You are here

Harvard Forest Data Archive


Species Distribution Modeling of Carnivorous Plants Worldwide

Related Publications



  • Lead: Aaron Ellison, Matthew Fitzpatrick, Nick Gotelli
  • Investigators:
  • Contact: Aaron Ellison
  • Start date:
  • End date:
  • Status: completed
  • Location: Global
  • Latitude: -90 to +90
  • Longitude: -180 to +180
  • Elevation:
  • Taxa: Aldrovanda sp., Brocchinia sp., Byblis sp., Catopsis sp., Cephalotus sp., Darlingtonia sp., Dionaea sp., Drosera sp., Drosophyllum sp., Genlisea sp., Heliamphora sp., Ibicella sp., Nepenthes sp., Pinguicula sp., Sarracenia sp., Sarracenia purpurea, Triphylophyllum sp., Utricularia sp.
  • Release date: 2019
  • Revisions:
  • EML file: knb-lter-hfr.332.1
  • DOI: digital object identifier
  • EDI: data package
  • DataONE: data package
  • Related links:
  • Study type: modeling
  • Research topic: biodiversity studies; conservation and management; ecological informatics and modelling; physiological ecology, population dynamics and species interactions; regional studies
  • LTER core area: populations
  • Keywords: carnivorous plants, climate change, demography, dispersal, distribution, modeling
  • Abstract:

    Forecasting how carnivorous plant species will respond to climatic change is a key issue in their conservation and management but presents a number of challenges. These challenges derive from interactions between the relatively simplistic statistical methods typically used to forecast species responses to climatic change, which to date have been limited mainly to species distribution models (“SDMs) and particular aspects of the ecology of carnivorous plants, including their rarity, habitat specialization, and limited dispersal ability. The small ranges and oftentimes low local abundance of carnivorous plants provide few occurrence records, which increase the potential for poorly or over-fitted SDMs and misspecification of relationships with their “optimal” environments. The unique habitats in which carnivorous plants often grow also are difficult to characterize using the basic temperature and precipitation data that often undergird SDMs. Rather, habitats in which carnivorous plants are common often are decoupled from broader climatic patterns (e.g., many retain high soil moisture even during seasonal drought) and may be associated with frequent disturbance. Last, dispersal limitation also may constrain range shifts of carnivorous plants as the climate changes. These three issues raise two related questions that are critical for understanding and forecasting the future of carnivorous plants. First, to what extent are current carnivorous plants distributions constrained by climate; and second, how readily, if at all, might carnivorous plants disperse to colonize new habitat as it becomes climatically suitable?

    We estimated the vulnerability of carnivorous plants to climatic change in light of challenges identified with SDMs in general and their particular application to these unique species. We combined two approaches: “ensembles of small models”, which attempt to deal with the challenges of fitting SDMs for data-limited species; and “bioclimatic velocity”, which is an estimate of how fast a species would have to migrate to track its climatic niche, to provide initial assessments of the vulnerability of carnivorous plants to climatic change. We also explored, for one carnivorous plant (Sarracenia purpurea) how to additionally incorporate demographic complexity into its SDM.

  • Methods:

    For our initial SDMs of carnivorous plants worldwide, we assembled a comprehensive database of occurrence records of carnivorous from multiple sources, including the Global Biodiversity Information Facility (GBIF), the Australia Virtual Herbarium, and numerous regional herbaria. We applied four quality control measures to the occurrence records before using them for modeling: [1] updated and standardized species names to the latest taxonomy and to identify hybrids, which were removed from further consideration; [2] species records without geographic (e.g., latitude–longitude) coordinates but with collection location information were georeferenced using GEOLocate 2.0; [3] removed occurrence records with [a] low spatial precision; [b] whose coordinates were far outside the known range; [c] whose coordinates matched the centroids of political boundaries; [d] or whose coordinates fell within botanical gardens or museums; [4] removed all spatial replicates (i.e., records for the same species with identical coordinates). The archived dataset contains those occurrence records that fell within the known native distribution of each species and which had sufficient spatial precision given the resolution of the climate data (WorldClim)— maximum temperature of the warmest month, mean temperature of the coldest quarter, annual precipitation, and precipitation seasonality—at 2.5 arc-minute spatial resolution.

    For scenarios of future climates, we used decadal averages of these same four bioclimatic variables for 2050 from 32 future climate simulations statistically downscaled to 2.5 arc-minute spatial resolution. The simulations included output from numerous general circulation models and one representative concentration pathway (RCP 8.5) developed as part of the IPCC AR5 and available from the Research Program on Climate Change, Agriculture and Food Security.1 28.5.3

    We modeled habitat suitability for carnivorous plants using MaxEnt version 3.3.3E as implemented in the dismo package in R. We fit models using the default values for all settings. We fit MaxEnt models (default values for all settings and optimized for model complexity) for any species with at least 10 spatially unique occurrence records and constrained selection of background data (points selected at random from the study region and also used to inform model fitting) to those areas within any terrestrial ecoregion in which the species has been observed and immediately adjacent ecoregions, as defined using the World Wildlife Fund’s definitions of terrestrial ecoregions. We partitioned the occurrence data for each species randomly 10 times into calibration (70%) and evaluation (30%) datasets, and models were run on each of the 10 resulting datasets. The multiple models for each species resulting from different random splits of the occurrence data into training and test partitions were combined into a single ensemble by weighted averaging.

    For more detailed SDMs of Sarracenia purpurea, we augmented the distribution records in hf332-01-carnivorous-plant-data.csv with additional records collected by Hannah Buckley (HF193) and data collected by Aaron Ellison and Nick Gotelli throughout New England (HF159).

  • Use:

    This dataset is released to the public under Creative Commons license CC BY (Attribution). Please keep the designated contact person informed of any plans to use the dataset. Consultation or collaboration with the original investigators is strongly encouraged. Publications and data products that make use of the dataset must include proper acknowledgement.

  • Citation:

    Ellison A, Fitzpatrick M, Gotelli N. 2019. Species Distribution Modeling of Carnivorous Plants Worldwide. Harvard Forest Data Archive: HF332.

Detailed Metadata

hf332-01: all carnivorous plant records

  1. id: unique record identifier
  2. family: plant family
  3. genus: plant genus
  4. species: plant species
  5. study name
  6. range: inside or outside of known range of the taxon
    • inside: inside of known range
    • outside: outside of known range
    • unknown: unknown
  7. num.precision: spatial precision of location (unit: meter / missing value: NA)
  8. longitude: longitude (unit: degree / missing value: NA)
  9. latitude: latitude (unit: degree / missing value: NA)

hf332-02: Sarracenia purpurea records only

  1. source: source
    • Buckley: from HF193
    • GBIF: from hf332-01-carnivorous-plant-data.csv
    • MAPP: from HF159
  2. lat: latitude (unit: degree / missing value: NA)
  3. long: longitude (unit: degree / missing value: NA)

hf332-03: distribution modeling function R script

  • Compression: none
  • Format: R code
  • Type: R code

hf332-04: distribution projecting function R script

  • Compression: none
  • Format: R code
  • Type: R code

hf332-05: distribution velocities function R script

  • Compression: none
  • Format: R code
  • Type: R code