---

title: "Ecological pathways connecting riverine drought to community change across space and time"
author: Kyle Leathers
date: August 13, 2025
output: md_document

Manuscript abstract

Climate change is intensifying droughts via reduced snowpack and accelerated snowmelt in high mountains globally, altering community structure in snow-dependent rivers. To predict impending ecological change in rivers, we must understand the importance of the abiotic and biotic mechanisms connecting hydrologic change to biodiversity change–and whether these mechanisms operate similarly across space and time. Here, we studied abiotic effects of drought and invertebrate communities in a minimally disturbed watershed in California’s Sierra Nevada. Our study employed a highly-replicated design of 60 nested sites (capturing microhabitat to reach-level variation) and over two decades of change (2002 to 2023) in a subset of sites, including the driest period on record. We used Spatial Stream Network (SSN) models and autoregressive (AR) models to partition the spatial and temporal variance into covariate-driven vs. autocorrelation effects. Structural equation modeling allowed us to identify causal pathways connecting hydrologic change to invertebrate community change. We found that drought-driven variation in temperature, water velocity, and fine sediment all explained variation in abundance in over a third of the species in the community. Notably, the influence of abiotic effects differed across space and time: no taxa had their variance explained by the same abiotic effect in the same direction across space and time, and total spatial variance explained by abiotic effects for each species had no relationship with its temporal counterpart. We also found that community dissimilarity across space was poorly explained by abiotic effects, while temporal dissimilarity was driven by differences in temperature and water velocity causing species turnover. Finally, we tested the scale-dependency of our inferences by changing the extent and resolution of our data (resampling from seasonal to interannual; from microhabitat to watershed-level data), and found that pathways of community change varied depending on scale and on whether comparisons were made across space or time. These differences between space and time likely arise from some ecological drivers operating more strongly in one dimension, and from spatial and temporal autocorrelation in species abundances masking environmental effects. Our study illustrates that projecting riverine community composition under future hydroclimates requires accounting for mechanism context-dependency over space and time.

Description of the Data and file structure

• URL of dataset- https://doi.org/10.5061/dryad.2rbnzs7xs
• Lead author for the dataset- Kyle Leathers
• Title and position of lead author- PhD student
• Organization and address of lead author- University of California Berkeley. Address: Mulford Hall, Berkeley, CA, 94720
• Email address of lead author- Kyle_leathers@berkeley.edu
• Additional authors or contributors to the dataset - David Herbst, Michael Bogan, Gabriela Jeliazkov, Albert Ruhi
• Organization associated with the data - University of California Berkeley
• Funding- Funding was provided by the Sequoia Parks Conservancy (Sequoia Science Learning Center Research Grant). A.R. and K.L. were further supported by UC Berkeley new faculty startup funds and by the US National Science Foundation award #1802714
• License- CC0
• Geographic location – This study took place in the western Sierra Nevada, California in Bull Creek within the Kings River Experimental Watersheds
• Time frame - Begin date 6/1/2002
• Time frame - End date 8/15/2023
• General study design- We took aquatic macroinvertebrate samples from four reaches in Bull Creek that were sampled 11 times from 2002-2023, encapsulating a range of low and high precipitation years and associated streamflow levels during the summer. We tested spatial effects by comparing four reach samples taken within the same year. Temporal effects were determined by comparing the community from a year to every other year in the same reach. We also tested for space-time effects where a reach’s community was compared to every other reach community regardless of year or location. Additionally, we examined if abiotic mechanisms of drought effects and community assembly differed spatially at the microhabitat scale. To test this, we took fine scale microhabitat samples in 2020 using a nested spatial design. Ten reaches within Bull Creek (including the four long term reaches) were sampled in 2020 with six microhabitat samples taken from each reach. We ran a piecewise structural equation model to elucidate abiotic and biotic mechanisms driving low-flow effects on stream invertebrate community dissimilarity.
• Methods description- To answer our research questions, we used samples from four reaches in Bull Creek that were sampled 11 times from 2002-2023, encapsulating a range of low and high precipitation years and associated streamflow levels during the summer. These samples include those previously taken at the reach scale (i.e., ~50-100 m) between 2002 and 2015 (discussed in Herbst et al. 2019), and those we sampled from the same four long-term reaches every year from 2020 to 2023. All samples were collected in the summer after snowmelt allowed access to sample, between June and August. We tested spatial effects by comparing four reach samples taken within the same year. Temporal effects were determined by comparing the community from a year to every other year in the same reach. We also tested for space-time effects where a reach’s community was compared to every other reach community regardless of year or location. Additionally, we examined if abiotic mechanisms of drought effects and community assembly differed spatially at the microhabitat scale. To test this, we took fine scale microhabitat samples in 2020 using a nested spatial design. Ten reaches within Bull Creek (including the four long term reaches) were sampled in 2020 with six microhabitat samples taken from each reach. Microhabitat samples came from three paired pool-riffles, so that each reach had three pool and three riffle samples. Microhabitat sites were chosen to capture natural, intra-reach variation in abiotic mechanisms.

