2 Nature order: this is any sequence, however arbitrary, based on appearance or other plant characteristic and is used here in a more general sense than the less practicable concept of ‘natural’ order or affinity, supposedly reflecting the underlying evolution-ary sequence or pattern, and often attempted in monographs. The primary and most demanding function of most field guides is identification, equivalent to a thesaurus rather than a dictionary; so field guides are most often in some sort of nature order with similar plants listed and illustrated together, and separate page indexes to names.
Placing species together in their taxonomic families provides a crude type of nature order, providing similar families are next to each other and species are arranged by similarity within families. However, there is no universal agreement on these sequences, so you will have to arrange families, genera and species in a sensible order yourself (see Box 5.6):
98 Plant Identification
BOX 5.6 ORTHODOX AND CUSTOMIZED SPECIES SEQUENCES FOR
NATURE ORDER GUIDES
Various family arrangements have been devised, and some are used to enforce a standard order on textbooks and herbarium cupboards. Although newer schemes may be more accurate and in tune with molecular data, some of the older sequences are followed in herbaria and may be appropriate in your guide for this reason. Well-known schemes are Hutchinson (1973), Dahlgren (1975), Cronquist (1988, as followed by Ribeiro et al, 1999), Mabberley (1997, an updated version of Cronquist, 1988), and the Angiosperm Phylogeny Group (APG, 1998, 2003; see www.mobot.org/MOBOT/research/APweb/).
The APG scheme is based on modern molecular analysis and is universally respected, but is not complete for all genera, and the family order is only fixed to within groups or
‘clades’, thus defining a tree of life showing how species, genera and families (that is, taxa) are related, but not prescribing any linear sequence of the taxa at the ends of the tree’s branches. In fact, there is no reason why natural patterns of overall similarity should form a spectrum or fit smoothly along a one-dimensional sequence (try arranging your friends and relatives in a line based on overall similarity); even two-dimensional diagrams of affinity (Dahlgren, 1975) can only ever be crude approximations of the multidimensional patterns of nature. It is therefore surprising how far it is possible to do just this for practical purposes. For field use, it is relatively easy to choose a few key field characters to define a sequence, just as a picture guide to people arranged by gender, height, and skin and hair colour would be far easier to use than one arranged by their names. Decide what spectrum of characteristics you and your users want to highlight in the primary sequence; but strive to make your sequence as similar as possible to an orthodox scheme as this will help you to summarize the features of families in one place, close to where the species are listed.
For example, for three different guides to Ghana’s trees:
•
Hawthorne (1995) listed Ghana’s tree species in genus-species order in a guide to ecological features (not a field guide) that users would be expected to reference by name.
•
In a more technical woody plant identification guide, Hawthorne (1990) and Hawthorne and Jongkind (2006) used a sequence of families, reasonably compatible with the APG groupings, but with heterogeneous families split and sets of small, similar families joined, in an order optimized for a smooth sequence of leaf and other vegetative characters.
•
For a tree photo field guide, (Hawthorne and Gyakari, 2006), the sequence was based purely on bark slash exudate, texture and colour.
Many less technical field guides are similarly ordered by habit, flower colour or other morphological characteristics, or preferred habitat. Beware that some such sequences have the disadvantage that one species – for example, from several habitats – may have to be mentioned in various parts of the book.
•
Facilitate the ability to identify through structured browsing by adopting a sequence related to plant form (see Box 5.6). Consider ordering the species in their families as a first approximation to nature order; but you may need to break up large families and merge small ones to make all groups manageable and useful. For guides in nature order, a key can be fully integrated with the species-specific descriptions, an efficient solution for several reasons (see this chapter’s ‘Conclusions’, page 119).
Identification 99
•
In a field guide with a primary order other than alphabetical names, all cross-references to species names in the book – for example, from similar species or keys –
should also specify the page number or equivalent (for example, ‘See species X, group 22b’ or ‘p99’). Do not frustrate readers by requiring them separately to look up the page in an index whenever a name is mentioned (as in the otherwise well-designed guide by Arbonnier, 2004).
The benefits and limits of identification by browsing Tropical plant guides designed to facilitate recognition – picture-rich books – have obvious potential, especially where literacy is limited or if the book has to be translated into several languages. As digital photography becomes more accessible, this potential is increasing. Picture-based books (with local name indexes) seem particularly appropriate where the guide aims to help people who know the plants in their own area to establish the scientific names. But what is the chance of making such guides promote accurate identification as a primary aim when many similar species, mostly unknown to the users, occur in the same area?
