Data and code from: Importance of elephants for dung beetle biodiversity and ecosystem functions

Ecologists theorize that removing highly connected species from ecological networks will trigger waves of coextinction, but empirical evidence is scant. We show that elephants are uniquely central to a generalized network of interactions between dung beetles and large herbivores in East Africa. Computer simulations based on this network predicted disproportionate dung-beetle losses if elephants go extinct, and a 15-year experiment supported this prediction: Excluding elephants diminished dung-beetle abundance, diversity, and ecosystem functions, whereas excluding smaller ungulates had little added effect. Dung-beetle abundance was depressed in nearby landscapes where livestock had displaced elephants, corroborating the experimental results. Our findings affirm the longstanding idea that big animals—which are inordinately extinction-prone—act as keystone species by provisioning many linked consumers with essential resources, thereby sustaining biodiversity and ecosystem services. # Data from: Importance of elephants for dung beetle biodiversity and ecosystem functions Dataset DOI: 10.5061/dryad.9p8cz8wwn ## Description of the data and file structure ### Data 1. 01_db_network.zip.  A compressed folder of 5 files required to run dung-beetle trophic association analyses. The .R file contains the script to run analyses; .csv files contain the network and associated trap array data. **Data files:** **1. network_matrix.csv:** weighted matrix of beetle-species (rows) × dung-type (columns) interactions generated from the cafeteria pitfall trap experiment. Value in each cell denotes frequency of each interaction (i.e., number of beetles per species caught on each dung type). **2. dung_preference_data.csv:** dung-beetle preference data derived from cafeteria pitfall trap experiment. Column Headings: * pit_opu: pitfall trap operational unit; a unique identifier for each pitfall trap deployed in the cafeteria experiment.     * dung_bait: specifies the identity of dung used as bait in each trap. * Season: specifies the survey period (two time points denoted as “Dry” or “Wet).          * site: specifies the location along (7 total sites) * n: number of dung beetle individuals collected in each trap.       * sp_rich: number of dung beetle species collected in each trap. * n_max: maximum number of dung beetle individuals collected in any trap.      * sp_max: maximum number of dung beetle species collected in any trap.           * d_dung:  standardized Kullback-Leibler distance for each dung bait. * ni_nmax: number of individuals per dung bait divided by the maximum observed on any bait (n_max).         * si_smax: number of species per dung bait divided by the maximum observed on any bait (s_max). **3. guild_spp_preferences.csv: dung-beetle preference data broken up by functional guild and size class.** Column Headings: * herbivore_spp: herbivore dung bait identity. * gut_R.NR: herbivore gut type (ruminant vs. nonruminant). * gut_HG.R.O: herbivore digestive physiology (hindgut vs. ruminant vs. other).       * dung_dens: average species-specific dung densities estimated using regular dung count surveys conducted between 2009-2022 (Alston et al. 2022, *Ecology*). * dry_dung_pile_mass_g: species-specific dry dung pile mass estimates (g) collated from the literature.     * dung_mass: total copromass (in kg) estimated by multiplying dung_dens by dry_dung_pile_mass_g. * dweller: proportion of dweller individuals caught on each dung type. * roller: proportion of roller individuals caught on each dung type. * tunneler: proportion of tunneler individuals caught on each dung type. * small: proportion of small-bodied (<5 mm) individuals caught on each dung type. * medium: proportion of medium-bodied (5-10 mm) individuals caught on each dung type. * large: proportion of large-bodied (>10 mm) individuals caught on each dung type. **4. pit_filtered_rarefaction_matrix.csv:** sample-based matrix of beetle-species (rows) × dung-type (columns). Value in each cell denotes the number of traps in which each beetle species was caught. ### Data S2. 02_extinction_simulations.zip. A compressed (zip) file of 6 files required to run topological and rewiring herbivore extinction simulations. The 3 .R files correspond to scripts required to run the extinction simulations; the .csv files contain the network and beetle or herbivore trait data.  **R Scripts:** Each simulation type (topological vs. rewiring) is listed as a separate, independent script (db_topological_extinction_simulations.R and db_rewiring_extinction_simulations.R, respectively). The third script file (rewiring_function.R) contains the functions required to run rewiring simulations.  **Data files:** **1. network_matrix.csv:** weighted matrix of beetle-species (rows) × dung-type (columns) interactions generated from the cafeteria pitfall trap experiment. Value in each cell denotes frequency of each interaction (i.e., number of beetles per species caught on each dung type). **2. network_spp_traits:** beetle species’ traits used to inform rewiring conditions in extinction simulations. Column Headings: * species: dung beetle species Latin name. * subfamily: dung beetle subfamily (Aphodiinae, Scarabaeinae).        * guild: dung beetle functional guild. * size_class: dung beetle size class, categorized into small (<5 mm), medium (5-10 mm), and large (>10 mm) based on body length. * mean_body_length_mm: average body length in mm. **3. lmh_traits.csv:** as guild_spp_preferences.csv in Data S1. ### Data S3. 03_uhuru_experiment.zip. A compressed (zip) file of 10 files required to analyze the herbivore exclusion experiment data, including 2 files containing the raw CO1 sequencing data. The .R file contains the script required to run all analyses associated with Fig. 3, Fig. S3, and Fig. S10; .csv files contain dung counts and dry mass values for herbivores in exclosures, dung beetle trap-level data across treatments, and raw sequencing results from DNA metabarcoding. **Data files:** **1. dung_surveys_2009-2022.csv**: dung counts of herbivores in exclosures from 2009-2022. Column Headings: * survey: unique numerical ID for dung count surveys (total 69 surveys). * month: survey month denoted as a numerical (1-12). * year: survey year. * day: survey day in month. * site: location of herbivore exclosure (North vs. South sites). * block: herbivore exclosure block (1-3) nested within site. * treatment: herbivore exclusion treatment (Open: accessible to all herbivores, Mega: excludes elephants and giraffes, Total: excludes all herbivores >=5 kg).    * plot: combined ID for site and treatment. * line: transect line within plot. * Columns 10-22: dung counts for each herbivore species. **2. herbivore_dung_mass.csv**: dry dung pile mass values for herbivore species used to calculate dung availability across herbivore exclosures. Column Headings: * species: herbivore species Latin name. * common_name: herbivore species common name.          * dry_dung_pile_mass_g: species-specific dry dung pile mass estimates (g) collated from the literature.           * references_used: references for literature search of dry dung pile mass estimates. **3. uhuru_db_trap_data.csv**: dung-beetle community data across herbivore exclosures. Column Headings: * trap_ID: pitfall trap operational unit; a unique identifier for each pitfall trap deployed in herbivore exclosures.          * level: location of herbivore exclosure (North vs. South sites). * plot_block: herbivore exclosure block (1-3) nested within level. * exclu_treatment: herbivore exclusion treatment (Open: accessible to all herbivores, Mega: excludes elephants and giraffes, Total: excludes all herbivores >=5 kg).     * dung_bait: identity of dung bait used in pitfall traps.      * total_abun: number of beetle individuals caught per trap. * dry_biomass: dry biomass (in g) of beetles caught per trap.        * n_spp: number of beetle species caught per trap. * comments: additional comments from field sampling. **4. uhuru_db_size_trap_data.csv**: same as uhuru_db_trap_data.csv, but broken up by beetle size class. **5. uhuru_db_guild_trap_data.csv**: same as uhuru_db_trap_data.csv, but broken up by beetle functional guild. **6. uhuru_trap_nmds_matrix.csv**: binary (presence-absence) data for each beetle species caught in traps in herbivore exclosures. Column Headings: * Columns 1-5: as in uhuru_db_trap_data.csv. * Columns 6-87: dung beetle species. **7. laikipia_herbivore_densities.csv**: herbivore abundance and biomass estimates compiled from survey data across Laikipia and used to calculate keystone metrics. Column Headings: * Site_exact: location of density estimate.        * Site_bin: scale of density estimate (e.g., local vs. district-wide).     * AirGround: method of data collection (aerial survey or ground-based).     * Mode: mode of data collection. * QualBasis * Species: herbivore species Latin name.                      * Common_name: herbivore species common name.   * Year1: first year of density estimate. * Year2: last year of density estimate. * Years: number of years across which herbivore densities were calculated. * UnitWt_kgL: herbivore unit weight (kg).      * density_km2: herbivore density (kg/km2). * biomass_kgkm2: herbivore biomass (kg/km2).         * Source: abbreviated reference source. * Full_reference: full reference source. **8. uhuru_db_raw_sequences.csv:** raw DNA sequences for dung beetle samples collected in herbivore exclosures, generated using a cocktail of COI primers (MLepF1/C_LepFolR). Associated sample metadata are provided in the uhuru_db_raw_community_matrix.csv file. Column Headings: * seq_id: unique sequence identifier. * sequence: DNA sequence. * phylum: dung beetle phylum. * class: dung beetle class. * order: dung beetle order.   * family: dung beetle family.       * subfamily: dung beetle subfamily. * tribe: dung beetle tribe. * genus: dung beetle genus. * species: dung beetle species name. * bid_ok: Best identity score with the closest sequence in the database that provided the best match. * EXTR#XXXX: unique identifier for each DNA sample. Empty cells in taxonomic columns (e.g. tribe, genus) indicate missing data due to low taxonomic resolution. **9. uhuru_db_raw_community_matrix.csv:** raw community matrix indicating the presence or absence of dung beetle species in each DNA metabarcoding sample collected from herbivore exclosures. Data are under the form of a sample x species matrix. Column Headings: * extraction_ID: unique identifier for each DNA sample. * trap_ID: pitfall trap operational unit; a unique identifier for each pitfall trap deployed in herbivore exclosures.          * level: location of herbivore exclosure (North vs. South sites). * exclu_treatment: herbivore exclusion treatment (Open: accessible to all herbivores, Mega: excludes elephants and giraffes, Total: excludes all herbivores >=5 kg).     * dung_bait: identity of dung bait used in pitfall traps.      * plot_block: herbivore exclosure block (1-3) nested within level. * All other columns are dung beetle species names. Presence of each species in DNA samples is indicated by a "1". ### Data S4. 04_result_congruence.zip. A compressed (zip) file of 2 files required to examine congruence between results from the herbivore exclusion experiment and dung-beetle network. The .R file contains the script required to run the analyses on result congruence; .csv file contains complementary data merging network and experimental results.  **Data files:** **db_spp_pref_responses.csv:** dung beetle species responses to herbivore exclusion, linking data from dung-beetle food web and herbivore exclosures. Column Headings: * species: beetle species in dung-beetle food web detected in herbivore exclosures. * n_network: total number of individuals per species in dung-beetle food web. * degree_network: beetle species’ degree in dung-beetle food web.    * elephant_pref_network: fraction of individuals captured on elephant bait in dung-beetle food web. * elephant_only: binary factor denoting species captured only on elephant bait. * extinct_mega: binary response indicating presence or absence of beetle species in *Mega* plots.     * n_extinct_mega_blocks: number of blocks where a species was captured in an *Open* plot but not in the *Mega* plot of the same experimental block. * prop_occur_ctrl_plots: proportion of total occurrences per beetle species in *Open* plots. * prop_diff_sumPAperplot_excl_ctrl: difference in frequency of occurrence between *Open* and *Mega* plots.  Empty cells in columns 8-9 (prop_occur_ctrl_plots, prop_diff_sumPAperplot_excl_ctrl) correspond to not applicable. ### Data S5. 05_uhuru_ecosystem_services.zip. A compressed (zip) file of 2 files required to analyze the dung beetle ecosystem services data from herbivore exclusion plots. **Data files:** **1. uhuru_db_ecosystem_services.csv:** dung-beetle ecosystem service data across herbivore exclosures. Column Headings: * sample_ID: unique identifier for each experimental dung pat deployed in herbivore exclosures.    * plot_ID: unique identifier for each herbivore exclusion plot.            * level: location of herbivore exclosure (North vs. South sites). * plot_block: herbivore exclsoure block (1-3) nested within level.      * exclu_treatment: herbivore exclusion treatment (Open: accessible to all herbivores; Mega: excludes elephants and giraffes; Total: excludes all herbivores >=5 kg). * beetle_treatment: beetle exclusion treatment nested within herbivore exclosures (O = unmanipulated open treatment accessible to all beetles; OS = unmanipulated open treatment but premixed with artificial seeds; C = wrapped with coarse mesh to exclude large beetles; F = wrapped with fine mesh to exclude all beetles). * beetle_treatment_SOCF: beetle_treatment recoded for treatment clarity. * dung_rep: experimental dung pat replicate. * termites: binary factor denoting whether experimental dung pat had evidence of termite activity.  * prop_dryloss_generic: proportional dry mass loss estimated using 54 control dung pats dried in the lab (unitless). * sqrt_prop_dryloss_generic: proportional dry mass square-root transformed for analysis (unitless). * large_seed_diff: proportion of large seeds removed. * small_seed_diff: proportion of small seeds removed.              * mean_DB_abund_trap: average beetle abundance per trap in herbivore exclosures.       * mean_DB_mass_trap: average beetle dry mass per trap in herbivore exclosures. **2. uhuru_dung_controls.csv:** laboratory control dung pats used to determine initial dry mass. Column Headings: * UID: unique identifier for each control dung pat.         * dung_treatment: control.  * wet_weight: initial wet weight (g).  * dry_weight: final dry weight (g).  ### Data S6. 06_COI_ref_accessions.zip. A compressed (zip) file of 2 files, one listing GenBank accession numbers for all sequences included in the COI local reference library; another with raw species matches and associated metadata for each DNA sample analyzed from herbivore exclusion plots. **Data files:** **1. COI_ref_accessions.csv:** GenBank accessions for COI sequences included in dung beetle reference library. Column Headings: * process_ID: unique sequence identifier.        * seq_length (bp): sequence length (base pairs).           * accession_no: GenBank sequence accession number. * species: dung beetle species Latin name (BOLD barcode identification number used as an interim identifier for those containing “sp.” in species name). **2. uhuru_co1_raw_sample_data.xlsx**: raw COI data used to calculate beetle species richness in herbivore exclosures. Column Headings: **Sheet 1: metadata** * extraction_ID: unique identifier for each DNA sample from herbivore exclosures.   * trap_ID: pitfall trap operational unit; a unique identifier for each pitfall trap deployed in herbivore exclosures.                * level: location of herbivore exclosure (North vs. South sites). * exclu_treatment: herbivore exclusion treatment (Open: accessible to all herbivores; Mega: excludes elephants and giraffes; Total: excludes all herbivores 5kg). * dung_bait: identity of dung bait used in pitfall traps. * plot_block: herbivore exclosure block (1-3) nested within level. **Sheet 2: raw_sample_data** * id: unique identifier for each beetle sequence. * sequence: full. * phylum: beetle phylum.          * class: beetle class. * order: beetle order.     * family: beetle family. * subfamily: beetle subfamily.  * tribe: beetle tribe.        * genus: beetle genus.    * species: beetle species. * Columns 11-957: unique identifiers for each DNA sample from herbivore exclosures. Empty cells in taxonomic columns (e.g. tribe, genus) indicate missing data due to low taxonomic resolution. ### Data S7. 07_db_scale_comparison.zip. A compressed (zip) file of 3 files required to compare dung-beetle abundance across Mpala and two livestock-dominated sites. The .R file contains the script required to run comparative analyses; .csv files contain the beetle abundance and dung count data for each site. **Data files:** **1. mpala_db_comp.csv:** dung beetle abundance data across three sites ('Mpala', 'Lekiji', and 'Koija'). Column Headings: * site: sampling location. * trap_ID: pitfall trap operational unit; a unique identifier for each pitfall trap deployed at each site. * block: unique identifier of each site trap block (total 9 per site; AR1, AR2, AR2, BR1, BR2, BR3, CR1, CR2, CR3). * replicate: trap-block replicate identifier. * dung_bait: identity of dung bait used in pitfall traps. * total_abun: number of beetle individuals caught per trap. * total_dry_mass_g: total beetle dry biomass (g) per trap. **2. mpala_comp_dung_counts.csv:** dung-beetle preference data derived from cafeteria pitfall trap experiment. Column Headings: * site: sampling location. * block: unique identifier of each site trap block (total 9 per site; AR1, AR2, AR2, BR1, BR2, BR3, CR1, CR2, CR3). * dung_type: herbivore dung type.  * dung_counts: number of dung piles per block * dry_dung_pile_mass_g: species-specific dry dung pile mass estimates (g) collated from the literature.    * total_dung_mass: dung counts multiplied by dry_dung_pile_mass_g.