. Typescript pp. 3. Arusha, Tanzania: Serengeti Research Institute.
15. CROZE, H. - A modified photogrammetric technique for assessing age-structures of elephant populations and its use in Kidepo National Park. 1972. E. Afr. Wildl. J., 10: 91-115.
16. CROZE, H. - Report on an aerial reconnaissance of parts of north-eastern Kenya, 5-13 September 1973. 1973. Report to the Kenya Game Department Typescript pp. 13. Nairobi: University of Nairobi.
17. CROZE H. - The Seronera bull problem. II. The Trees. 1974. E. Afr. Wildl. J., 12: 29-47.
18. CROZE, H., GWYNNE, M.D. and JARMAN, P.J. - A proposal for a global ecological monitoring programme with initial emphasis on arid and semi-arid lands. 1973. Cyclostyled pp. 10. Nairobi: UNDP/FAO Habitat Utilization Project.
19. DIX, R.L. An application of the Point-Centered Quarter method to the sampling of grassland vegetation. 1961. J. Range Mgmt, 14: 63-69.
20. DUNCAN, P. - The ecology of the topi. 1974 Ph. D. thesis. Nairobi: University of Nairobi.
21. DUNCAN, P., GWYNNE, M.D. and JARMAN, P.J. - The status of the Lamu-Garissa topi population. 1975. Parts I - IV (in press).
22. EHRLICH, P.R. - The population bomb. 1968. New York: Sierra Club/Ballantine.
23. FAO. - Rangeland surveys in Kenya. 1967-1973. A series of 13 reports covering a total of 174,000 km2 of the country. Nairobi: UNDP/FAO Kenya Range Management Project.
24. FAO. - Kenya rangeland surveys. 1973. AGP: SF/KEN 11 Tech. Rep. 5: 1-38. Rome: Food and Agriculture Organization of the United Nations.
25. FIELD, C.R. - Elephant ecology in the Queen Elizabeth National Park, Uganda. 1971. E. Afr. Wildl. J., 9: 99-123.
26. FIELD, C.R. - The food habits of wild ungulates in Uganda by analysis of stomach contents. 1972. E. Afr. Wildl. J., 10: 1742.
27. FOLLIS, T.B. and SPILLETT, J.J. - A new method for rumen sampling. 1972. J. Wildl. Mgmt, 36: 1336-1340.
28. GERRESHEIM, K. - The Serengeti landscape classification. 1974. Nairobi: Serengeti Ecological Monitoring Programme/African Wildlife Leadership Foundation.
29. GITTINS, R. - Trend-surface analysis of ecological data. J. Ecol., 56: 845-869.
30. GLOVER, J., ROBINSON, P. and HENDERSON, J.P. - Provisional maps of the reliability of annual rainfall in East Africa 1954. Q J1 R. met. Sac., 88: 602-609.
31. GODDARD, J. - Food preferences of black rhinoceros in the Tsavo National Park. 1970. E. Afr. Wildl. J., 8: 145-161.
32. GOODALL, D.W. - Some considerations m the use of point quadrats for the analysis of vegetation. 1952. Aust. J. scient. Res., Ser. B, 5: 141.
33. GRIMSDELL, J.J.R. - Age determination of the African buffalo, Syncerus caffer Sparrman. 1973. E. Afr. Wildl. J., 11: 31-53.
34. GRIMSDELL, J.J.R. - Ecological monitoring, in Handbook of african wildlife ecology, edit. by J.J.R. Grimsdell and H. Russell. 1975. Nairobi: African Wildlife Leadership Foundation.
35. GWYNNE, M.D. - Plant physiology and the future in Tropical pastures, 1966, edit. by W. Davies and C.L. Skidmore. London: Faber and Faber.
36. GWYNNE, M.D. - Selective food intake by some East African herbivores. 1971. AGP: SF/KEN 11 Work Pap. Nairobi: UNDP/FAO Kenya Range Management Project.
37. GWYNNE, M.D. - The use of video-tape recorders in aerial census and monitoring programmed 1972. Typescript pp. 4. Nairobi: UNDP/FAO Kenya Range Management Project.
38. GWYNNE, M.D. - Habitat plant/animal relations in Range ecology, livestock production and wildlife ecology research. 1974. AGP: SF/KEN 11 Tech Rep. 4. Rome: UNDP and FAO.
