Diario de jhorthos

Update: I have made much progress with the software and it seems to work well and is reasonably robust. I look forward to releasing something to a public repository before too long (currently it is October 2022). In the meantime, if you want to request a correlation list for some favorite taxon let me know and I will do it for you, but only plants please.

Apparently this roughly corresponds to "phytosociology" which honestly sounds like a stupid name.

Teaser: in the PNW, which fern is geographically most associated with Polystichum munitum?
A) Struthiopteris spicant (deer fern)
B) Athyrium filix-femina (lady fern)
C) Polypodium glycyrrhiza (licorice fern)

After much incompetence, I finally have a preliminary Python program that can take iNaturalist observations (as provided to GBIF) and produce Pearson correlation coefficients for observation counts over a grid of latitude/longitude bins. Just in case anyone happens to read this and is interested. :-)

e.g. this output roughly makes sense:

--taxonA "Polystichum munitum" --taxonB "Athyrium filix-femina"
Latitude 32.0 to 52.0, increment 0.10, bins 200
Longitude -130.0 to -116.0, increment 0.10, bins 140
17926 taxon A
6026 taxon B
Correlation coefficient 0.673


compared to

--taxonA "Polystichum munitum" --taxonB "Myriopteris gracillima"
Latitude 32.0 to 52.0, increment 0.10, bins 200
Longitude -130.0 to -116.0, increment 0.10, bins 140
17926 taxon A
704 taxon B
Correlation coefficient 0.089


There are some issues I am not sure how to deal with, in particular how to handle grid blocks where both species have no observations (for now I just leave them in).

Oh, I also have a program that measures physical distances (each observation for taxonA gets a distance to the closest observation of taxonB).

General bias issues:
1) ease of recognition/identification
2) charisma/size
3) geographical accessibility, most problematic in rugged mountain and severe desert areas

(thinking about these)

Common or statistically enriched:

Adiantum aleuticum, Androsace nivalis, Anemone drummondii, Aspidotis densa, Castilleja elmeri, Chaenactis thompsonii, Claytonia megarhiza var. nivalis, Ivesia tweedyi, Lewisia columbiana, Lomatium cuspidatum, Oreocarya thompsonii, Poa curtifolia, Polystichum lemmonii, Sedum rupicola, Viola purpurea

17 taxa, index taxon Viola cuneata

Viola cuneata, Ceanothus pumilus, Pseudotrillium rivale, Chamaecyparis lawsoniana, Erythronium citrinum, Collinsia linearis, Hastingsia serpentinicola, Sedum laxum, Castilleja elata, Castilleja brevilobata, Viola hallii, Calochortus howellii, Pinguicula macroceras, Lithospermum californicum, Arctostaphylos nortensis, Micranthes howellii, Darlingtonia californica

Taxa sorted by correlation to index taxon.

23 taxa, index taxon Larrea tridentata, min weight 5.0

Acourtia nana, Agave lechuguilla, Alloberberis haematocarpa, Cercocarpus breviflorus, Condalia warnockii, Dasylirion wheeleri, Dasyochloa pulchella, Echinocactus horizonthalonius, Echinocereus coccineus, Echinocereus dasyacanthus, Echinocereus roetteri, Echinocereus stramineus, Echinocereus viridiflorus, Juniperus monosperma, Koeberlinia spinosa, Larrea tridentata, Mammillaria grahamii, Parthenium incanum, Prosopis glandulosa, Rhus microphylla, Sclerocactus uncinatus, Thymophylla acerosa, Yucca faxoniana

Current analysis and lists at Google docs:

Spatial clustering pages

NOTE - I am still actively trying to decide the best geographic parameters and cutoffs for cluster definitions, so these lists will be changing from time to time for now.

See https://en.wikipedia.org/wiki/Species_distribution_modelling

This is partially reproducing more academic schemes/clusters but I think this will still have value because it leverages this huge new crowd-sourced data set, which is growing by about 2x every year as of 2021. (Plus so far it seems to be hell trying to figure out where/how to access previous information. Scientific publication has its values but accessibility to the non-professional is not among them.)

I am building up quite a catalog of these, still not very well documented but quantitatively valid (they are output from clustering software, not personal lists). I list a few here and most are on Google Doc, indirectly linked above.

Most of these come out of a complete spatial analysis of western US plants (roughly west of Denver), the south half of BC, and a moderate slice of northern Mexico, based on research grade iNaturalist spatial data (so far at a grid resolution of about 4 km). Occasional taxa have a confusing name because the data come from GBIF, which seems to use older taxonomic names.

The lists may exclude taxa you expect to see (mostly when the number of observations is too small), and they may include taxa you did not expect (for example in the Wenatchee Serpentine cluster, Anemone drummondii is not serpentine specific, but it grows well on serpentine in the Wenatchee Mountains). They also may include taxa that are highly correlated with the community, but only over a part of the community range. If the geographic bounds used to do the original graph analysis cut through the range of a taxon, the correlations reported will not account for observations outside those bounds (there might be taxa that have a much broader range, though within the geographic bounds used the correlation is valid). Finally, there is an issue with taxa that are automatically obscured because the reports intentionally introduce random error in their positions.

Edge weight in the correlation graph is defined as 1 / (1 - PearsonCorrelationCoefficient), capped at 100. For example, nearly perfect correlated taxa have a connection (edge) weight of 100, correlation 0.9 gives weight 10.0, correlation 0.75 gives weight 4.0, correlation 0.5 gives weight 2.0, etc.

This will likely be an ongoing project and volunteers to help (especially interpreting what the heck these cluster are or mean) are welcome. I have a ton of data not shown here and am also happy to share software.

calcareous (cedar) glade communities of Tennessee and Kentucky region failed to be useful because the large majority of endemics are listed as endangered and the coordinate uncertainty associated with such observations is much larger than the typical calcareous glade itself. (plus most of them are under buildings now anyway)

23 taxa, index taxon Agastache breviflora

Agastache breviflora, Ageratina herbacea, Asclepias hypoleuca, Castilleja patriotica, Commelina dianthifolia, Cyperus fendlerianus, Echeandia flavescens, Glandularia chiricahensis, Hedeoma hyssopifolia, Heliopsis parvifolia, Ipomopsis macombii, Lactuca graminifolia, Linum neomexicanum, Lithospermum cobrense, Lupinus neomexicanus, Monarda citriodora, Polemonium pauciflorum, Psacalium decompositum, Salvia microphylla, Solanum stoloniferum, Stevia serrata, Tephrosia thurberi, Vicia pulchella

Second cluster (have not tried yet to figure out why they separate):
22 taxa, index taxon Rubus neomexicanus

Acmispon oroboides, Asclepias hypoleuca, Castilleja nelsonii, Cologania angustifolia, Commelina dianthifolia, Erythranthe cinnabarina, Galium mexicanum, Hedeoma hyssopifolia, Ipomopsis macrosiphon, Lathyrus graminifolius, Lithospermum macromeria, Monarda citriodora, Phaseolus parvulus, Phaseolus pedicellatus, Ribes pinetorum, Rubus neomexicanus, Salvia arizonica, Scrophularia parviflora, Sedum cockerellii, Solanum stoloniferum, Stevia serrata, Vicia pulchella