The niche for picture-based field guides to tropical plants is outstandingly unfilled.
For the major patterns of variation in the plant life of typical rainforests, picture browsing alone allows at least a very useful first approximation (see Box 5.7). Some analytical guides or sections of guides, and longer-term education rather than piecemeal identification alone, are also essential if such guides are to be accurate for difficult groups. Even the most enthusiastic users need to be encouraged to study specific, often very subtle details, and there comes a point when to avoid analytical text for the sake of some users is to introduce a serious handicap in the potential for accuracy for others. If some users cannot read, and accurate identification of many difficult species is important, you and your committee of planners should question whether the combination of these users plus your field guide can solve the problem. More usually, even a low percentage of users who can read a guide in an optional analytical way will justify adding some technical, analytical text to support the images for difficult species.
Keys and other ways of diagnosing species by their characters The classic method for botanical identification by analysis is the diagnostic key, which has not changed radically since the 18th century (see Pankhurst, 1975, 1991, 1993; Tilling, 1984; and Edwards and Morse, 1995). A key is basically a ‘decision tree’, a figurative treasure hunt, where the answer to each question leads to another question until the right name is reached (the key might be formatted or phrased as a pair of optional statements; but each step can still be considered a question). Each key question filters out some of the remaining species until only one species is left. Keys can be displayed or formatted in various ways (see Table 5.1 and Plate 2, centre pages).
There might be one character considered per question, with character states representing the possible answers; or two or more characters might be combined into one key question – for example, if a combination of two makes the dichotomy workable in different seasons, or the individual characters define a set of species less reliably than their combination. The character states define the possible statements in a key; so, in order to make a truly dichotomous key, all multi-state characters can be translated to binary ones. It is often better for categorical and some numeric types of character to have the three or more optional answers all listed together (‘polychotomous’) at a partic-
100 Plant Identification
BOX 5.7 CAN PICTURE-BASED GUIDES BE ACCURATE
IN SPECIES-RICH PLACES?
Can browsing of picture-rich guides be accurate? This issue is one of those investigated by the UK Department for International Development’s (DFID) Forestry Research Programme (FRP) Field Guide Project R7367 (see Case study 8.1, page 184). The three main sets of trials in this project are all directly relevant to the question.
A trial photo guide to large trees in the Ghanaian rainforest was made showing bark slash and other tree details, containing 128 picture cards, one or more species per card.
With about 30 minutes of training, users from villages and towns with little prior knowledge of local trees could use this through structured browsing of pictures to identify more trees (about 80 per cent accuracy averaged across all trials) than professional tree spotters knew without the guide. The tree-spotters themselves were able significantly to improve their own performance with the same guide. The poor-performing minority of photographs could easily have been anticipated: they had slash patterns that looked very similar to other species and the minor differences visible were not consistent ones.
We also ran picture recognition tests in a wide range of Grenadian vegetation, from montane forest with many similar species, to drier lowlands where the differences between many species (stilt-rooted mangroves to fruiting mangoes) were easy to recognize.
Obviously, the levels of accuracy in the easy patches were much higher, reaching more than 80 per cent even with vanloads of schoolchildren. In the difficult vegetation, more than half of the 20 sample plants were still correctly matched. Intriguingly, people preferred photographs to drawings or paintings, but were not significantly more accurate with them overall. There was a slight trend for standard botanical drawings, which we presume induce a more analytical approach, to perform better with difficult species.
Various illustrative devices and even minimal basic text can be added to improve the results slightly: practice, user need and experience would also improve accuracy in the longer term. However, tests on Cola species on Mount Cameroon revealed the limitations.
Only about half of all Cola treelets from 19 species were matched correctly to their pictures, and it made little difference whether the pictures were photographs of live or herbarium specimens, or line drawings, or with or without small marks to highlight important details.
In summary, browsing helps almost any user distinguish plants within a heterogeneous set (that is, most crucial differences are easily seen), encouraging novices to realize that plant identification can be easy. A general average of 50 to 80 per cent species accuracy covering the studious, the gifted, the bored and the occasional half-drunk should be possible in a species-rich rainforest by browsing pictures whose primary order has been optimized for this task, and which are broken into smaller groups of 20 or less species.
For difficult groups of plants, one should not expect a browsing/recognition approach to yield accurate results.
ular point in the key. In any case, users consider one question at a time before moving onto another question in a predefined order.