39. GWYNNE, M.D. and BELL, R.H.V. - Selection of vegetation components by grazing ungulates in the Serengeti National Park. 1968. Nature, Land., 220: 390-393.
40. GWYNNE, M.D. and ROBERTSHAW, D. - The effects of solar radiation load on feeding and drinking behaviour of Sahiwal cattle in Kenya (in press), 1976.
41. GWYNNE, M.D. and SMITH, K. - Proposals for the Kiunga Marine National Park and Reserve. 1974 a. Cyclostyled pp. 13. Nairobi: Kenya Game Department.
42. GWYNNE, M.D. and SMITH, K. - Proposals for the Dodori River Game Reserve. 1974 b. Cyclostyled pp. 11. Nairobi: Kenya Game Department.
43. GWYNNE, M.D. and SMITH, K. - Proposals for the Boni Forest Game Reserve. 1974 c. Cyclostyled pp. 12. Nairobi: Kenya Game Department.
44. HEYTING, A. - Discussion and development of the Point-Centered Quarter method of sampling grassland vegetation. 1968. J. Range Mgmt, 21: 370-380.
45. H.M.S.O. - Handbook of meteorological instruments. 1956. London: Her Majesty's Stationery Office.
46. HUXLEY, J.S. - Eastern Africa: the ecological base. 1962. Endeavour, 21 (82) : 98-107.
47. JARMAN, P.J. - Diets of large mammals in the woodlands around Lake Kariba, Rhodesia. 1971. Oecologia (Bert.), 8: 157-178.
48. JARMAN, P.J. and GWYNNE, M.D. - Feeding ecology of impala. 1976 (in press).
49. JEFFERS J.N.R. fed.). - Mathematical models in ecology. 1972. Oxford: Blackwell Scientific Publications.
50. JOLLY, G.M. - Sampling methods for aerial censuses of wildlife populations. 1969 a. E. Afr. agric. For. J., 34 (Special Issue): 4649.
51. JOLLY, G.M. - Treatment of errors in aerial counts of wildlife populations. 1969 b. E. Afr. agric. For. J., 34 (Special Issue): 50-55.
52. KREULEN, D. - Wildebeest habitat selection on the Serengeti Plains, Tanzania, in relation to calcium and lactation. 1975 (in press).
53. KRUUK, H. - Development in different areas of the Serengeti National Park: a draft proposal for management. 1970. Typescript pp. 2. Arusha, Tanzania: Serengeti Research Institute.
54. KRUUK, H. - The spotted hyaena. 1972. Chicago: University of Chicago Press.
55. Lamprey, H.F. - Ecological separation of the large mammal species in the Tarangire Game Reserve, Tanganyika. E. Afr. Wildl. J., 1 : 63-92.
56. LAWS, R.M. - Age criteria for the African elephant, Loxodonta a. africana. 1966. E. Afr. Wildl. J., 4: 1-37.
57. LAWS, R.M. - Aspects of reproduction in the African elephant, Loxodonta africana. 1969. J. Reprod. Fort. Suppl., 6: 193-217.
58. LEESE, M.N., STRANGEWAYS, I.C. and TEMPLEMAN, R.F. - Field performance of Institute of Hydrology automatic weather station. 1973. Rep. Inst. Hydrol. U.K., 21: 1-33.
59. LEUTHOLD, W. - Preliminary observations on the food habits of gerenuk in Tsavo National Park, Kenya. 1970. E. Afr. Wildl. J., 8: 73-84.
60. McKAY, A.D. and FRANDSEN, P.E. - Chemical and floristic components of the diet of zebu cattle (bos indicus) in browse-grass range pastures in a semiarid upland area of Kenya. Trop. Agric., Trin., 46 : 279-292.
61. MILNER, C. and HUGHES, R.E. - Methods for measurement of the primary production of grassland. 1968. IPB Handbook No. 6. Oxford: Blackwell Scientific Publications.
62. MONTEITH, J.L. - Survey of instruments for micrometeorology. 1972. IBP Handbook No. 22. Oxford: Blackwell Scientific Publications.