We calculated water temperature in our study as the average temperature 30 days before the sampling date of a year. We estimated water temperature before 2020 using historic air temperature records beginning in 2004 within the Bull Creek watershed. We found a strong linear relationship (i.e., R2 >0.95) between air and water temperature averaged over the 30 days before sampling in our data from 2020-2023, so we estimated past water temperature using the resulting regression equation from the relationship. Air temperature records were not available from 2002-2003, so we used Parameter-elevation Relationships on Independent Slopes Model (PRISM) downscaled air temperature daily estimates for the 30 days prior to sampling for those years (PRISM Climate Group 2024). We corrected for overestimation in PRISM air temperature estimates relative to recorded Bull Creek air temperature using a linear regression that had strong support. High frequency water temperature sensors were deployed prior to sampling in 2020 to capture average thermal conditions leading up to sampling. An MX2202 temperature sensor (Onset Computer Corporation) was placed in each microhabitat site at approximately July 11, over a month before invertebrate sampling. Multiple STIC (Stream Temperature, Intermittency, and Conductivity) loggers were placed in long term reaches from 2021-2023 (Chapin et al. 2014). Other abiotic mechanisms were measured at the time of sampling, including water velocity, the percent of fine sediment substrate, and discharge. Water velocity was estimated from 2002-2015 with a Global Water flow probe FP111 at the 50 transect sites. Water velocity in 2020 was measured five times in each microhabitat using a USGS pygmy flow meter connected to an Aqua CMD Current meter digitizer. After 2020, water velocity was estimated using a combination of the USGS pygmy flow meter and measuring the velocity of a floating leaf at five points in a cross-stream transect for each reach. Measurements were averaged for tests done at the reach scale. Prior to 2021, reach estimates of velocity and sediment were a weighted average of pool and riffle values based on the proportion of pool and riffle habitat. The proportion of fine substrate particles (i.e., less than 0.1 mm diameter) was estimated in every reach. Prior to 2020, visual estimates of fine sediment coverage were made 50 times in each reach divided among 10 cross-stream transects. A square foot grid with 25 equally spaced points was used in 2020 to estimate the substrate size under each point in the microhabitat prior to sampling. Substrate coverage was visually estimated from 2021-2023 prior to sampling; observations were taken every 10 m in the 100 m reach at each of the 11 sampling locations. Finally, we measured discharge in every reach using the velocity cross-sectional area method (Ode et al. 2016). Microhabitat discharge was assumed to be equal within a reach.

Stream invertebrates were sampled using a 250 μm D-frame net until 2020 and a 500 μm D-frame net beginning in 2021. This sampling difference likely did not affect our results; past work in streams has found that community samples collected with a 250 μm and 500 μm mesh are generally comparable (Herbst and Silldorff 2006, Buss and Borges 2008). All samples were stored in ~70% ethanol. We used a rotating-drum splitter on samples in the laboratory to split the sample into smaller fractions before sorting and identifying at least 500 individuals from each sample under a stereomicroscope. Subsamples were completely processed to avoid bias regarding the size of individuals picked and identified. Invertebrates were identified to the highest resolution possible, typically genus or species level. Total taxa abundance was corrected by multiplying the counted abundance by the inverse of the fraction of the sample identified. All samples at the reach scale were made comparable by correcting density estimates to 1 m2 and ensuring all aggregate reach samples contained a number of subsamples from riffle or pool habitats proportional to the prevalence of those habitats. The 2021-2023 samples consisted of 11 evenly spaced surber samples throughout a 150 m reach; sampling alternated between the right, center, and left of the channel. In other years, samples were taken and identified from pools and rifles separately, but the proportion of riffle:pool habitat area in the reach was recorded from 2002-2015. We used this proportion to calculate a weighted average of the aggregate reach community. Abundance was corrected in the same manner in 2020; the riffle pool ratio was approximated by matching discharge in 2020 with the historic year from 2002-2015 that had the closest discharge for each reach. The riffle pool ratio of the selected historic year at the same location could then be used to correct abundance records.

