Introduction

Coastal and inland flooding can facilitate the novel spread of aquatic invasive species (AIS) into drainages during and after severe storm events. Flooding can transfer species longitudinally and laterally within a drainage (Amoros and Borentte 2002; Bunn and Arthington 2002), yet a relatively understudied lateral pathway is when a flood temporarily connects adjacent drainages by breaching a divide. For example, during the ‘Great Flood of 1993,’ drainages along Upper Mississippi River basin were connected by floodwaters leading to dispersal of many fish, invertebrates, and plants species (Allen 1993; Galat et al. 1998). Lateral connections between hydrologic units can also be caused by storm surge along coasts where sea levels temporarily increase hydrologic connectivity along shorelines, bays, and estuaries for up to several meters of elevation inland (Guntenspergen et al. 1995) leading to downstream spread of aquatic species into bays and estuaries (Hill et al. 2021). Species dispersal during a storm surge is contingent on its salinity tolerance and the possibility of reaching another waterbody (Schofield et al. 2007; Fuller et al. 2010; Ożgo et al. 2016). Storm surge also causes large-scale habitat loss or alteration from shoreline migration and vegetation scouring (Sirianni et al. 2022; Kroes et al. 2023). As global climate conditions continue to change, the magnitude and frequency of storm events and flooding are expected to increase (Brooks 2013; Ornes 2018), enhancing the potential for the spread of AIS.

Several AIS have been found to disperse during floods and are considered invasive to respective regions within the U.S., including fishes, invertebrates, and vascular plants (Dean et al. 2024a; Dean et al. 2024b). Examples of invasive fish species include Grass Carp, Ctenopharyngodon idella (Bain et al. 1993), Silver Carp, Hypophthalmichthys molitrix, Bighead Carp, H. nobilis (DeGrandchamp et al. 2007), Flathead Catfish, Pylodictis olivaris (Jones and Noltie 2007), and Nile Tilapia, Oreochromis niloticus (Schofield et al. 2007). Flood-dispersed invasive invertebrate examples include Giant Applesnail, Pomacea canaliculata (Seuffert and Martin 2012), Golden Mussel, Limnoperna fortunei (de Amo et al. 2021), and Australian Redclaw Crayfish, Cherax quadricarinatus (Williams et al. 2001). Numerous aquatic invasive plants can be transported on flood waters (Nilsson et al. 2002, 2010; Boedeltje et al. 2003). Examples of flood-dispersed invasive plants include Giant Salvinia, Salvinia molesta (Fuller et al. 2010), Alligatorweed, Alternanthera philoxeroides (Phillips 2022), Common Reed, Phragmites australis (Minchinton 2002; Galatowitsch et al. 2016), and Waterprimrose Willow, Ludwigia hexapetala (Skaer Thomason et al. 2018). The life history traits of flood-dispersed species could aid in opportunistic, flood-induced dispersal across drainage divides. There is a need to understand which AIS could spread via flood water to aid managers in combatting AIS spreading across distinct drainages and management jurisdictions.

To help stakeholders address these collective needs, the U.S. Geological Survey (USGS)’s Nonindigenous Aquatic Species (NAS) database team created novel Flood and Storm Tracker (FaST) maps. The NAS database team publicly released their first FaST map after the flood event caused by Hurricane Harvey in Texas in 2017 (Pfingsten et al. 2017). As of the summer of 2023, the NAS database team has published 16 FaST maps for 14 hurricanes and one major flood event since the summer of 2017 (https://nas.er.usgs.gov/viewer/Flooding). For each flood event, a FaST map highlights locations of low elevation points along drainage divides where high flood waters potentially exceeded the divide elevation to create temporary corridors for freshwater aquatic species dispersal into neighboring drainages. The maps also show pre-storm occurrences of non-native aquatic taxa to indicate species that potentially spread during flooding. Managers can use FaST maps to prioritize early detection and rapid response efforts on AIS, a national goal of the Federal government and its partners (Reaser et al. 2020), by locating areas indicated on FaST maps that show where invasive species potentially dispersed to new watersheds.