63. MUNN, R.E. - Biometeorological methods. New York and London: Academic Press.
64. NEWBOULD, P.J. - Methods for estimating the primary production of forests. 1967. IBP Handbook No. 2. Oxford: Blackwell Scientific Publications.
65. NORTON-GRIFFITHS, M. - Draft statement on the management policy to be adopted in the Serengeti National Park with respect to fire control. 1970. Typescript pp. 2. Arusha, Tanzania: Serengeti Research Institute.
66. NORTON-GRIFFITHS, M. - Serengeti ecological monitoring program. 1972. Nairobi: African Wildlife Leadership Foundation.
67. NORTON-GRIFFITHS, M. - Counting the Serengeti migratory wildebeest using two-stage sampling. 1973. E. Afr. Wildl. J., 11: 135-149.
68. NORTON-GRIFFITHS, M. - Reducing counting bias in aerial censuses by photography. 1974. E. Afr. Wildl. J., 12: 245-248.
69. NORTON-GRIFFITHS, M. - The numbers and distribution of large mammals in Ruaha National Park, Tanzania. 1975. E. Afr. Wild l. J., 13 (in press).
70. NORTON-GRIFFITHS, M., BUNNING, J. and KURJI, F. - Woodland changes in the Serengeti ecosystem. 1973. Typescript pp. 5. Arusha, Tanzania: Serengeti Research Institute.
71. NORTON-GRIFFITHS, M., HERLOCKER, D. and PENNYCUICK, L. - Patterns of rainfall in the Serengeti ecosystem. 1975. E. Afr. Wild l. J., 13 (in press).
72. NORTON-GRIFFITHS, M. and PENNYCUICK, L. - Trend surface analysis (MX23). 1974. Tech. Pap. Inst. Dev. Stud. Nairobi, 8: 1-10.
73. OWEN, J.W. - Draft statement on the management policy to be adopted in the Serengeti National Park with regard to tourist lodges. 1970. Typescript pp. 5. Arusha, Tanzania: Serengeti Research Institute.
74. PEARSON R.L. and Miller, L.D. - The biometer: a hand-held grassland biomass meter. 1973. Incid. Rep. Cola. State Univ., 6: 1-24.
75. PENNYCUICK, C.J. - Methods of using light aircraft in wildlife biology. 1969. E. Afr. agric. For. J., 34 (Special Issue): 24-29.
76. PENNYCUICK, C.J. - The shadowmeter: a simple device for controlling an aircraft's height above the ground. 1973. E. Afr. Wildl. J., 11: 109-112.
77. PEREIRA, H.C. - A cylindrical gypsum block for moisture studies in deep soils. 1951. J. Soil Sci., 2: 212-223.
78. PEREIRA, H.C. and HOSEGOOD, P.H. - Comparative water-use of softwood plantations and bamboo forest. 1962. J. Soil Sci., 13: 299-313.
79. PRATT., D.J., GREENWAY, P.J. and GWYNNE, M.D. - A classification of East African rangeland, with an appendix on terminology. 1966. J. appl. Ecot., 3: 369-382.
80. PRATT, D.J. and GWYNNE, M.D. feds.). - Rangeland ecology and management in East Africa. 1975. London: University of London Press.
81. RABCHEVSKY, G. - Remote sensing of the earth's surface. 1970. J. remote Sens., 4: 1-32.
82. RAINY, M.E. - Research proposal: a study of the ecology of Samburu pastoralism. 1969. Typescript pp. 16. Nairobi: University of Nairobi.
83. RIJKS, D.A. and OWEN, W.G. - Hydrometeorological records from areas of potential agricultural development in Uganda. 1965. Entebbe: Ministry of Mineral and Water Resources.
84. RODGERS, W.A. - The lion (Panthera leo) population in the eastern Selous Game Reserve. 1974. E. Afr. Wildl. J., 12: 313-317.
85. ROUSE, J.W. and RITER, S. - ERTS experiments. 1973. Trans. Geosci. Electronics GE-11 (I): 3-77.
86. SAVIDGE, J. - Aerial census techniques in estimating wildlife populations. 1973. Typescript pp. 6. Nairobi: UNDP/FAO Kenya Wildlife Management Project.
