The NHMUK Chinese type collection was digitised and transcribed as part of the GPI project. The type specimen dataset contains 3,737 records, collected by more than 200 collectors, with geographical coverage from almost all of the provinces of China. The GPI project digitisation included taking a high quality digital image and transcribing labels, but did not include georeferencing (to give the specimens’ collection coordinates). The aim of the current project was to georeference the whole collection as close to an exact locality as possible, although we had to acknowledge that this would not be possible in some cases where the label is missing detailed locality information. A secondary aim was to identify as many unique site localities as possible to simplify and avoid duplication when georeferencing. Many collectors have collected multiple specimens at the same location, sometimes even returning to the same location at a later date. At some sites, multiple collectors have collected specimens, with some specifically visiting sites to follow an earlier collector’s collection routes. In such examples, we can use the same locality information for multiple collection events or specimen locations. It is important to identify these sites, as examples of repeated collecting over long periods of time are comparatively rare and these data are crucial for projects that use specimens to understand diversity and evolutionary changes.

During the process of georeferencing, we found that the existing transcribed label data did not accurately represent what was on the specimen label. Careful checking of the label information showed that up to 80% of the locality data transcribed was inaccurate, therefore the label information had to be checked for each specimen and re-transcribed where necessary.

One of the challenges in transcribing these specimens was the transliteration (and occasionally translation) of Chinese. As a language completely different from the European languages and writing systems, transliteration is not an easy task. Transliteration (romanisation) of Chinese has changed a lot in the last couple of hundred years, especially in the 20th century, covering the time period when many of the specimens in this project were collected. While it is beyond the scope of this paper to give a comprehensive overview of Chinese translation and romanisation, we hope to give the reader a short primer and where they can find more information.

Handwritten labels are harder to interpret, so marking uncertainties on the transcription is helpful for subsequent users of the data. In case of uncertainty by the first transcriber, someone else can look at the label or georeferencers themselves can take a second look on the label.

Prior to transcription, we agreed on and used a set of formal abbreviations, often historical and based in Latin, for sections that cannot be transcribed as the information is missing from the label. This is fairly common in natural history collections (Table 2).

Blank fields should only appear for records which have not been transcribed yet. If the transcriber is uncertain of the information, they can indicate that with brackets, square brackets or dots, whichever the institutes prefers to use in their collection system (Groom et al. 2019). We used the following indications (the bold words in Table 2), we also used “Illegible” for illegible collector names and “[illegible]” for illegible locality information. In case of uncertainty, we used “[word ?]”, indicating that we are unsure of the transcription of the word (to the extent that we could not say whether it is a locality, collector etc).

As supplementary data, we created a dataset about the collectors from whom we have specimens in the Museum’s Chinese type collection (Suppl. material 1). This includes a small number of collectors who turned out not to have collected in China, but whose collections had been mistakenly filed in the Chinese geographical region. Since their collection was also georeferenced as part of this project, we have kept them in the dataset. For all the collectors, their Global Plants Collector profile is included where available, as well as their Wikidata biography page, for example, the page for Henry Fletcher Hance (Wikidata Contributors 2019) . From the 265 collectors, 180 of them had an existing Wikidata profile. For 37 collectors, we created their page, based on biographical information. For the remaining 48 collectors, the biographic information was too limited to be able to create a Wikidata page.

While transcribing, we checked the collector’s active working years and possible expedition descriptions to correctly identify them. We also compared label information to their biography information on Global Plants. For collectors with no Global Plants profile, we tried to find biographic information via Google or other sources. For Chinese collectors, Plants of China (Hong and Blackmore 2015) was especially helpful. Access is limited on Google books, but there is a copy of it in the NHMUK’s library. It provided a list of Chinese collectors, with names in pinyin, Chinese characters, variants of their names, birth and death dates and sometimes biographic information.

In botanical collections, we also have specimens collected by people who were not botanists by profession, but collected plants and sent them to other botanists or institutes. These collectors were biologists, entomologists or, quite often, other professions such as missionaries, doctors and surgeons. For these collectors, other sources were useful to find biographic information, depending on their affiliation. For missionaries, the best sources for biographic information are documents issued by the corresponding church about their missionary activities. Obituaries are also a good source, as they often describe the deceased person’s life in detail.

Our criteria for creating Wikidata profiles were based on having the following information available: full name or family name with initials at least; one or more categories of biographic information with a reference, for example, birth date, birth place, institute they worked for or graduated at etc.

Unfortunately for certain collectors, we only had limited or ambiguous information on when they were definitely alive or who they were. For these collectors, we were unable to create a profile at this point.

Following transcription, the next step is georeferencing. In this project, we followed a method shown in the flow diagrams below (Fig. 4). Although we always tried to find coordinates for the exact locality, this was sometimes not possible due to incomplete labels. In situations like these, we used centroid coordinates for the closest locality we could confirm, such as a country or a province. The protocol we used can be found in the article by Blagoderov et al. 2017 (Supplementary material 2). These centroid coordinates were mathematically defined using ArcGIS, using the Geometry tool, in ArcMap, ArcGIS. ESRI, ArcGIS Software.

