Has everyone overlooked Rumex salicifolius?

data
county records
analysis
Wisconsin
floristics
A first pass through likely county records data. Where do the candidates cluster, and which species keep showing up?
Published

July 16, 2026

Overlooked and underreported

Likely County Records

species expected but not yet documented

July
2026

Issue no. 1 ecology & miscellany

Has everyone overlooked Rumex salicifolius?

The basis for the above map and the following discussion is Likely Records. This takes Wisflora occurrence data from each county and finds species absent in the ‘home’ county but present in neighboring counties out to 50 miles, border to border. The figure above shows the number of taxa occurring in at least two of the neighboring counties but not the home county. It turns out there’s a lot of potential county records out there.

38733 species/county pairs, across 72 counties reveals some patterns and trends showing where botanists are looking, and what they’re looking for. I wondered which counties have the longest lists of potential species, how strong the signal is for the potential species, and what species kept showing up as absent even though they’re common in neighboring counties.

Which counties have the most potential species?

Figure 1

Longer lists here don’t necessarily mean collection gaps – they can also mean a county has a lot of well-recorded neighbors, as is likely the case with Dodge (Dane), although Dodge has its own relatively limited public land and associated plant collections. High collection counts are likely associated with both high population and colleges, universities, and museums which might house herbaria.

Other confounding factors are the size and location of the county. Most of the counties here are small, including Pepin–Wisconsin’s smallest. The effect is multi-factored; small counties are near other small counties, so they have a high neighbor count within 50 miles. Small areas are less likely to accumulate higher species richness than large areas. Furthermore, these counties generally occur on the interior of the state. Counties on the border have fewer neighbor counties from which to pull.

A final consideration is that in a county like Menominee–which includes the Menominee Indian Reservation–plant collections for western science may be less prevalent.

What is the signal strength?

n_neighbor_counties is the strongest signal in this dataset, which counts how many neighbor counties within 50 miles already have the species confirmed. A potential record with eight or nine neighbors is much more likely to be present than one with two neighbors. Again, geography and chance play a role here. Large counties and border counties may have fewer neighbors within 50 miles. Douglas and Bayfield counties illustrate this point as they are big counties at the northern end of Wisconsin, maxing out at 12 and 10 neighbor counties respectively. However, Columbia County has 32 neighbors counties for Veronica serpyllifolia, the most neighborly individual record in the dataset.

Warning: Removed 1 row containing missing values or values outside the scale range
(`geom_bar()`).
Figure 2

The distribution skews toward a high number of species appearing from a low number of counties. As an example, the southern counties might include a high richness of northern species, and vice versa. Nevertheless, there are at least 7,000 potential records with at least four neighbor counties.

The inset highlights the genera that are most common in occurences with 10+ neighboring counties. These genera are among the most cryptic and difficult to identify, so it makes sense that collections might occur piecemeal, likely following taxonomic experts.

Which species are overlooked?

Some species appear as ‘likely’ in a checkerboard statewide. For instance, one of the species just missing the list above is Festuca saximontana. The note from Wisflora illustrates this point, “It may be more common, but it is difficult to tell apart from F. rubra and F. trachyphylla, so collectors may ignore it, or specimens may be misidentified.” Other likely candidates may be in difficult-to-identify genera or may be overlooked non-native species that are less likely to be collected.

However, many of the species in the graph may show a natural checkerboard distribution, such as Primula mistassinica, finding refuge in distinct microhabitats that are small and irregular statewide.

Figure 3

Many of the species in this chart seem to be inherently patchy and rare, and thus might not make great candidates to search for.

Observed but not collected

The dataset also includes GBIF occurrence counts for each species/county pair, broken into total GBIF records and a subset from iNaturalist. This lets us ask which species have been purportedly documented in a county but never collected as an herbarium specimen.

Figure 4

This chart shows the top species by count that have been observed on iNat or GBIF but not in an herbarium. It’s fascinating that this group is almost entirely species used in prairie restorations, pointing to their success, reach, and uncertain status for plant collections. It also includes some likely landscaping/yard plants such as Picea glauca and Mertensia virginica.

The geography of undercollecting

Reaching back to the nature of the study here, some patterns stand out in the charts below. This confirms our assumption that smaller counties tend to accumulate more likely records, partly because of more neighbors and partly because a smaller area naturally harbors fewer total species. Interior counties also pull candidate species from neighbors on all sides, while border counties have fewer neighbors.

Figure 5

The negative trend makes sense for the above-described reasons. Some interesting trends here are that the counties below the trend line are often home to colleges and universities. Meanwhile, Wisconsin’s largest county (Marathon) is large but is located centrally in the state and has lots of neighbors.

Figure 6

The relationship here is slightly more convoluted, but counties touching or near the state border are generally in the lowest tier of likely records rather than the highest. The exceptions are interesting, and include poor Pepin and Lafayette. How many Wisconsinites could pick out Lafayette on a map?

Everything above runs directly against data/county_records.csv – the same file behind the Likely Records table.