• Laboratory, field, or other analytical methods- In order to answer our first and second research questions, we first examined how taxa responded to abiotic mechanisms of drought across space at the microhabitat scale using a Spatial Stream Network (SSN) model. We used a SSN model for the most common taxa in the study (i.e., 19 taxa present in at least half the microhabitat sites and half of the reach samples taken over time) to test how much variation in taxa abundance was explained by spatial autocorrelation, abiotic mechanisms, or was left unexplained. We also examined how community abundance was explained by space and abiotic mechanisms using SSN models. To test the importance of temporal autocorrelation and covariate effects across time, we used generalized least squares (gls) models explaining variation in taxa abundance at the reach scale with abiotic mechanisms as explanatory variables and first order autoregression [AR(1)]. Standardized effects for each abiotic mechanism of drought from our SSN and gls results were compared using linear regression to determine if abiotic mechanisms have a consistent effect across space and time. We also compared spatial autocorrelation (i.e., upstream distance autocorrelation) and temporal autocorrelation (ɸ) of all taxa based on model results. We used the codyn R package to examine change in turnover, reordering, richness, and evenness between communities across space, time, and space-time (Avolio et al. 2019). We always used the RAC_difference function, even for temporal comparisons to ensure comparability and because we wanted temporal comparisons to include samples more than one time step away. For our final question, on the causal pathways ultimately connecting drought to community dissimilarity across space and time, we used piecewise structural equation models (pSEM). Each pSEM was composed of five linear models; four models predicted each RAC component by abiotic mechanism differences between communities, and one model explained community dissimilarity between communities with the RAC components. We made pSEMs comparing communities across space, time, and space and time at the reach scale, along with a pSEM for microhabitat comparisons across space. We calculated the absolute difference in mean water temperature, water velocity, and percent of fine sediment between two samples to use the delta as covariates in models.
• Quality control- Data was recorded in hard copies and digitally to reduce the risk from mistyped data. Data was plotted to look for outliers that were erroneous and paper copies were referenced to ensure values were correct.

Description of files used in analysis and included on Dryad

Details for: Biotic_KREW_SSN_Table_dryad.csv

• Dataset description: Csv file of dataframe extracted from SSN object. The .ssn file is needed to run an SSN and can be found by unzipping the lsn.ssn.zip file.

