Rapid Vegetation Survey
“Best bet” Land-use Systems
Thematic reports
Impact of different land uses on biodiversity
An Intensive Biodiversity Baseline Study in Jambi Province,Central Sumatra, Indonesia
Unique id: 2
Source file: D:\Projects\ASB\ASB Country and Thematic reports - xml\Above ground biodiversity assessmet WG\C-Sec1-3.xml
Authors: A.N. Gillison
--------------------------------------------------------------------------
Introduction:
Evidence for the need to conserve biodiversity is well
established in the literature and is reflected in the international Convention
on Biological Diversity that has addressed a series of issues for attention by
its signatories (CBD 1994). Despite the
agreed urgency to develop a framework for biodiversity conservation, there is,
as yet, no operational definition for biodiversity. According to Weitzman
(1995), the implementation of any plan to preserve biodiversity is hampered by
the lack of an operational framework or an objective function, and “We need a more-or-less consistent and
useable measure of the value of biodiversity that can tell us how to trade off
one form of diversity against another”.
Miller and Lanou, (1995) also maintain
“The value of biodiversity is
determined largely by the interaction between human society and biodiversity”. This implies that among other things, there
should be a dynamic link between biodiversity and productivity for human needs.
The World Bank (1995) asserts it is necessary to integrate biodiversity
concerns into national decision making, but the mechanisms for this remain
elusive. In
These pressures highlight both the need for a working
definition of biodiversity and a cost-efficient, generic tool for its
assessment that can be used in turn to inform policy planners and managers.
While the species remains the sole currency unit for biodiversity assessment
(Heywood and Baste 1995) there will be
little progress (cf. Wulff
1943). Species richness and abundance
used alone and without other attributes of behaviour and performance can
seriously misinform and impede biodiversity assessment. Parity in richness does
not guarantee equivalence in either genetic composition or response to
environment. Partly for this reason, an emerging school of thought now
considers assessment should include functional features or types as well as
species. (Box 1981, Gillison 1981, 1988,
Nix and Gillison 1985, Cowling et
al. 1994a,b, Huston, 1994, Collins, S.L. and Benning, T. 1996, Martinez
1996, Woodward et al. 1996). Varying
definitions of functional types are so far most commonly associated with guilds
(Bahr 1981, Gillison 1981, Huston 1994, Gitay and Noble 1996, Mooney 1996,
Shugart 1996, Smith, 1996, Smith et al.,
1996, Gillison and Carpenter 1997), but as
A new quantitative method has been developed for
characterising vascular plants according to a set of 35 Plant Functional
Attributes that describe a plant as a three component ‘coherent’ (sensu Vogel 1991) or functional model. This consists of the photosynthetic envelope, modified
Raunkiaerean life form (Raunkiaer 1934) and above-ground root system. The
method uses a semantic rule set and grammar (Gillison and Carpenter, 1997) to
generate a theoretically finite set of unique PFA combinations for the world’s
vascular plants. Any one combination is termed a functional modus. Using this rule set, about 7.2
million combinations or modi are
possible, although it is thought that in reality the number is closer to 4,000.
There is no a priori interdependence
between modi and species; as the
mapping is many-to-many, i.e. more than one modus
can occur within a species and vice versa
. The advantage of functional over
solely species-based methodsis that the former can be universally applied by
observers with limited botanical and ecological experience. It can be used to
compare functional characteristics of individuals and sets of individuals
independently of species, e.g. where taxa may be geographically disjunct but
may possess similar adaptations to environment.
In a comparative study of methods of characterising site productivity
and growth patterns in
Richness in species and unique modi can be a useful complementary descriptor of habitat. But while these contribute to characterising
biodiversity, they do not by themselves reflect evenness or dominance of
individuals per species such as the frequently used diversity indices of Shannon-Wiener and Simpson (Magurran 1988).
Many diversity indices have been developed, but the search goes on (Cousins
1991, Majer and Beeston 1996). The great majority are based on species
abundance and at best are usually regarded simply as another species-based,
stand attribute. A problem for survey in tropical forests is that to generate
such indices requires time-consuming counts of individuals which is not
cost-effective. To help circumvent this problem, Gillison et al. (Appendix 2.1) have developed a complementary measure of functional diversity based on the
numbers of modi per species for each
plot. This differs from approaches by others (e.g.