In this study, our goal was to describe the development of FaST maps to highlight the risk of non-native aquatic taxa dispersing across drainage divides by flood waters in the United States and its island territories. This involved three research objectives: (1) identifying regions threatened by storms by their hydrologic drainage units and pre-storm occurrences of non-native aquatic taxa in each region; (2) using USGS water height data and a storm surge model to determine if flooding connected adjacent drainage basins; and (3) determining locations of potential dispersal corridors (i.e., hydrologic connections) between river basins that non-native aquatic taxa might use for migration in the flood-impacted area. We made two assumptions in the construction of FaST maps: (1) at a certain threshold above ground level and proximity to a drainage basin border, inter-basin transfer of water occurs, and (2) given access to a water corridor with sufficient height above ground, aquatic taxa have a non-zero probability of dispersing along the corridor. Collectively, the FaST maps are a first effort to address a relatively understudied pathway of AIS spread across a large landscape and provide managers with information to prioritize AIS tracking at locations impacted by storms and flooding.

Methods

Study area

Our study area included portions of the conterminous U.S., Alaska, Hawai’i, and U.S. territories that were impacted by severe storms that resulted in major flooding across multiple watersheds. Areas most frequently impacted by flooding included peninsular Florida, the northern Gulf of Mexico, the Atlantic coast, and Puerto Rico. While hurricanes and tropical storms combined with storm surge led to the most severe flooding, major rain events that led to flooding were also considered. We defined major rain events as having the potential for inter-basin transfer of flood waters and aquatic taxa. For example, the Midwest (specifically Iowa, Nebraska, Oklahoma, and Arkansas) received significant flooding in 2019 due to historic precipitation levels.

Spatial framework

Our spatial framework was the USGS Watershed Boundary Dataset, which defines discrete hydrologic units as the extent of surface waters that drains to an outlet or connection point within a dentritic stream network, or to multiple outlet points within a non-dendritic network (USGS 2019b; Jones et al. 2022). HUCs are not independent units as many are longitudinally connected at stream confluences within the network or comprise an entire lake with multiple connected adjacent HUCs (e.g., Lake Erie). We used 10-digit HUCs (watersheds) as the standard drainage delineation for each map with a few exceptions. Early in FaST methodology development, we used 8-digit HUCs (subbasins) in 2017 for Hurricanes Harvey, Irma, and Nate due to consistency with 8-digits HUCs being a standard scale in the NAS database (Daniel et al. 2017, 2018). We used 12-digit HUCs (subwatersheds) in 2018 for Hurricane Lane because the Hawaiian Islands had relatively fewer 8- and 10-digit HUCs compared to the U.S. mainland, while the islands could be divided into the most HUCs at the 12-digit scale. We used the dates of each flood event provided by the USGS Flood Event Viewer (USGS 2023a) for the temporal framework of flood duration in our maps. Digital elevation models for the conterminous U.S. and its territories were obtained from the USGS 3D Elevation Program (USGS 2021). Although higher resolution data (e.g., 1–3 m spatial resolution) could be used at a local or regional level, we used 1/3 arc-second grid resolution (~ 10 m spatial resolution) at HUC borders due to coverage availability for the entire U.S. Elevation measurement error was assumed to be within 0.5 m. Maps were developed using Esri ArcMap 10.7.1 (ESRI 2020).

Biological data

We obtained pre-flood occurrences of aquatic species using the NAS database for each FaST map (USGS 2023b; see Supplemental Information for a full list of aquatic taxa from FaST maps). Species records included freshwater-obligate vertebrates, invertebrates, and plants. Species records from the NAS database date back to the 1790s and include occurrence information for all U.S. states and territories. We included taxon records from established inland populations (omitting marine or mammal species with some marine bivalve and crustacean exceptions) that were not eradicated or extirpated and that were present prior to respective flood events. We included occurrence records with spatial accuracy at least approximated to localities in the flood-impacted areas, while we excluded records with spatial accuracy approximated to centroids of states, counties, and HUCs.