Variables

• FullCode: Idenifying code of each 60 microhabitat sites. The first number refers to the reach (also referred to as cluster). the second letter distinguishes a pool (P) from a riffle (R). The last number names the riffle-pool pair the site was in.
• Cluster: Reach number
• Pool_Riffl: Distinguishes a pool (P) from a riffle (R)
• Pool_Rif_1: The riffle-pool pair the site was in.
• Fraction: Fration of sample that was identified
• Columns 6-155: Each column is the abundance of a taxa found at the microhabitat scale. Names are shortened due to ArcGIS constraints in places, but name corrections are listed below. Otherwise column name is the accurate taxonomic name
  • Acariforme ->Acariformes
    -Amiocentru -> Amiocentrus_aspilus
    -Ampumixis_ -> Ampumixis_dispar
  • Apsectrota -> Apsectrotanypus
  • Atrichopog -> Atrichopogon
  • Bezzia_sen -> Bezzia_sensu_lato
  • Brundiniel -> Brundiniella
  • Calineuria -> Calineuria_californica
  • Canthocamp ->Canthocamptidae
  • Ceratopogo->Ceratopogon
  • Ceratopsyc->Ceratopsyche
  • Cordulegas->Cordulegaster_dorsalis
  • Corynoneur->Corynoneura
  • Cricotopus->Cricotopus_Orthocladius
  • Despaxia_a->Despaxia_augusta
  • Dysmicoher ->Dysmicohermes
  • Ecclisomyi->Ecclisomyia
  • Ephemerell->Ephemerella
  • Eukiefferi->Eukiefferiella_brehmi
  • Eukieffe_1->Eukiefferiella_brevicalcar
  • Eukieffe_2 ->Eukiefferiella_devonica
  • Eukieffe_3->Eukiefferiella_gracei
  • Hesperoper->Hesperoperla_pacifica
  • Heterocaud->Caudatella heterocaudata
  • Heteroplec->Heteroplectron_californicum
  • Heterotany->Heterotanytarsus
  • Heterotris->Heterotrissocladius
  • Hydrobaenu->Hydrobaenus
  • Hydropsych->Hydropsyche
  • Isoperla_s->Isoperla_sobria
  • Kogotus_Ri->Kogotus-Rickera group
  • Krenosmitt->Krenosmittia
  • Lepidostom->Lepidostoma
  • Macropelop->Macropelopia
  • Micropsect ->Micropsectra
  • Microtendi->Microtendipes_pedellus
  • Microten_1->Microtendipes_rydalensis
  • Monodiames ->Monodiamesa
  • Nanocladiu->Nanocladius_balticus
  • Nanoclad_1->Nanocladius_parvulus
  • Nematomorp ->Nematomorpha
  • Ochrotrich->Ochrotrichia
  • Oligochaet->Oligochaete
  • Optioservu->Optioservus
  • Ordobrevia->Ordobrevia_nubifera
  • Parakieffe->Parakiefferiella
  • Paraleptop->Paraleptophlebia
  • Parametrio->Parametriocnemus
  • Paraphaeno->Paraphaenocladius
  • Paratendip->Paratendipes
  • Parorthocl->Parorthocladius
  • Phaenopsec->Phaenopsectra
  • Polycentro->Polycentropus
  • Polypedilu->Polypedilum_fallax
  • Polypedi_1->Polypedilum_laetum
  • Polypedi_2->Polypedilum_scalaenum
  • Psectrocla->Psectrocladius_sordidellus
  • Pseudochironomus->Pseudochironomus
  • Pseudodiam ->Pseudodiamesa
  • Psychoglyp->Psychoglypha
  • Rheocricotopus->Rheocricot
  • Rheotanyta ->Rheotanytarsus
  • Rhyacophil->Rhyacophila_alberta
  • Rhyacoph_1->Rhyacophila_atrata
  • Rhyacoph_2 ->Rhyacophila_betteni
  • Rhyacoph_3->Rhyacophila_brunnea
  • Rhyacoph_4->Rhyacophila_grandis
  • Rhyacoph_5 ->Rhyacophila_narvae
  • Sanfilippo->Sanfilippodytes
  • Serratella->Serratella_levis
  • Stempellin->Stempellina
  • Stempell_1->Stempellinella
  • Symposiocl->Symposiocladius
  • Synorthocl->Synorthocladius
  • Thienemani->Thienemanniella_cf_xena
  • Thienema_1->Thienemanniella_fusca
  • Thienneman ->Thiennemannimyia
  • Turbellari->Turbellaria
  • Tvetenia_b ->Tvetenia_bavarica
  • Tvetenia_d->Tvetenia_discoloripes
  • Yoraperla_->Yoraperla_nigrisoma
  • Yphria_cal->Yphria_californica
  • Zavrelimyi->Zavrelimyia
• ID2: Idenifying code of each 60 microhabitat sites. The first number refers to the reach (also referred to as cluster). The second number names the riffle-pool pair the site was in. The last letter distinguishes a pool (P) from a riffle (R).
• Substrate_: Fraction of fine sediments in the site substrate.
• Substrate1: Fraction of cobble sediment in the site substrate.
• Substrate_1: Fraction of sand sediment in the site substrate.
• SubstrateC: Fraction of substrate not covered by anything in the site substrate.
• Substrat_2: Fraction of substrate covered by a thin layer of fine sediment in the site substrate.
• Substrat_3: Fraction of substrate covered by wood or woody debris in the site substrate.
• Substrat_4: Fraction of substrate covered by algae in the site substrate.
• Conductivi: Conductivity measured in microsiemens/cm.
• Canopy: Proportion of canopy that is covered.
• Width_cm: Width of the stream at the microhabitat site (cm)
• Velocity_c: Average water velocity (cm/s) at microhabitat
• Depth_in: Average water depth (inches) at microhabitat
• P_R: Whether the site is a pool (P) or riffle (R) microhabitat
• P_R_N: Pool-riffle pair number within the reach. Adjacent pool-riffles have the same number
• Reach: Code for reach the sample was taken from.
• mean_month: The average temperature 30 days before the sampling date, in degrees Celsius.
• max_month_: The average maximum temperature 30 days before the sampling date, in degrees Celsius.
• min_month_: The average minimum temperature 30 days before the sampling date, in degrees Celsius.
• elevation_: Elevation of the microhabitat site in meters
• Q_L_s: Water discharge in Liters per second
• r_meanT: Reach average of mean_month
• r_maxT: Reach average of max_month_
• r_minT: Reach average of min_month_
• Stream_siz: stream size category, including mainstem (M), intermediate sized (I), and headwater (H)
• Cluster_P_: Code for cluster/reach number and whether the microhabitat is a pool or riffle
• alpha: species richness of the site sample
• Total_N: Total abundance estimated in site sample
• EvennessJ: Shannon's Evenness (Pielou’s J)
• Lat_1: Latitude of site
• Long_2: Longitude of site
• r_cond: Average reach conductivity measured in microsiemens/cm
• r_canopy: Average reach proportion of canopy that is covered
• NEAR_FID: The ObjectID of the nearest feature
• NEAR_DIST: The distance from the input feature to the near feature
• NEAR_X:The x-coordinate of the location on the near feature that is closest to the input feature.
• NEAR_Y:The y-coordinate of the location on the near feature that is closest to the input feature.
• NEAR_ANGLE:The angle of the line at the FROM_X and FROM_Y location that connects the input features to the near feature.
• rid: Reach identifier unique to each stream line segment. Created in STARS
• ratio: the distance ratio from the end of an edge to the point location. Created in STARS
• upDist: The upstream distance between the stream outlet (i.e., the most downstream location in the stream network) and each of the edges and sites.
• afvArea: Proportion of watershed stream length that is upstream of this site. 1 is the watershed outlet and 0 is the origin edge of a tributary with nothing upstream.
• locID: location ID assigned in STARS.
• netID: Network identifier (netID) assigned to the edges, sites, and prediction sites attribute tables to differentiate between two edges with the same binary ID. Made in STARS.
• pid: point ID assigned in STARS.