87. SCHALLER, G. - The Serengeti lion. 1972. Chicago: University of Chicago Press.
88. SCOTT, R.M., WEBSTER, R. and LAWRANCE, C.J. - A land system atlas of western Kenya. 1971. Christchurch, Hampshire: University of Oxford.
89. SINCLAIR, A.R.E. - Serial photographic methods for population, age and sex structure. 1969. E. Afr. agric. For. J., 34 (Special Issue): 87-93.
90. SINCLAIR, A.R.E. - Long term monitoring of mammal populations in the Serengeti: census of non-migratory ungulates, 1971. 1972. E. Afr. Wildl. J., 10: J., 11: 307-316.
91. SINCLAIR, A.R.E. - Population increases of buffalo and wildebeest in the Serengeti. 1973 a. E. Afr. Wildl. J., 11: 93-107.
92. SINCLAIR, A.R.E. - Regulation and population models for a tropical ruminant. 1973 b. E. Afr. Wildl. J., 11: 307-316.
93. SINCLAIR, A.R.E. and DUNCAN, P. - Indices of condition in tropical ruminants. 1972. E. Afr. Wildl. J., 10: 143-149.
94. SINCLAIR, A.R.E. and GWYNNE, M.D. - Food selection and competition in the East African buffalo (Syncerus caffer Sparrman). 1972. E. Afr. Wildl. J., 10: 77-89.
95. Starker-Leopold, A. - Research policy in the National Parks of Tanzania. 1970. Cyclostyled pp. 18. Arusha: Tanzania National Parks.
96. STEWART, D.R.M. - Analysis of plant epidermis in faeces: a technique for studying the food preference of grazing herbivores. 1967. J. appl. Ecol., 4: 83-111.
97. SWANK, W.G., WATSON, R.M., FREEMAN, G.H. and JONES, T. (eds.). - Proceedings of the workshop on the use of light aircraft in wildlife management in East Africa. 1969. E. Afr. agric. For. J., 34 (Special Issue): 1-111.
98. THORNTON, R.F. and MINSON, D.J. - The relationship between apparent retention time in the rumen voluntary intake, and apparent digestibility of legume and grass diets in sheep. 1973. Aust. J. agric. Res., 24: 889-898.
99. TUCKER, C.J., MILLER, L.D. and PEARSON, R.L. - Measurement of the combined effects of green biomass, chlorophyll, and leaf water on canopy spectro-reflectance of the short grass prairie. 1973. Proc. Res. Cont. remote Sens. Earth 2; pp. 27. Tullahoma : University of Tennessee.
100. TURNER, M.I.S. - Draft of management policy regarding poaching in the Serengeti. 1970. Typescript pp. 2. Arusha, Tanzania: Serengeti Research Institute.
101. VAN DYNE, G.M. - The ecosystem concept in natural resources management. 1969. New York: Academic Press.
102. VAN DYNE, G.M. - An integrated ecological research programme, in Mathematical models in ecology, edit. by J.N.A. Jeffers, 1972. Oxford: Blackwell Scientific Publications.
103. VAN DYNE, G.M. and Torrell, D.T. - Development and use of esophageal fistula: a review. 1964. J. Range Mgmt, 17: 7-19.
104 VESEY-FITZGERALD, D.F. - Utilization of the habitat by buffalo in Lake Manyara National Park. 1969. E. Afr. Wildl. J., 7: 131- 145.
105. VESEY-FITZGERALD, D.F. - Browse production and utilization in Tarangire National Park. E. Atr. Wildl. J., 11: 291-305.
106. VESEY-FITZGERALD, D.F. - Utilization of the grazing resources by buffaloes in the Arusha National Park, Tanzania. 1974. E. Afr. Wildl. J., 12: 107-134.
107. WARREN-WILSON, J. - Inclined point quadrats. 1960. New. Phytol., 59: 1-8.
108. WARREN-WILSON, J. - Estimation of foliage denseness and foliage angle by inclined point quadrats. 1963. Aust. J. Bat., 11: 95-105.
109. WATSON, R.M. - The population ecology of the wildebeest (Connochaetes taurinus albojubatus Thomas) in the Serengeti. 1967. Ph. D. thesis. Cambridge: University of Cambridge.
110. WATSON, R.M. - Aerial photographic methods in censuses of animals. 1969 a. E. Afr. agric. For. J., 34 (Special Issue): 32-37.