Flood data

Flood data included coastal storm surge and inland flooding data (see Supplemental Information for hyperlinks to source material). The flood data used in our models differed between inland and coastal systems and varied by storm. For each flood event, we modeled flood conditions with observed information available from existing products. For inland systems, we used USGS streamgage water heights above flood stage (Eberts et al. 2019; USGS 2022) and USGS high-water marks with approved (i.e., non-provisional) heights above ground level (Koenig et al. 2016; USGS 2023a). Water level measurement error was assumed to be within 0.5 m.

For coastal systems, observed storm surge heights above ground level were provided by USGS storm-tide sensors deployed prior to major hurricanes (McCallum et al. 2013). Storm surge water level estimation error was assumed to be within 1.0 m. We relied on modeled storm surge data when observational data for coastal flooding was absent. Storm surge was estimated using gridded outputs from National Oceanographic and Atmospheric Administration’s (NOAA) Sea, Lake, and Overland Surges from Hurricanes (SLOSH; NOAA 2022b) model. SLOSH is a low-resolution hydrodynamic model created by the National Weather Service to estimate tropical storm surge elevations based on hypothetical, historical, and predicted storms (Jelesnianski et al. 1992; Glahn et al. 2009). The grids have cells of varying size, with smallest size closest to the shore (e.g., a range of 120–2200 m for the New Orleans, Louisiana basin). SLOSH uses storm parameters (pressure, location, direction, speed, and radius of maximum winds) to calculate storm surge and winds on each cell of a gridded basin. SLOSH does not include rainfall quantities, river flow, or wind-driven waves, but accounts for the initial tide level (usually a mean and a high tide level; Jelesnianski et al. 1992). The accuracy of the SLOSH surge heights forecasts has been assessed and found to be within 20% for significant surge height values when compared with high-water marks for multiple storms (Jelesnianski et al. 1992; Glahn et al. 2009; Jarvinen and Lawrence 1985). Wind fields predicted by SLOSH were, in most cases, within 6% of observed values (Houston et al. 1999). When compared to a high-resolution hydrodynamic model, ADCIRC, the surge values from SLOSH were different by 12% on average, for the most dangerous storms simulated (Lin et al. 2010).

Developing FaST maps

The initial flood-impacted areas were estimated based on the storm paths for hurricanes accessed from the NOAA National Hurricane Center (NOAA 2022a). That area was further refined by combining high water flow data from USGS streamgage heights above flood stage and above-ground water height data from high-water marks, storm-tide sensors, and the NOAA SLOSH model. For each FaST map, we plotted the locations of streamgages above flood stage and water heights above ground level (both included respective measurement errors of 0.5 m) by HUC along the inland storm path. HUCs were selected and data were collected for areas that had storms pass over/along the coastline (e.g., a storm entered land or re-entered the ocean). Initial draft maps were created within less than twenty days of major storm or flood event. These maps were then revised a few months later if more data were collected from the flood-impacted area. We quantified flood frequency across FaST maps by comparing the number of overlapping HUCs in flood-impacted areas for all FaST maps in our study.

Longitudinal connections between an upstream to downstream HUC-pair were modeled when two neighboring HUCs had hydrologic connections determined by the downstream flow network in the USGS National Hydrography Dataset (USGS 2019a). We also modeled downstream movement at reservoirs with streamgage heights above flood stage due to water being released from dams to reduce upstream flood risk. The flow of flood waters and thus potential species spread via longitudinal connections was assumed unidirectional from upstream to downstream HUCs along streams (Fig. 1). While we acknowledge the possibility for fishes to actively swim upstream, we choose to predict downstream-only movement for all taxa at longitudinal connections due to high-flow conditions during floods creating an impediment to upstream movement. Active swimmers (e.g., salmonids) would be just as likely to swim upstream during non-flood conditions unless the flood waters overcame physical barriers such as dams. Although we tracked the longitudinal connections by selecting downstream HUC pairs to be highlighted on the FaST maps, we did not assign a point on the maps to represent a longitudinal connection and we also did not include longitudinal connections from HUCs that lacked a downstream connection.