It is one of the tenets of RBA that for practical purposes there should be indicators or surrogates of more complex plant and animal assemblages. Whether this is a realistic assertion is a continuing source of debate (Cranston and Hillman 1992, Reid et al. 1993, Pearson 1995, Howard et al. 1996), and there is often questionable theoretical support for targeting so-called keystone species (Tanner et al. 1994). There is nonetheless an increasing need for reduced attribute sets that can be used to carry other information such as the status of key pollinators and seed dispersers that may not be available at the time of survey (Miller et al. 1995) To demonstrate indicator efficiency requires calibration from very intensive baseline studies of taxa and functional types at a comprehensive range of spatial, temporal and environmental scales. Such baseline studies are almost non-existent in complex tropical environments. Ongoing studies within the context of ASB show varying correlative trends. In a baseline study of Sumatran rain forests, Gillison et al. (1996) showed that while plant biodiversity increased with elevation from 500 to 900m asl, the converse was true for insects and birds. While such confounding effects can be accommodated by appropriate regression models and multiple discriminant formulations, predictive models of biodiversity based on environmental correlates such as elevation clearly need to be carefully evaluated before being used by managers. It follows that environmental context and scale are important in designing field studies of biodiversity (see also He, et al., 1994,).
Most practitioners now concede the landscape matrix is critical to supporting biodiversity (cf. Forman and Godron, 1986, Franklin 1993), and this has been central to survey design and data collection across all the ASB and CIFOR ercoregional benchmark sites. Because disturbance is a critical determinant of biodiversity (Petraitis et al.., 1989, van der Maarel 1993, Phillips et al.. 1994), factors such as agriculture, shifting cultivation and forest fragmentation (Grime 1979, Bierregard et al., 1992, Sayer and Wegge 1992, Margules and Gaston 1994, Brooker and Margules 1996) should be considered when designing a survey. For this reason, the ASB sites have been located specifically to sample a range of dynamic conditions, along successional gradients of land use from pristine rain forest, logged-over forest, plantations to degraded grasslands. Although the issue of plot size is a continuing source of debate in plant ecology, recent studies show that for plant diversity, useful information can be recorded from plots as small as 50 x 2m (Parker and Bailey 1992, Parker and Carr 1992, Parker et al. 1993) and 40 x 5m. (Gillison et al. 1996). The advantage of ‘small and many’ vs. ‘few and large’ is that the former is more cost-effective when sampling variation in biodiversity at landscape level (cf. Keel et al. 1992). Variation of this kind demands cost-effective survey techniques (cf. Margules and Haila 1996). Because the distribution of plants and animals is determined mainly by environmental gradients, gradient-based techniques using the gradsect approach offer one means of sampling such variation (Gillison and Brewer 1985). With gradsects, sites are located according to a hierarchical nesting of assumed physical environmental determinants such as climate, elevation, parent rock type, soil, vegetation type and land use. This approach has been shown to be more cost-efficient than purely random or purely systematic (e.g. grid-based) survey design (Gillison and Brewer 1985, Austin and Heyigers 1989). As gradients themselves are being sampled, this will enhance the efficiency of extrapolative spatial models.
Issues of biodiversity conservation inevitably raise important questions of site representativeness. For a programme involved in the selection of ‘best-bet’ options for biodiversity and productivity, a manager may need to choose between different locations to ensure optimal management. For this a range of sophisticated computer-based solutions already exists. These are based mostly on species occurrence but may include environmental features such as land classes (Nicholls and Margules 1993, Pressey et al. 1996, Csuti et al.. 1997, Pressey et al. 1997). Other species-based approaches use additional levels of higher taxa (Prance 1995) or a measure of ‘phylogenetic distance’ to include taxic richness or genealogical relationships as embodied in taxonomic classifications, typically by a weighting of the relative number of species per genus, genera per family etc. (Vane-Wright et al. 1991, Williams et al.. 1992, Faith 1992, 1993, 1995). A problem with species-dependent approaches of this kind is that for many tropical lowland forests, species identification is difficult and time-consuming. In addition, the majority of these algorithms require expertise that is frequently lacking in developing countries. For this reason, and because functional types can be more easily identified than species, Gillison et al., (unpublished 1998) developed an analagous concept of ‘functional distance’ based on modi (outlined in Annex I). The algorithm is being incorporated in a new data-entry software package PFAPRO designed to run on a PC as a Windows application (Carpenter and Gillison, unpublished 1998). When data from a series of plots containing functional modi have been entered, PFAPRO has the facility to generate a distance matrix on demand. By this method, managers can readily identify levels of similarity between plots or landscape units.