In a ‘multi-access key’, also known as a polyclave, the user does not have to follow a predefined sequence of questions. They are based on complete lists of taxa and their characters, and therefore require more data to construct than simple keys which need only focus on the details in specific questions. The user selects the characters to be defined and the order in which they are considered. Missing characters on incomplete specimens can be avoided. Although these are mostly implemented as dynamic keys,
Identification 101
static versions can be made where one character perhaps defines columns and the other rows in a table, the answer being in the cell where the chosen row crosses the chosen column. Users still think about one character, or question, at once; but by placing the options in a table, the species that show various combinations of characters are clearly exposed. More than two characters can be considered in this way, for instance in a triangular table. A multi-access table may also subdivide into successive and more specific options, with rows and columns splitting (see Box 5.8); but such polyclaves can soon become too complicated and inefficient to use.
Table 5.1 Various formats of dichotomous or similar keys a. Numbered
596 Leaves
opposite
597
Leaves alternate
599
597
Flowers red
Species A
Flowers blue
598
598
Leaves serrated
Species B
Leaves entire
Species C
599
Leaves serrated
Species D
Leaves entire
Species E
b. Indented
Leaves opposite
Flowers red
Species A
Flowers blue
Leaves serrated
Species B
Leaves entire
Species C
Leaves alternate
Leaves serrated
Species D
Leaves entire
Species E
c. Indented and
596a Leaves opposite
numbered
597a Flowers red
Species A
597b Flowers blue
598a Leaves serrated
Species B
598b Leaves entire
Species C
596b Leaves alternate
599a Leaves serrated
Species D
599b Leaves entire
600a Leaves truly alternate Species E
600b Leaves sub-opposite Go to 596
d. Numbered, with
596 Leaves
opposite
597
return number
Leaves alternate
599
597
Flowers red
Species A
Flowers blue
598
598
Leaves serrated
Species B
Leaves entire
Species C
599
(596) Leaves serrated
Species D
Leaves entire
Species E
Note: Return number only necessary when the ‘parent’ question is not immediately above: helps retracing steps.
102 Plant Identification
e. Flow diagram, with boxes and text
Species A
Red
Flower
Species B
colour?
Yes
Opposite
Blue
Serrated?
Leaf
Start here
arrangement?
No
Species C
Species D
Alternate
Yes
Serrated?
No
Species E
Note: Although they can be very clear, flow charts occupy a surprising amount of space compared to a normal key with the same information content.
f. Flow diagram with pictures
Identification 103
g. Numbered, as with
596
Leaves opposite
597
(a), with icons to
designate
character states
Leaves alternate
599
597
Flowers red
Species A
Flowers blue
598
598
Leaves serrated
Species B
Leaves entire
Species C
599
Leaves serrated
Species D
Leaves entire
Species E
Note: As with (e), suitable for showing a limited amount of information in a simple, clear but spatially inefficient way.
h. Table with
Leaves opposite
Leaves alternate
successi
ve division
of columns only
Flowers red
Flowers blue
Leaves
Leaves
serrated
entire
Leaves
Leaves
serrated
entire
Species A
Species D
Species E
Species B
Species C
Note: Obviously, this type of tabular key can have added images as well.
i. Circular version of hierarchical divided table shown in (g) Note: The illustrated hummingbird guide is basically a circular list with words associated with each picture. In more complex versions, an inner circle represents the first choices of a key; each of these semi-circles may then be split into smaller sectors of the circle, leading to a species name in the outer rim.
Source: Texas Parks and Wildlife Publications
104 Plant Identification
BOX 5.8 DAWKINS’S GRAPHICAL FIELD KEYS OF UGANDA TREES
TWO-DIMENSIONAL MULTILEVEL TABLE
Dawkins’s (1951) multi-access key to Uganda trees based on bark and leaf details has inspired many writers of tropical tree guides. The primary table was divided in the following manner. The numbers refer to groups; further details were given for each species on a separate page. Names were put in these boxes as well. Box 6 of Table 5.1 is filled out as an example.
Table 5.2 Dawkins’s primary table
Source: adapted from Dawkins (1951)
Hawthorne (1990) used a similar approach, but added drawings and indented keys to vegetative features to separate the almost 700 species in Ghana. Pictorial multi-access keys (see Figure 5.1) were added as a graphical alternative for those who preferred it.
However, most users in Ghana preferred the simpler indented keys to these multi-access keys and the latter also take up more space and more time to make. They should be seen as often something of a gimmick, more of aesthetic satisfaction to the key creator than of benefit to practical users.
Plant Identification Page 18
