The Ocean Biogeographic Information System (OBIS) is an open-access data and information system for marine biodiversity for science, conservation and sustainable development.
In this example we will use R libraries obistools
and robis
to search data regarding marine turtles occurrence in the South Atlantic Ocean.
Let's start by loading the R-to-Python extension and check the database for the 7 known species of marine turtles found in the world's oceans.
%load_ext rpy2.ipython
%%R -o matches
library(obistools)
species <- c(
'Caretta caretta',
'Chelonia mydas',
'Dermochelys coriacea',
'Eretmochelys imbricata',
'Lepidochelys kempii',
'Lepidochelys olivacea',
'Natator depressa'
)
matches = match_taxa(species, ask=FALSE)
matches
scientificName | scientificNameID | match_type | |
---|---|---|---|
0 | Caretta caretta | urn:lsid:marinespecies.org:taxname:137205 | exact |
1 | Chelonia mydas | urn:lsid:marinespecies.org:taxname:137206 | exact |
2 | Dermochelys coriacea | urn:lsid:marinespecies.org:taxname:137209 | exact |
3 | Eretmochelys imbricata | urn:lsid:marinespecies.org:taxname:137207 | exact |
4 | Lepidochelys kempii | urn:lsid:marinespecies.org:taxname:137208 | exact |
5 | Lepidochelys olivacea | urn:lsid:marinespecies.org:taxname:220293 | exact |
6 | Natator depressa | urn:lsid:marinespecies.org:taxname:344093 | exact |
We got a nice DataFrame back with records for all 7 species of turtles and their corresponding ID
in the database.
Now let us try to obtain the occurrence data for the South Atlantic. We will need a vector geometry for the ocean basin in the well-known test (WKT) format to feed into the robis
occurrence
function.
In this example we converted a South Atlantic shapefile to WKT with geopandas, but one can also obtain geometries by simply drawing them on a map with iobis maptool.
import geopandas
gdf = geopandas.read_file('data/oceans.shp')
sa = gdf.loc[gdf['Oceans'] == 'South Atlantic Ocean']
atlantic = sa['geometry'].loc[0].to_wkt()
sa['geometry'].loc[0].exterior
%%R -o turtles -i atlantic
library(robis)
turtles = occurrence(
species,
geometry=atlantic,
)
names(turtles)
Retrieved 2000 records of 2309 (86%) Retrieved 2309 records of 2309 (100%) [1] "id" "decimalLongitude" [3] "decimalLatitude" "eventDate" [5] "institutionCode" "collectionCode" [7] "catalogNumber" "datasetName" [9] "phylum" "order" [11] "family" "genus" [13] "scientificName" "originalScientificName" [15] "scientificNameAuthorship" "obisID" [17] "resourceID" "yearcollected" [19] "species" "qc" [21] "aphiaID" "speciesID" [23] "scientificNameID" "class" [25] "lifestage" "coordinatePrecision" [27] "basisOfRecord" "sex" [29] "individualCount" "bibliographicCitation" [31] "dynamicProperties" "associatedReferences" [33] "datasetID" "geodeticDatum" [35] "identificationRemarks" "modified" [37] "occurrenceID" "occurrenceRemarks" [39] "occurrenceStatus" "ownerInstitutionCode" [41] "recordNumber" "specificEpithet" [43] "taxonomicStatus" "taxonRank" [45] "taxonRemarks" "type" [47] "vernacularName" "locality" [49] "accessRights" "coordinateUncertaintyInMeters" [51] "eventID" "footprintWKT" [53] "language" "references" [55] "waterBody" "depth" [57] "minimumDepthInMeters" "maximumDepthInMeters" [59] "continent" "eventTime" [61] "recordedBy" "fieldNumber"
set(turtles['scientificName'])
{'Caretta caretta', 'Chelonia mydas', 'Dermochelys coriacea', 'Eretmochelys imbricata', 'Lepidochelys kempii', 'Lepidochelys olivacea'}
Note that there are no occurrences for Natator depressa (Flatback sea turtle) in the South Atlantic. The Flatback sea turtle can only be found in the waters around the Australian continental shelf.
With ggplot2
we can quickly put together a of occurrences over time.
%%R
library(ggplot2)
turtles$year <- as.numeric(format(as.Date(turtles$eventDate), "%Y"))
table(turtles$year)
ggplot() + geom_histogram(
data=turtles,
aes(x=year, fill=scientificName),
binwidth=5) + scale_fill_brewer(palette='Paired')
One would guess that the 2010 count increase would be due to an increase in the sampling effort, but the drop around 2010 seems troublesome. It can be a real threat to these species, or the observation efforts were defunded.
To explore this dataset further we can make use of the obistools
' R package. obistools
has many visualization and quality control routines built-in. Here is an example on how to use plot_map
to quickly visualize the data on a geographic context.
%%R
library(dplyr)
coriacea <- turtles %>% filter(species=='Dermochelys coriacea')
plot_map(coriacea, zoom=TRUE)
However, if we want to create a slightly more elaborate map with clusters and informative pop-ups, can use the python library folium
instead.
import folium
from pandas import DataFrame
def filter_df(df):
return df[['lifestage',
'institutionCode',
'individualCount',
'sex',
'eventDate']]
def make_popup(row):
classes = ('table table-striped table-hover table-condensed '
'table-responsive')
html = DataFrame(row).to_html(classes=classes)
return folium.Popup(html)
def make_marker(row, popup=None):
location = row['decimalLatitude'], row['decimalLongitude']
return folium.Marker(location=location, popup=popup)
from folium.plugins import MarkerCluster
species_found = sorted(set(turtles['scientificName']))
clusters = {s: MarkerCluster() for s in species_found}
groups = {s: folium.FeatureGroup(name=s) for s in species_found}
m = folium.Map()
for turtle in species_found:
df = turtles.loc[turtles['scientificName'] == turtle]
for k, row in df.iterrows():
popup = make_popup(filter_df(row))
make_marker(row, popup=popup).add_to(clusters[turtle])
clusters[turtle].add_to(groups[turtle])
groups[turtle].add_to(m)
m.fit_bounds(m.get_bounds())
folium.LayerControl().add_to(m);
embed_map(m)
%matplotlib inline
inst = turtles['institutionCode'].value_counts()
inst.plot.bar(figsize=(13, 3.75));
We can get fancy and use shapely to "merge" the points that are on the ocean and get an idea of migrations routes.
First let's use obistools
QA/QC functions to eliminate the points that fall on land.
%%R -o land
land <- check_onland(turtles)
plot_map(land, zoom=TRUE)
turtles.set_index('id', inplace=True)
land.set_index('id', inplace=True)
mask = turtles.index.isin(land.index)
ocean = turtles[~mask]
Now we can use shapely's buffer to "connect" the points that are close to each other to visualize a possible migration path.
from folium.plugins import Fullscreen
from shapely.geometry import MultiPoint
from shapely.ops import polygonize
m = folium.Map()
Fullscreen().add_to(m)
for turtle in species_found:
df = ocean.loc[ocean['scientificName'] == turtle]
positions = MultiPoint(
list(zip(
df['decimalLongitude'].values,
df['decimalLatitude'].values)
)
).buffer(distance=2)
folium.GeoJson(
positions.__geo_interface__,
name=turtle,
tooltip=turtle,
style_function=style_function(color=colors[turtle]),
).add_to(m)
m.fit_bounds(m.get_bounds())
folium.LayerControl().add_to(m);
m
One interesting feature of this map is Dermochelys coriacea's migration between Brazilian and African shores.
More information on Dermochelys coriacea and the other Sea Turtles can be found in the species IUCN red list.