Data collected during this project will be used to generate
and test spatial models of key sets of taxa and functional types and to couple
these with productivity patterns based on land use. For this purpose a
potential mapping software package DOMAIN (Carpenter et al. 1993) will be used. Unlike other packages such as BIOCLIM
(Busby 1991) or CLIMEX (Sutherst and
Maywald 1985) that are either climate-dependent or require detailed, process-based
knowledge about the species in question, DOMAIN uses any georeferenced data
that are considered important in influencing performance of an individual. This
may include environmental data used to construct a gradsect–based survey.
DOMAIN then accepts known distribution points for specific taxa or functional
types and constructs an environmental envelope for these using environmental
correlates and a distance measure based on the Gower metric. It then computes a
grid-based distribution map of according
to the similarity matching of each pixel or grid with the original
environmental domain. DOMAIN has been used in previous baseline studies in
Most vegetation classification and survey methods incorporate a combination of broad structural variables coupled with seasonality (deciduousness) and a list of dominant species, e.g.’Very tall evergreen Dipterocarp forest’. While this is useful for many geographic purposes it is insufficiently diagnostic for management purposes. In addition, structurally similar vegetation types are usually annotated by regionally different plant species. Within a region, vegetation described according to vegetation structure may be adequate for describing animal habitat but similar structure in separate global ecoregions are not necessarily ecologically equivalent. For ecologically sensitive classifications additional, response-based attributes such as adaptive features or plant functional attributes (PFAs) provide added value. As PFAs are generic and largely independent of species, they can be used to make ecological comparisons between geographically remote areas where environments and adaptive features may be similar but where species differ.
2.2 Methods:
The Plant Functional Attribute proforma (modified from Gillison, 1988 and updated by Gillison and Carpenter, 1997) was used to record site physical features [georeference by GPS in degrees, minutes and seconds; slope percent (clinometer); elevation (m) (digital aneroid altimeter); aspect in degrees (compass); parent rock type; soil type; vegetation structure, (mean canopy height (m), crown cover percent, basal area (m2ha-1); litter depth (cm); Domin scale cover-abundance estimates of wood plants <2m tall and Domin estimates of bryophytes; all vascular plant species and plant functional types (PFTs]. As described by Gillison and Carpenter (1997), Plant Functional Types or PFTs or functional modi are combinations of essentially adaptive morphological or functional attributes (e.g. leaf size class, leaf inclination class, leaf form and type (distribution of chlorophyll tissue), coupled with a modified Raunkiaerean life form and type of above-ground rooting system. PFTs are derived according to a specific grammar or rule set from a minimum set of 35 functional attributes. An individual with microphyll-sized, vertically inclined, dorsiventral leaves supported by a phanerophyte life form would be a PFT expressed as MI-VE-DO-PH. Although they tend to be indicative of a species, they are independent of species in that more than one species can occur in one PFT and more than one PFT in a species. PFTs allow the recording of genetically determined, adaptive responses of plant individuals that can reveal infraspecific as well as interspecific response to environment (e.g. LUTs) in a way that is not usually contained in a species name. They have a major advantage in that, because they are generic, they can be used to record and compare data sets derived from geographically remote regions where, for example, adaptive responses and environments may be similar but where species differ. The data are recorded along a 40x5m strip transect located along the contour.
The data were compiled in a laptop computer using a recently developed software package, PFAPro (Gillison and Carpenter, unpublished). PFAPro facilitates compilation according to the rule set developed by Gillison and Carpenter (1997). It also facilitates the summary analysis of meta-data as well as producing graphs of relationships between different plant and vegetation variables. Using PFAPro, data logged for each 5x5m quadrat allow the generation of cumulative species and PFT totals per unit area and this allows the subjective inspection of asymptotic curves that can indicate whether or not a plot is an adequate sample of the vegetation or LUT (See Annex 1, Fig.1).