Details for: Bull_TimeSeries.csv

• Dataset description: The dataset includes reach scale abundances of macroinvertebrates and abiotic conditions of reaches in the study.

Variables

• Reach: Code for reach the sample was taken from.
• Lowest: Taxonomic identification of invertebrate
• Reach_Abundance_m2: Invertebrate density (# individuals estimated per square meter)
• Reach_velocity_cm_s: Average reach water velocity in cm/s
• Reach_pct_f: Average percent of fine sediments in the reach (i.e., <0.1 mm diameter)
• Q_L_s: Water discharge in Liters per second
• Year: Year sample was collected
• Reach_tempC:The average temperature 30 days before the sampling date, in degrees Celsius.

Details for: Drought_covariates.csv

• Dataset description: Microhabitat abiotic conditions used in determining abiotic differences for RAC analysis.

Variables

• FullCode: Idenifying code of each 60 microhabitat sites. The first number refers to the reach (also referred to as cluster). the second letter distinguishes a pool (P) from a riffle (R). The last number names the riffle-pool pair the site was in.
• Velocity_cm_s: Average microhabitat water velocity in cm/s
• Substrate_F_P: Percent of fine sediments in the microhabitat site
• Q_L_s: Water discharge in Liters per second for the reach. Q is assumed to be constant in a reach.
• r_meanT: Mean water temperature at the reach scale a month before the sampling date, in degrees Celsius
• max_month_water_temp: the average of each daily maximum water temperature in the 30 days prior to sampling at the microhabitat scale
• mean_month_water_temp: Mean water temperature at the microhabitat scale a month before the sampling date, in degrees Celsius. Data at sites where a sensor was lost is the reach average value.