Fig. 1
figure 1

Theoretical Flood and Storm Tracker map for a coastal area impacted by flood waters. Species occurrences (red circles) are present prior to a flood event. Selected (dark blue) HUC10 (hydrologic unit code 10) is the Cooper River drainage in South Carolina

Full size image

Bidirectional lateral connections between neighboring HUCs are an important flood connection for this study as they provide opportunities for novel species dispersal into drainages where hydrologic connections are absent under non-flood conditions. We defined lateral connections using water level observations and model-derived water heights greater than the most proximal HUC border elevations (Fig. 1). For the observed water heights, we used storm-tide sensor or high-water mark heights with water levels 0.5 m (measurement error) greater than the HUC border elevation within a 1.0 km radius of where water heights were measured. We chose a maximum 1.0 km distance from measured water heights to HUC borders assuming that water heights would be uniform at that distance. This 1.0 km flood radius from measured water heights does not account for landscape topography between the elevations at the water height measurement and the HUC border. Instead, we assume elevation gradients (i.e., slopes) to be low enough between those two elevations that major flooding could occur within that 1.0 km distance. We did not consider streamgage water levels for modeling lateral connections due to the ambiguity in interpreting flood stage at each streamgage (i.e., streamgage water levels at or above a flood stage is not synonymous with the level at which water breaches stream banks and overflows). For model-derived water heights, we used SLOSH-modeled storm surge heights greater than 1.0 m (measurement error) above the HUC border elevation where model-derived heights intersected the HUC border. We chose a conservative assumption for lateral connections with the modeled storm surge heights compared to actual water height measurements because of the larger uncertainty in the SLOSH model results compared to field measurement error.

Non-native aquatic taxa observed in flood-impacted areas prior to flood events were included in each map (e.g., Fig. 2). For each taxon, HUCs in FaST maps were delineated by introduction status, i.e., either the HUC had that taxon present prior to the flood event or that taxon could potentially be introduced to the HUC via a hydrologic connection. We highlighted the HUC with the pre-flood taxon occurrence and the HUC with the potentially flood-dispersed taxon occurrence to visualize the risk of species spread by flood waters on FaST maps. We removed any taxa from the maps that had native ranges in HUCs with inter-basin flood connections or taxa that did not have any potential flood-dispersed introductions. We recorded the relationship between species, HUCs, and introduction status in a matrix for all FaST maps.

Fig. 2
figure 2

Example FaST map for Hurricane Ida in the northern extent over the Mid-Atlantic U.S. showing occurrences and potential spread due to flood dispersal of New Zealand mudsnail, Potamopyrgus antipodarum

Full size image

Results

We developed 16 FaST maps documenting major storm and flood events (15 between 2017 and 2022 and one hindcast in 2012) which were published on the NAS database as of 30 July 2023 (Table 1; see Supplemental Information for hyperlinks to individual FaST maps). Flooding was mapped along the Atlantic coast in five maps, along the Gulf coast in five maps, both Atlantic and Gulf coasts in three maps, solely inland and non-hurricane-related in one map, and in Pacific (Hawai’i) and Caribbean (Puerto Rico) islands in one map each (Fig. 3). All maps were made in response to a hurricane except the one for Midwest Spring Flood in 2019. Two maps were developed for Hurricane Ida for the convenience of displaying the spatial extent and regional grouping of flooding into southern and northern portions. The Gulf and South Atlantic coasts had two to four separate events where flood-impacted areas overlapped during 2017–2022, while the North Atlantic coast had two to three overlapping events from Hurricanes Sandy in 2012, Isaias in 2020, and Ida in 2021 (Figs. 2, 3). Hurricanes Sally and Delta in 2020 were the only mapped events that shared flood-impacted areas in the same year.