111. WATSON, R.M. - The planning of flights and the handling of time-serial data. 1969 b. E. Afr. agric. For. J., 34 (Special Issue): 70-78.
112. WATSON, R.M. - The south Turkana Expedition papers II. A survey of the large mammal population of South Turkana. 1969 c. Geogrl. J. 135: 529-546.
113. WATSON, R.M. - Results of aerial livestock surveys of Kaputei Division, Samburu District and Northeastern Province. 1972. Nairobi: Statistics Division, Ministry of Finance and Planning.
114. WESTERN, D. - The structure and dynamics of the Amboseli ecosystem. 1973. Ph.D. thesis. Nairobi: University of Nairobi.
115. WESTERN, D. - Road development plans for Amboseli National Park. 1974. Typescript. Nairobi: Kenya National Parks.
116. WESTERN, D. and THRESHER, P. - Development plans for Amboseli. 1973. Cyclostyled pp. 97. Nairobi: International Bank for Reconstruction and Development.
117. W.M.O. - Guide to hydrometeorological practices. 1965. Geneva: World Meteorological Organization.
118. WOODHEAD, T. - Studies of potential evaporation in Kenya. 1968 a. Nairobi: Water Development Department and East African Agriculture and Forestry Research Organization.
119. WOODHEAD, T. - Studies of potential evaporation in Tanzania. 1968 b. Dar-es-Salaam: Water Development and Irrigation Division and East African Agriculture and Forestry Research Organization.
120. YULE, G.E. - The wind bloweth where it listeth. 1920. Cambr. Rev., 41: 184-186.
121. ZAPHIRO, D. - The use of light aircraft to count game. 1959. Wild Life, Nairobi, 1 (4): 31-36.
Fig. 1: Map of East Africa showing the areas which have been studied using Systematic Monitoring Techniques. A - C were examined predominantly by Systematic Ground Survey (SOS) supplemented with aerial sample or total counts of animals. 1-8 were monitored using Systematic reconnaissance Flights (SRF) supplemented with ground studies and interpretation of the Earth Resources Technology Satellite (ERTS) imagery.
A. Selous Game Reserve. Synecological survey study for the Tanzania Game Division by Rodgers: Department of Zoology, University of Nairobi; 1967 to present.
B. Amboseli Game Reserve (now National Park). Synecological study of the structure and dynamics of the Amboseli basin by Western (1973): Department of Zoology, University of Nairobi; 1968 to present.
C. Samburu. Study of the ecology of Samburu pastoralism by Rainy: Department of Zoology, University of Nairobi; 1970 to present.
1. Serengeti ecosystem. The Ecological Monitoring Programme of the Serengeti Research Institute (cf. Norton-Griffiths 1972). Multi-disciplinary studies -1964 to present; SRF - 1969 to 1972.
2. Ruaha National Park. SRF for large mammal numbers and distribution (cf. Norton-Griffiths, 1975, in press): Serengeti Research Institute; 1972 to 1973.
3. Lamu/Southern Garissa District. SRF by the UNDP/FAO Kenya Range Management Project, UNDP/FAO Habitat Utilization Project, Kenya Game Department and University of Nairobi, combined with ground sampling and ERTS correlations (cf. Duncan, Gwynne, Jarman, in press): 1973 to present.
4. Southern Kajiado. Ilkisongo Monitoring Project: an extension of (B) above to include the ecology of the entire Amboseli National Park ecosystem, including Maasai pastoralists; SRF and ERTS correlations; by Western and Croze; Department of Zoology, University of Nairobi and New York Zoological Society; 1973 to present.
5. Kajiado District. Investigations of land-use potentials with reference to wildlife populations; SRF and some ground sampling; UNDP/FAO Kenya Wildlife Management Project; 1974 to present.
6. Tsavo National Park ecosystem. Ecological monitoring by Cobb (1975) of the Tsavo Research Project; SRF and SGS: Kenya National Parks; 1973 to 1974.
7. East Rudolf National Park. SRF for Kenya National Parks: UNDP/FAO Kenya Range Management Project (Gwynne, Norton-Griffiths and Duncan); 1973.
8. Tana River/Kilifi Districts. SRF predominantly for elephant data by Allaway (under the supervision of Croze); Department of Zoology, University of Nairobi and the Research Division, Kenya Game Department; 1974 to present.