In addition to site physical data, simple totals of species, PFTs and vegetation structural variables, PFAPro can be used to generate a range of diversity indices for PFTs (Shannon-Weiner, Simpson and Fisher’s alpha). The calculations are not trivial as, unlike diversity indices for species that are based on abundances of individuals per species, the PFT indices are derived on the number of species per PFT. Since the species to PFT relationship is many-to-many, this must be taken into account when calculating diversity. The method is described more fully in Appendix 2.1.
Four observers (ecologist and assistant, botanist (x2) and two laborers) collected plant voucher material later identified and curated at the Herbarium Bogoriense. A complete set of identified species and associated PFTs is listed in Annex III, Table 3. This method facilitated sampling even the most complex rain forest plot of 177 species in less than three hours. Photographic records were made of each plot. A sub-set of these has been scanned and will be cross-referenced with the data set.
Results:
The data were analysed according to the methods described above and in Part B. The most useful interpretations came from multidimensional scaling in which a two vector solution was extracted from plot data (Part B, Annex I, Fig. 7). This graph shows a zone of maximum biodiversity richness that is associated with jungle rubber. The peak in richness can be explained in part by the greater variety of available ecological niches in this agroforestry system compared with pristine rainforest. The analyses are based on a minimum data set of mean canopy height, basal area, species richness, PFT richness and a ratio of species numbers to numbers of PFTs or modi. Cumulative species, modi and species/modi richness area curves per 40x5m plot are indicative of vegetation type per LUT (Part B, Annex I, Fig.1 (1-7)). More detailed results from analyses of combined sets of taxa and functional types are described in the synthesis (Section 11). Other analyses dealing with variations on compositional structure of species, PFTs and vegetation structure and their relation to LUT will be dealt with in a later report. Plant taxa and functional types for each LUT are listed in Annex III Table 3. Summary data are listed in Table 2.1 and estimates of green biomass are given in Table 2.2 below. Relationships between vegetation and LUTs are described briefly in Section 3 below.
Table 2.1.
Summary of Taxa and Plant Functional Types (Modi) per LUT
|
No. |
Site |
Family |
Genus |
Species |
Uniq Sp/Plot |
Modi |
|
1 |
BS1 |
44 |
82 |
103 |
102 |
37 |
|
2 |
BS2 |
43 |
81 |
104 |
100 |
36 |
|
3 |
BS3 |
32 |
48 |
50 |
50 |
20 |
|
4 |
BS4 |
45 |
83 |
111 |
108 |
39 |
|
5 |
BS5 |
43 |
82 |
117 |
112 |
38 |
|
6 |
BS6 |
26 |
35 |
42 |
42 |
27 |
|
7 |
BS7 |
25 |
43 |
48 |
46 |
33 |
|
8 |
BS8 |
37 |
60 |
68 |
65 |
37 |
|
9 |
BS9 |
31 |
52 |
58 |
54 |
30 |
|
10 |
BS10 |
53 |
97 |
115 |
111 |
47 |
|
11 |
BS11 |
49 |
89 |
100 |
97 |
41 |
|
12 |
BS12 |
6 |
10 |
11 |
11 |
10 |
|
13 |
BS13 |
6 |
7 |
7 |
7 |
5 |
|
14 |
BS14 |
7 |
12 |
15 |
15 |
12 |
|
15 |
BS15 |
8 |
19 |
19 |
19 |
13 |
|
16 |
BS16 |
22 |
40 |
43 |
42 |
34 |
|
|
Total |
477 |
840 |
1011 |
981 |
459 |
|
|
|
|
|
|
|
|
|
|
Unique Total |
91 |
320 |
_ |
557 |
216 |
Table 2.2.
Green biomass per Land Use Type*
|
Site.no |
LUT |
Av.kg/m2 |
stdev |
coefvar |
SEM |
C-t/ha |
|
BS01 |
NF |
0.133 |
0.079 |
0.594 |
0.028 |
0.533 |
|
BS02 |
NF |
0.000 |
0.000 |
* |
0.000 |
0.000 |
|
BS03 |
LOF |
0.000 |
0.000 |
* |
0.000 |
0.000 |
|
BS04 |
LOF |
0.045 |
0.083 |
1.854 |
0.029 |
0.179 |
|
BS05 |
LOF |
0.007 |
0.020 |
2.828 |
0.007 |
0.028 |
|
BS06 |
HTI |
0.247 |
0.159 |
0.642 |
0.056 |