Table 1 Flood events by location, start and end dates, area of potential impact (km2), and number of non-native aquatic taxa and U.S.-prohibited aquatic taxa in each impacted area
Full size table
Fig. 3
figure 3

Flood events from 2012 and 2017–2022 (see Table 1 for respective dates and location descriptions). Storm path and change in strength is indicated by lines and arrows (NOAA National Hurricane Center). Flood impact area is labeled by storm or flood event. Overlapping flood impact areas are shown as a different color for Hurricanes Sally and Delta in 2020 (panel g)

Full size image

Most lateral connections were found along ocean coastlines and near river confluences. The number of lateral connections between HUCs ranged from four (Hurricane Sally in the North Atlantic coast) to 83 (Hurricane Dorian in the South Atlantic coast) with a median of 25 lateral connections and an interquartile range of 13 and 46 lateral connections (Table 2). Inter-basin connections between different 4-,6-, and 8-digit HUCs were highest due to Hurricanes Harvey and Irma, and lowest for island storms (Maria and Lane) and for Hurricanes Sally and Delta, which had the lowest number of lateral connections; few novel connections were between HUCs that did not already have a pre-flood hydrologic connection (Table 2). Hurricane Harvey caused nine pairs of HUCs to connect in southeast Texas, and Hurricane Ida (North) caused five pairs of HUCs to connect across New York City and Long Island. The major novel lateral connections were between the Lower Colorado River and the Navidad River, the San Bernard River and the Lower Brazos River, and the Bronx-East Rivers and the Rockaway Inlet-Jamaica Bay (Table 3). These lateral connections were novel due to not sharing a coastal confluence or downstream connection. Some lateral connections were near anthropogenic culverts like canals where water flow could be bidirectional, such as along the Louisiana Gulf coast. Longitudinal (downstream) connections between HUCs ranged from zero (Hurricanes Maria in Puerto Rico and Lane in Hawai’i) to 363 (Hurricane Sally in the Gulf and South Atlantic) with a median of about 95 downstream connections and an interquartile range of 31 and 267 downstream connections. Most longitudinal connections were located along main stem (> 5th order) rivers and at the edge of large waterbodies such as Lake Pontchartrain. Locations of the longitudinal and lateral connections for all storms can be found in Pfingsten et al. (2023) and on the USGS NAS database (https://nas.er.usgs.gov/viewer/Flooding/).

Table 2 The number of lateral and longitudinal connections between neighboring HUCs Hydrologic Unit Codes; USGS, 2019b) by flood events from 2012 and 2017–2022
Full size table
Table 3 Names of river basins by flood event that had potential novel lateral connections with an adjacent river basin
Full size table

Areas of flood impact ranged from 8680 km2 (Hurricane Maria in Puerto Rico) to 161,829 km2 (Hurricane Irma in Florida and coastal Georgia and South Carolina) with a median area of 82,189 km2 and interquartile range of 35,601 and 105,383 km2 (Table 1). When we combined individual maps (Fig. 3), we were able to identify overlap in flood impacted areas across the study region (Fig. 4). Drainages of the South Atlantic and Gulf coasts had the greatest overlap in flood impacted areas across four flood events. In the South Atlantic, this included the Santee and Lower Pee Dee River drainages. In the Gulf coast, this included the Lower Mississippi, Pearl, Pascagoula, Choctawhatchee, and Apalachicola River drainages (Fig. 4). Other areas such as the Lower Calcasieu, Pensacola Bay, Lower St. Johns, Lower Pamlico, Lower Roanoke, Delaware, and the Lower Hudson River drainages had overlapping impacted areas across three flood events, while most coasts along the Gulf of Mexico and Atlantic Ocean had overlap in up to two flood events.