Fig. 2a and b: 10 cm isohyets for the mean dry season (a) and mean wet season (b) in the Serengeti National Park, Tanzania, and the surrounding ecosystem. Spots show the location of monthly storage rain gauges; stippling indicates higher ground (1,800- 1,500 m). From Norton-Griffiths et al. (1975).
Fig. 3: Time per bite as affected by grass height. Goat-c consistently take a longer time per bite and can be considered less efficient at grazing. From Gwynne (1974).
Fig. 4: Elephant use (any form of browsing) of five tree species expressed as a function of the frequency of each of the tree species recorded in ground sample plots. Datum points above the 5 % chi-square probability line (+) indicate species used more than expected, i.e. preferred; points below (-), less than expected, i.e. avoided. Solid circles - Acacia senegal solid triangles - Acacia tortilis; circles Commiphora trothe; triangles - Albizia harveyi; squares - Balanites aegyptiaca; crosses - Acacia clavigera. From Croze (1974).
Fig. 5: Actual dry season distribution of topi (a) and Hunter's antelope (b) in the Lamu - southern Garissa area of Kenya; data obtained during a single SRF survey flight. From Duncan, Gwynne and Jarman (1976).
Fig. 6: Broad scale SRF generated vegetation map of the Lamu-southern Garissa area of Kenya. Animal distribution in Fig. 5 should be compared with this map.
A= dry Acacia dominated bushland
E = woodland with grassland glades
F = closed canopy woodland with no grassland
H = woodland with abundant Hyphaene palms
From Duncan, Gwynne and Jarman (1976).
Fig. 7: The wet season (a) and dry season (b) surfaces of best fit (P < 0.01) for the occurrence of Hunter's antelope in the Lamu- southern Garissa area of Kenya. From Duncan, Gwynne and Jarman (1976).
Fig. 8: The wet season (a) and dry season (b) surfaces of best fit (P < 0.01) for cattle distribution in the Lamu-southern Garissa area of Kenya. The distribution of cattle is relatively static as compared with seasonally mobile species such as the topi and Hunter's antelope (cf. Fig. 7). From Duncan, Gwynne and Jarman (1976).
Fig. 9: Livestock routes in the Lamu - southern Garissa area of Kenya as recorded during SRF surveys. (Gwynne, unpublished data.)
Fig. 10: Cattle trail showing the characteristic parallel ribbed appearance which can readily be identified from the air during SRF (photo Gwynne).
Fig. 11: Preliminary vegetation map from a SRF in southern Kajiado District, Kenya. Identified units are:
Fig. 12: Green biomass isopleths (20 gm m-2) in the Serengeti National Park, Tanzania, in the early wet season (a) and the dry season (b). Areas of > 60 gm m-2 are indicated by hatching. Data collected from quasi-systematic flights using a digital photometer. (After MacNaughton, unpublished data.)
Fig. 13: Quartic surface of dry season elephant distribution in Samburu District, Kenya, from a trend surface analysis of data collected during a non-systematic reconnaissance flight. Isophant lines indicate elephants observed per 10 x 10 km grid square. From Croze (1973).
Fig. 14: Population parameters from vertical aerial photography. Population age-structures may be calculated from measurements of elephant body lengths made on the photographs. Black inset: vertical axis is frequency; horizontal axis is age (1-60) in years. After Croze (1972).
Fig. 15: Aerial photographs showing change in woody vegetation cover due to the action of fire and elephants, over a nine-year period in the Serengeti National Park, Tanzania. After Gerrisheim in Norton-Griffiths (1972).
Fig. 16: A portion of soil map (a) of the Lamu - southern Garissa area of Kenya prepared from ERTS imagery using both colour composites and single band images. This map may be compared with the ERTS MSS Band 5 image shown (b). (Gwynne, unpublished data.) See colour section at back of book.
Fig. 17: Estimates of grass greenness in the Southern Kajiado District of Kenya from (a) Systematic Reconnaissance Flight and (b) ERTS imagery:
Fig. 18: Distribution of Maasai cattle in Southern Kajiado District of Kenya, from a Systematic Reconnaissance Flight (cf. simultaneous grass greenness of Fig. 17).