Figure 4.  

Flow diagram on how to approach labels in different languages.

During the georeferencing process, we found that some of the ‘Chinese’ specimens are likely to be from different geographical regions. Dealing with an area with a long history of border changes, it is possible that the collection locality was part of China at the time of collection. For mountainous or remote areas, like Tibet, it is also possible that the collector(s) were unaware of the border crossing. However, mistakes can happen at any time and it occasionally happens that a sheet was mistakenly filed in a certain geographic region when, in fact, it belongs to a different region, sometimes not even bordering the one we are working with. Therefore, we did our best to carry out our own investigations, to make sure we did not spend time looking for a locality in a completely different country. When in doubt, we searched the collector’s biography on JSTOR Global Plants, to see which countries they worked in and compared this to our label information to determine where we should look to find the correct collection locality.

When the specimen had a Chinese label, we transcribed the Chinese characters into pinyin, to make the search easier, then used Google to locate the correct collection site.

Where labels were in English, we followed the procedure show in Fig. 5.

Figure 5.  

Flow diagram on how to approach labels written in English.

If the specimen were collected before the 1950s, the first step was to check if the collector has published a book about their travels or experiences or whether this was covered in any other publication. JSTOR’s database is a good source for these publications, as well as archive.org and occasionally Google books. Publications about collecting activities often provide multiple spellings of collection sites for easier identification and, where available, coordinates as well. In books written by collectors themselves, we often find descriptions of collection sites and how they got there and therefore, following the description of the route on Google maps, we can often positively identify the collection locality.

When there was no book/publication available or the locality was not mentioned, we needed to identify the romanisation system used by the collector. This was a manual process using the commonly-used systems in Table 1, but we also used less common, usually outdated systems described in Legeza (1968), Legeza (1969).

The most complex and problematic sites were the ones where there was no book or publication available and the locality name did not match any of the commonly-used romanisation systems. For these labels, please find the suggested procedure in the Alternative sources to find localities section.

It is worth mentioning that by familiarising ourselves at the start of the project with the most common or relevant geographic terms on the labels (i.e. city, river, mountain, valley etc.), we can more easily determine what part of the label has the locality information and name. We used a pre-made list as a ‘cheatsheet’, including not only the Mandarin Chinese words for the English equivalent, but the same words’ spellings under different romanisation systems and/or in different languages (e.g. French).

For French labels, follow Figure 4.1, but use the following steps if the answer for “Is there extra non label information about Collector” is "No":

1. Check a EFEO : pinyin comparison chart (easily accessible via Google)

2. Use comparison table to transcribe locality name into modern pinyin form

3. Google/Geonames search to find the locality

If the collection pre-dates the EFEO system:

4. EFEO comparison table is still usable, as the system was based on the transcription of French missionaries

If none of the methods helps, see Alternative methods to find localities paragraph.

For labels in languages other than English, French or Chinese, the flow diagram on Figure 4.1 can be followed and, if there is no extra non-label information about the Collector, see the section on alternative methods.

As we had labels written in various languages, in order to make work easier, we created a reference table with the most frequent generic localities in various languages (Table 4).

The label says “Herb. H. F. Hance”, meaning it is from the herbarium of H. F. Hance. H. F. Hance is Henry Fletcher Hance and it says “Ipse Legit” on the bottom of the label, meaning “collected himself”. The collection date is 1866. If we check Hance’s biography, in 1866 he was in Guangdong Province in China as a Vice-consul.

The verbatim information is: “Ad coenobium buddhisticum Filoi tsz secus fl. North River, prov. Cantoniensis.”

As the collection date is 1866, we cannot use Wade-Giles or any other known romanisation chart. What we do know that “prov. Cantoniensis” is approximately the current Guangdong Province (Wikimedia Foundation, Inc. 2019). “secus fl. North River” means that it is “near North River” (= Bei river). (See the river map)

In the The Journal Of Botany British And Foreign Vol-v (Hance 1867), there is a paragraph about the place in the “On Liquidambar Formosana, Hance” section by H. F. Hance.

“"in an excursion I made with Mr. Sampson up the North River to the Tsing-yune Pass, about 120 miles above Canton, in the magnificent dense woods encompassing the renowned Buddhist monastery of Filoi-tsz"

Tsing-yune Pass = Qingyuan Pass, which is near Qingyuan city towards Canton (Guangzhou) on the North River. At that time, it was common to travel on the rivers in China. Hance was stationed in Guangzhou at the time. He also mentions Mr. Sampson, who is Theophilus G. Sampson. Many of his collected specimens were given to H. F. Hance and the Filoi-tsz Buddhist monastery often appears for specimens he collected.

There is another good description about the area mentioning the same sites in the book Through China with a camera by John Thomson (Thomson 1899).