Details for: KREW_biotic_data3.csv

• Dataset description: The dataset includes the biotic data for microhabitat samples in 2020.

Variables

• FullCode: idenifying code of each 60 microhabitat sites. The first number refers to the reach (also referred to as cluster). the second letter distinguishes a pool (P) from a riffle (R). The last number names the riffle-pool pair the site was in.
• Lowest: Finest resolution identification for an invertebrate group
• Cluster: Reach number
• Pool_Riffle: Distinguishes a pool (P) from a riffle (R)
• Pool_Riffle_N: The riffle-pool pair the site was in.
• Fraction: Fraction of the original sample sorted identified. At least ~500 individuals were identified from each sample, or the entire sample was identified if fewer than 500 individuals were present.
• Abundance_corrected: The estimated total abundance the taxon in the sample. The number of individuals was multiplied by the inverse of the fraction of the sample that was sorted.

Details for: Bull_Time_Series_drought_covariates.csv

• Dataset description: The dataset includes abiotic conditions at the reach scale for the four long term reaches in the study.

Variables

• Reach: Code for reach the sample was taken from.
• Year: Year sample was collected in.
• Reach_pct_f: Average percent of fine sediments in the reach (i.e., <0.1 mm diameter)
• Reach_tempC: The average temperature 30 days before the sampling date, in degrees Celsius.
• Reach_velocity_cm_s: Average reach water velocity in cm/s
• Q_L_s: Water discharge in Liters per second for the reach. Q is assumed to be constant in a reach.

Details for: response_summary_combined.csv

• Dataset description: The dataset includes the results of SSN models explaining variation in community abundance, richness, and evenness.

Variables

• FactorLevel: Model term, one of model intercept, water velocity (Velocity_c), fine sediment cover (Substrate_), or mean water temperature (mean_month)
• Estimate: Effect size
• std.err: Standard error of effect size
• t.value: T-value
• prob.t: p-value of factor.
• response: Biotic community metric which was explained in the model.

Details for: Kings_Q_july_dryad.csv

• Dataset description: The dataset includes discharge data over time for the four long term Bull Creek reaches and the King's River to which Bull Creek is a tributary.

Variables

• Q_av: Average July discharge (L/s)
• Q_av_log: Log of Q_av. NA values are present when data was not collected or used in analysis (2017-2019) and when discharge measurements were not taken (2004-2006).
• year: Year
• Location: Location, either Bull Creek reach or Kings River
• Reach: Reach name within Bull Creek. The letter refers to whether the reach is a mainstem (M), intermediate sized (I), or headwater (H). Reach is NA for the Kings River because there are no reach designations there.
• Year_Q_type: North Fork Kings River data comes from site 11218400 which had over 80 years of data. This was used to classify water years by the percentile of discharge in the time series, where drought is less than 10% and High is above 90%. Year Q type was not determined using Bull Creek reaches, so these values are left as NA.

Details for: KREW_Reach_distance_df.csv

• Dataset description: The dataset includes the distance between all reaches studied in Bull Creek

Variables

• Reach1: First reach compared
• Reach2: Second reach compared
• distance_m: distance in meters between reaches

Details for: taxa_varcomp_up_combined.csv

• Dataset description: The dataset includes results from SSN models explaining spatial variation in the most common taxa within Bull Creek across both space and time (found in over 50% of samples)

Variables

• FactorLevel: Model term, one of model intercept, water velocity (Velocity_c), fine sediment cover (Substrate_), or mean water temperature (mean_month)
• Estimate: Effect size
• std.err: Standard error of effect size
• t.value: T-value
• prob.t: p-value of factor.
• taxa: Taxonomic identification of invertebrate
• Covariates (R-sq): The proportion of variance explained by drought covariates
• Exponential.tailup: The proportion of variance explained by spatial upstream autocorrelation
• Nugget: The proportion of variance that is left unexplained

Details for: Taxa_autocorrelation_df_comb.csv

• Dataset description: The dataset includes drought environment effects and the influence of autocorrelation across space and time on macroinvertebrate abundance. Results from both SSN and GLS models are provided.

