Humans instinctively classify the variation they see in the natural world and apply names to those classes (e.g., “mango”). If the mental model of a class is the same for the speaker and the listener, effective communication is possible. Cultural naming systems, or folk taxonomies, are the same as scientific taxonomies in this way, but the latter also contain processes for explicitly connecting a name to an organism or set of organisms. For each scientific name there is a single type specimen in a museum somewhere which is the official bearer of the name and which can be reviewed by others. Additionally, taxonomists may also publish detailed accounts of a group of organisms that share the same name: they create an explicit circumscription or taxonomic concept.
The situation is further complicated because many folk names refer to groups of species, not just single species (the basic unit of biological variation). Again, “mango” may be applied to a range of species in Asian countries where there are more than one; in the Americas, where mangoes are not native, there is generally only the species Mangifera indica. Similarly, “oak” may refer to tens or of species in some countries. In many cultures, where a broad, multi-species name is used, additional specifiers are added to indicate a single species, e.g., “gambel oak”.
Because the creators and users of Greenstand images come from many cultures, and are generally not experienced at looking at the variation of plants, we need to use scientific names to communicate about the plants. When we can, we need also to refer to particular species, because there may be great ecological variation among species in a genus; for example not all species in Acacia (a genus), are ecologically similar to Acacia mangium (a species).
The process of finding the scientific species name for a plant in a Greenstand photo can be difficult. While some species are distinctive and common and a tropical botanist may know then instantly, some will be morphologically close to other species, and may be uncommon. The first step in naming the species is to search through the images for different species. Since planters often only plant a few species at a time, searching requires subsampling the full time range and spatial range of plantings in a geographical area.
The next step is to assign morphotype (or morphospecies) codes to a putative species, based on the first image or first few images. We are using eight character codes. These codes can be anything and need not be related to the name of a suspected species, though if the species is indeed recognized the code may be derived from the first four letters of the genus and species. The value of these codes is that i) if a species name is misapplied at first, and later changed, the reference ID code stays unchanged (this happens a lot!), and ii) if a morphotype is not recognized, it can still have a code; see, e.g., INDE0002 in the herbarium, “Indeterminate #2”, a fairly common species in Haiti that we still do not have a scientific name for.
Some morphotypes used by Greenstand explicitly contain several species that cannot be reliably distinguished from images, such as EUCA0000 and CITR0000. This is not desirable, because of the potentially great range of ecological and cultural variation among these lumped species, but unavoidable, and is preferable to using a number of narrower morphotypes which were unreliably applied.
After creating ‘bins’ (folders for images) with morphotype codes, we match subsequent images, shuffling them into a folder based on their appearance. For instance if we are looking at 1,000 captures from Haiti, we might end up with 20 folders. Generally, a few species will account for most of the individuals, with some only having a few images, and some only having one. As the numbers build up and we have more opportunity to view variation among individuals, we may need to split a folder into two, or lump two folders into one. We can then create a synonym table, keeping one code current or active, and linking other, now unused, codes to it.
Eventually we try to match morphotypes across countries and continents. Some restoration species are ubiquitous around the world, some are local.
The final step is to try to identify the species to scientific name. This requires internet searches, going to (real or virtual) libraries, and communication with experts in an area. If ‘side data’ about the local name of a plant could be captured, this would help a lot. Once we can link to a scientific name, we can access and incorporate large amounts of external information (distribution in different countries, ecological characteristics, risk of invasiveness, etc.); for those with a database background: the scientific name becomes a key permitting links to global data tables of species information.
As this data structure develops we are capturing it in a database (in the form of a single XML document), and making the data available as a dynamically generated virtual herbarium, with country-based sub-herbaria (e.g., for Haiti).
This herbarium is now a resource for Greenstand volunteers using CVAT, who are giving the eight character taxon codes to the images they review.