About

about safeHavens

This package helps germplasm curators identify and communicate areas of interest to collection teams for targeting new accessions. For common species, it provides multiple sampling approaches for curators to choose from for each individual taxon they aim to collect. It also provides an additional sampling approach for rare species. The package allows for integration into existing workflows, that is reading in accession data from databases or Excel, comparing them to occurrence data, and writing out results into both tabular (e.g.. CSV, a database, Excel, etc.) and simple spatial data formats (e.g., geopackages, shapefiles, etc.), that can be shared with collection teams and work on handheld electronic devices (e.g., Qfield (free, preferred), ESRI products (paid, legacy)). It allows germplasm curators to have sophisticated spatial capabilities without the need to be GIS experts or paid software.

‘common’ species sampling approaches

Each approach is based on fundamental genetics and ecological theory, and loosely aligns with the concepts in landscape genetics/genomics. In practice, all functions attempt to capture the geographic and environmental variation across a species range. To this end, they evolve from simply partitioning a species range into n equal area, or regularly spaced areas, to modeling isolation by distance, resistance, and environment, or sampling from existing spatial data such as ecoregions or seed transfer zones. The geographic distance functions largely rely on Sewall Wright’s concept of Isolation by Distance (1943), the now empirically supported idea that populations that are more geographically proximate will be more genetically similar than populations that are further apart. All approaches in the package are a priori, with the exception of the EnvironmentalBasedSample function, which relies on a species distribution model (SDM) to inform sampling locations and is based on the concept of Isolation by Environment (Wang & Bradburd 2014); however this still does not have field valdiation. None of these sampling approaches are meant to supplant genetic diversity-based sampling, but we recognize that having these data is rare, and even gathering them would often miss the opportunity to make opportunistic seed collections along the way.

The methods have various trade-offs in terms of computational and environmental complexity, although all are designed to be run on a standard desktop or laptop computer, and processing up to a couple thousand of species overnight, with the faster methods, should be possible. The table below presents the currently implemented sampling scheme and the user-facing function associated with them.

Function	Description	Comp.	Envi.
PointBasedSample	Creates points to make grids over area	L	L
EqualAreaSample	Breaks area into similar size pieces	L	L
OpportunisticSample	Using PBS with existing records	L	L
IBDBasedSample	Breaks species range into clusters	H	L
IBRSurface	Breaks species range into clusters	H	M
PolygonBasedSample	Using existing ecoregions or STZs	L	H
EnvironmentalBasedSample	Uses correlations from SDM to sample	H	H
KMedoidsBasedSample	For rare species with known occurrences	M	M

geographic distance

The first two functions, PointBasedSample, and EqualAreaSample, are flavors of the same process, where we partition the species range into geographic clusters of similar sizes.
The OpportunisticSample method is a special case of PointBasedSample, where existing collection records are used to help guide the sampling scheme, in other words additional samples are designed around the existing samples. OpportunisticSample uses the same underlying logic as PointBasedSample, but first ‘removes’ areas around existing collection records, and then fills in the remaining gaps with new sampling locations The PointBasedSample and EqualAreaSample functions would be used when establishing a new collection strategy for a species, while OpportunisticSample would be used when augmenting an existing collection strategy.

IBDBasedSample also relies on geographic distance, but in lieu of using the continuity of geographic space as its primary method, it focuses on the discontinuity of space and uses distance matrices and clustering to identify portions of the species range that are closer to each other than to others. This method is slightly more computationally expensive than the previous geographic distance methods.

using regions

The PolygonBasedSample may be the most commonly encountered method in North America, and in various formats, is driving two major germplasm banking projects in both the Midwest and Southeastern United States, as well as at a high level, composing the way that numerous native seed collection coordinators are structured in the West. This method uses environmental variation as an implicit guide to target populations for seed collections; that is, the different ecoregions serve as stratification agents. In broad strokes, the general thinking is that these regions represent continuous transitions in the environment faced by the species, and populations across these ranges will be differently adapted to these environments. It can be used with either ecoregion or seed transfer zone based data. However, it relies on existing spatial data products, which may or may not be relevant to the ecology of the species.

resistance distance

The IBRSurface workstream allows users to parameterize a general cost surface for a landscape. The least-cost paths between occupied portions of the species range will then calculated and clustered. Spatial data from the end of this process can be passed into PolygonBasedSample if more than one collection per zone are desired.

environmental distance

EnvironmentalBasedSample, is both the most computationally expensive and the most environmentally explicit. This function fits a Species Distribution Model (SDM), generated via a generalized linear model (glmnet), supported by this package, to cluster populations based on environmental variables related to their observed distributions and the spatial configuration and distance between them. On paper, this draws together all aspects of the above functions; however no empirical testing of this approach has been implemented.

Spatial data from the end of this process can be passed into PolygonBasedSample if more than one collection per zone are desired.

rare species sampling approach

All of the above methods are designed for common species, that is species with more than 50 or 100 unique occurrence records. Each method returns a polygon geometry that provides guidelines for the general regions where the species should be sampled. However, for rare species, where the occurrence of individual populations are tracked, and which are generally collected along maternal lines, requiring considerably more field effort to gather seed, a supplemental approach exists that suggests the priority in which individual populations (‘points’) can be sampled.

The KMedoidsBasedSample method utilizes either geographic distances alone or geographic distances in conjunction with a distance matrix developed from a few key environmental variables to suggest which populations should be prioritized for seed collections. This method is based on the concept of maximizing the dispersion of selected points in geographic and environmental spaces to best capture the overall variation present in the species range.

installation

safeHavens can be installed directly from github, using either devtools or remotes. Note that the software relies on a few other packages that may require additional system dependencies, such as rgeos and rgdal. It also requires a working installation of GDAL, PROJ, and GEOS on your system. This can occasionally be finicky to install, but once you have them effectively all open-source geospatial softwares are available to you.

safeHavens can be installed from GitHub with remotes or devtools. Some users have issues with devtools on Windows, but remotes tends to work well.

remotes::install_github('sagesteppe/safeHavens') 

# install.packages('devtools') 
# devtools::install_github('sagesteppe/safeHavens')