Variables

• taxa: Taxonomic identification of invertebrate
• Parameter: Model term, one of water velocity, fine sediment cover, or mean water temperature
• Estimate_space: Effect size of SSN spatial model
• prob.t: p-value of factor.
• Estimate_direction_space: Indicates whether the spatial effect was not significant, positively significant (Positive), or negatively significant (Negative)
• Exponential.tailup: The proportion of variance explained by upstream distance autocorrelation in the SSN model
• Covariates..R.sq.: The proportion of variance explained by drought covariates in the SSN model
• Temp_sd: Standard deviation of mean water temperature across space
• Substrate_sd: Standard deviation of substrate across space
• Velocity_sd: Standard deviation of water velocity across space
• SD_taxa: Standard deviation of taxa abundance across space
• standardized_effect_space: Spatial effect size standardized by using the ratio of the standard deviations of the predictor and response variables, for increased comparability.
• Value: Effect size of GLS temporal model
• Std.Error: standard error of Value
• t-value: T-value
• p-value: p-value of factor.
• lower: Lower confidence interval for Value
• Phi: Temporal autocorrelation in the GLS model
• upper: Upper confidence interval for Value
• Rsq_df: Variation explained by the drought covariate in the GLS model
• Estimate_direction_time: Indicates whether the temporal effect was not significant, positively significant (Positive), or negatively significant (Negative)
• SD_taxa_time: Standard deviation of taxa abundance across time
• Temp_sd_time: Standard deviation of mean water temperature across time
• Substrate_sd_time: Standard deviation of substrate across time
• Velocity_sd_time: Standard deviation of water velocity across time
• standardized_effect_time: Temporal effect size standardized by using the ratio of the standard deviations of the predictor and response variables, for increased comparability.

Details for: RAC_B_C_sub.csv

• Dataset description: The dataset includes the results of rank abundance curve comparisons of communities using RAC_difference across time, space, and space-time.

Variables

• composition_diff: Difference value between 0 and 1. 0 indicates identical composition while 1 indicates a complete difference betwwn the communities.
• Comparison: Whether the communities were compard at the reach scale (across Space, Time, or Space-Time) or the microhabitat scale across space (Space_2020)
• FullCode: Microhabitat identification code of the first site in the comparison. The first number refers to the reach (also referred to as cluster). the second letter distinguishes a pool (P) from a riffle (R). The last number names the riffle-pool pair the site was in.
• FullCode2:Microhabitat identification code of the second site in the comparison. The first number refers to the reach (also referred to as cluster). the second letter distinguishes a pool (P) from a riffle (R). The last number names the riffle-pool pair the site was in.
• Reach_Year: Reach and year sampled of the first site in the comparison. The first letter indicates reach size, whether the reach is a mainstem (M), intermediate sized (I), or headwater (H). NA values are present in microhabitat Space_2020 comparisons because comparisons are made at the micrhabitat rather than reach scale.
• Reach_Year2:Reach and year sampled of the second site in the comparison. The first letter indicates reach size, whether the reach is a mainstem (M), intermediate sized (I), or headwater (H). NA values are present in microhabitat Space_2020 comparisons because comparisons are made at the micrhabitat rather than reach scale.

Details for: RAC_Output_Intra_annual_time3.csv

• Dataset description: The dataset includes the results of rank abundance curve comparisons of communities using RAC_difference across time in intra-annual comparisons of the Summer and Fall in Bull Creek within four long term reaches.