In that direction, away from Guangzhou, on the North River, near Qingyuan, there is one well-known Buddhist temple, the Feilai Temple (飞来寺). In the spelling Hance used, the “tsz” stands for “寺” (pinyin: sì or si4), meaning “temple”.

At the end, we double checked that all the information matches. The Feilai Temple is in Guangdong Province, on the North River. From the paragraph written by H. F. Hance, we know he travelled there in 1864 and also that, in 1866, he was living in Guangzhou. The Feilai Temple was built in AD 520 and it still stands. Therefore, we concluded that the “Filoi-tsz” on the label is indeed the Feilai Temple.

There is no collection date, but a collection event code: 9158 and a collector's name: Henry.

The verbatim (collection locality) information is: Yunnan, Man-mei; alt. 7000'.

The collector could be Benjamin Couch Henry or Augustine Henry. Augustine Henry is more likely, as he indeed collected in Yunnan Province. For more certainty, we looked at the historical literature, including the Protologue as well. Unfortunately, there is no first name mentioned. We then searched for duplicates of the type, with the same species, collection number and locality. Fortunately, there were three other specimens and the collector marked “A Henry” on all three. Therefore, it is highly likely that the collector is Augustine Henry (1857 – 1930). He was English speaking and living around the time when the Wade-Giles romanisation system was developed. We do not have a collection date, but the Protologue was published in 1898, therefore, the collection has to be earlier. We cannot determine what romanisation system A. Henry used, but we can exclude the EFEO, as he was not French speaking and the Gwoyeu Romatzyh, which was developed in the 1920s. Using a romanisation comparison chart (Pinyin.info), we can see that “Man” and “mei” is written the same way by all systems. It is safe to assume that we can search for the name “Manmei” without transliteration. The dash/hyphen is not used in the Chinese language, only in the Western romanisation systems, while the current pinyin system does not use it either.

Using the Geonames search engine for “Manmei” in Yunnan Province, we get five records, three of them an exact match to our name on the label. From the Protologue and the duplicates, we get the extra locality information “S. of Red River” or “Mts. S. of Red River”. From the three records, two of them are South of the Red River and we can reason that 漫美 (Manmei) is the most likely, as it is closest to the south bank of the Red River. However, we cannot be certain and, without the exact date he visited the location, at this point, we cannot identify the locality with absolute certainty.

Transcription

• To get transcription with the required quality, adequate time and training need to be provided to the transcribers.
• Due to the complexity of herbarium sheets, the transcription process needs to be closely monitored by curators, providing help where needed.
• Transcription guidelines are needed for questionable labels and label information and to support transcription in other languages, for example, to provide keywords.

Georeferencing

• When dealing with collections with non-European languages, georeferencing (as well as transcription) needs language and transliteration guidelines.
• Standards for locality georeferencing need to be implemented, as well as standards for recording missing and illegible information.

Project management

• The size of the collection should not be underestimated. Providing the necessary amount of time is essential. Generally, a quality over quantity approach leads to better results for transcription and georeferencing, reducing future re-work. It is challenging to estimate the time and resources required for a compex georeferencing project, especially in externally-funded projects. We would recommend agreeing a per record time limit for georeferencing and recording which could not be resolved within this limit. Some specimens may be more easily georeferenced once others have been done or could be georeferenced if there is time remaining at the end of a project.

Data Sharing

• It is hard for us to share georeferencing protocol information, including data sources that were used for individual records. We inconsistently recorded this information in our collections management system and when it is recorded, it is not always done in a structured way that can be easily shared on our data portal. This reduces the re-usability of potentially useful data that other collections could use.

For this project, collection size, project duration and the need for re-work on previous transcription meant that we could not spend as much time on georeferencing as originally planned. There are specific areas which need more time to determine and verify specific localities, for example, the “Lien chau” river in the vicinity of Guangzhou. In the future, we now know that we need to approach other countries / geographical regions with caution, as the transcription quality could be lower than expected. It would be a useful to build a database of collector’s publications prior to starting a future project, especially if certain collectors are very frequent in a collection. Many publications are available in the institutional libraries, but they are more useful in a machine- and human-readable digital format, where one can search the text to find the information needed. Many of these publications are available online for free or at least partially. We found it useful to have these publications ready when we started to work the collection, to speed up the transcription process.

This project was useful in terms of understanding that the quality of the initial transcription project and the data from it, was not as expected. We cannot set a tight schedule on transcription projects, as this will lead to poor quality data. Providing sufficient time and training and working closely with a curator, can help to reduce transcription errors.

Accurate transcription can help us make statistical analyses on the collection, for example, what areas were most collected; or which collectors provided how many specimens. We can concentrate on more prolific collectors first, mapping their localities, since they likely collected for a longer period covering more areas. Mapping those localities could help us to map the rest of the collection. This is especially true if those collectors used spellings that are not in use anymore.

We hope that the collated dataset about the collectors and the methodology provided in this paper, provide a helpful guide for future projects working on East Asian collections, as well as a starting point for other regions, based on non-Latin alphabets.