Variables

• Year: Year samples were collected
• Reach: Reach name within Bull Creek. The letter refers to whether the reach is a mainstem (M), intermediate sized (I), or headwater (H)
• species_diff: RAC_difference output from comparing species identity (Turnover). Difference value between 0 and 1. 0 indicates identical composition while 1 indicates a complete difference betwwn the communities.
• evenness_diff_abs: RAC_difference output from comparing community evenness. Difference value between 0 and 1. Zero indicates identical composition while 1 indicates a complete difference betwwn the communities.
• richness_diff_abs: RAC_difference output from comparing species richness differences between two communities. Difference value between 0 and 1. 0 indicates identical composition while 1 indicates a complete difference betwwn the communities.
• rank_diff: RAC_difference output from comparing ordering of species abundance in community (Reordering). Difference value between 0 and 1. 0 indicates identical composition while 1 indicates a complete difference betwwn the communities.
• Comparison: Comparison was made across time.
• Substrate_F_P_siteS: Summer average reach fraction of fine sediments in the site substrate.
• r_meanT_siteS: Summer average reach water temperature (C) in the prior month.
• Velocity_cm_s_siteS: Summer average reach water velocity (cm/s).
• Q_L_s_siteS: Summer reach discharge in (L/s).
• Velocity_cm_s_siteF: Fall average reach water velocity (cm/s).
• Substrate_F_P_siteF: Fall average reach fraction of fine sediments in the site substrate.
• Q_L_s_siteF: Fall reach discharge in (L/s).
• r_meanT_siteF: Fall average reach water temperature (C) in the prior month.
• Velocity_delta: Difference in water velocity between the Summer and Fall.
• Substrate_F_delta: Difference in fine sediment cover between the Summer and Fall.
• Q_L_s_delta: Difference in discharge between the Summer and Fall.
• r_meanT_delta: Difference in water temperature between the Summer and Fall.
• composition_diff: Community dissimilarity between communities. Zero indicates identical composition while 1 indicates a complete difference betwwn the communities.

Details for: space_time_50_taxa.csv

• Dataset description: Table of taxa that are present in at least 50% of the spatial microhabitat samples and 50% of the temporal reach samples.
• Lowest: Finest resolution taxonomic identification
• n_space: Number of samples that contain the taxa in the spatial microhabitat data. Over 29 samples present needed to have at least half
• total_abundance: Total spatial abundance across all spatial samples
• mean_abundance: Average abundance of the taxa in a spatial microhabitat based on sites that had at least one individual.Equivalent to total_abundance divided by n_space.
• n: Number of samples that contain the taxa in the temporal data. Over 21 samples present needed to have at least half

Details for: lsn.ssn.zip

• Dataset description: Contains SSN geospatial object that is analyzed in KREW_biotic_SSN_dryad.R using previous version of R. R version 4.2.1 and earlier should be able to analyze the file. A more detailed description of the .ssn directory and its contents is provided in Peterson and Ver Hoef (2014). This was made using the Spatial Tools for the Analysis of River Systems (STARS) toolset. More information available from Peterson, E. E. and Ver Hoef, J. M. 2014. STARS: An ArcGIS Toolset Used to Calculate the Spatial Information Needed to Fit Spatial Statistical Models to Stream Network Data . Journal of Statistical Software 56(2): 1–17.

This directory contains:

• edges.gpkg: edges in GeoPackage format. A network identifier, netID, is added that is unique to each subnetwork.
• sites.gpkg: observed sites in GeoPackage format (if present). Three new ID columns are added that are unique to the measurement (pid), the location (locID), and the network (netID).
• netID.dat files for each distinct network, which store the binaryID values for line segments in edges.
• Distance between sites in dist.net

Code/Software

Code is available on Zenodo at https://doi.org/10.5281/zenodo.13324182.

KREW_biotic_dryad.R contains the code used to process, analyze, and plot data for the manuscript generally. Questions about the code can be addressed to Kyle Leathers - kyle_leathers@berkeley.edu

KREW_biotic_SSN_dryad.R contains the code used to process and analyze Spatial Stream Network models for individual species and community metrics. Questions about the code can be addressed to Kyle Leathers - kyle_leathers@berkeley.edu

Code can be cited as:
Leathers, K. W., Herbst, D., Bogan, M. T., Jeliazkov, G., Ruhi, A. 2025. Ecological pathways connecting riverine drought to community change across space and time. Zenodo, code. https://doi.org/10.5281/zenodo.13324182.

Sharing/access Information

Data in csv files uploaded on Dryad is available for other use. Please cite data as:
Leathers, K. W., Herbst, D., Bogan, M. T., Jeliazkov, G., Ruhi, A. 2025. Ecological pathways connecting riverine drought to community change across space and time. Dryad, dataset. https://doi.org/ 10.5061/dryad.2rbnzs7xs.