Fig. 4
figure 4

Overlap of 16 flood events from 2012 and 2017–2022. Flood impact area is derived from high flow, high water, and coastal storm surge at 8-, 10-, or 12-digit U.S. Geological Survey Hydrologic Unit Codes (USGS, 2019b) depending on the event (see Table 1)

Full size image

The richness of non-native aquatic taxa in flood-impacted areas ranged from 27 taxa (Hurricane Michael, Florida Panhandle) to 162 taxa (Hurricane Ida, North Atlantic Coast) with a median of 77 taxa and an interquartile range of 68 and 113 taxa (Table 1). The number of non-native aquatic taxa that are prohibited to be sold, owned, or transported in the U.S. under the U.S. Code of Federal Regulations (CFR; 7 CFR 360.200; 50 CFR 16.11–15) and were also present in any of the flood-impacted areas ranged from zero taxa (Hurricane Michael, Florida Panhandle) to six taxa (Hurricane Sally, Gulf and South Atlantic coasts; Hurricane Irma, Peninsular Florida and coastal Georgia and South Carolina) with a median of three taxa and interquartile range of two and four taxa (Table 1). The matrices of species introduction status by HUC for these maps were documented in Pfingsten et al. (2023).

Discussion

Our FaST maps are the first efforts to display the potential inter-basin dispersal of multiple aquatic species from flood events at large spatial scales using water heights and flood-impacted areas to make these predictions. Many aquatic species are prone to disperse through flood waters, and these FaST maps and their framework highlight the areas where non-native aquatic species are present and the likelihood for floods is high. These maps provide natural resource managers with information on which species may have invaded and where to begin surveillance efforts.

Frequently flooded areas

Most FaST maps were based on hurricane-induced flooding. This bias towards one flooding mechanism (i.e., storm surge) was the reason most of the mapped flood events were on coastal zones. The exceptions were due to heavy rainfall events as far inland as the Arkansas and Missouri River Basins including rainfall during hurricanes. Many of those precipitation events were too narrow in spatial scale to justify creating a FaST map due to little if any evidence of flooding that crossed drainage basins. Storm surge events are monitored by multiple agencies that provide public data on water heights at coasts. Thus, many of the repeated flood events along the Gulf and South Atlantic coasts can be explained by the prevalence of storms that make landfall in the U.S. Most hurricanes that made direct hits on the mainland of the U.S. from 1851–2004 were off the coasts of Florida, Texas, Louisiana, and North and South Carolina (see Table 10 in Blake et al. 2005).

Given the topography along drainage divides, there is a relatively low chance of lateral, or temporary, hydrologic connections compared to longitudinal, or less ephemeral, downstream hydrologic connections during localized flood events. This is due to the higher elevation separating neighboring watersheds relative to the elevation at confluences where hydrologic units typically border at stream connections from higher to lower order streams (Jones et al. 2022). During the flood event triggered by Hurricane Harvey in southeast Texas in 2017, the main stems of the Lower Colorado and San Bernard River watersheds were ephemerally connected at upstream, non-coastal portions that under normal stream flow conditions would not connect (Chen et al. 2020). We documented this same novel lateral connection but from high-water marks near the Colorado River through the East Matagorda River basin to the San Bernard River. While this is a rare event compared to most coastal and downstream flooding, we did find 15 potential lateral connections that connected adjacent, hydrologically independent river basins.

High elevation gradients have a decreased risk of flooding and lateral species movement compared to low elevation areas near coasts due to topographic barriers like valleys and precipitation runoff downstream (e.g., Hurricanes Maria in Puerto Rico, Lane in Hawai’i, and Florence in the Carolinas). These areas with high variation in topography may experience high stream flow during heavy precipitation, leading to longitudinal connections at downstream HUCs, while many of the low-lying coastal plains in flood-impacted areas experienced accumulated rainfall from upstream in the watersheds combined with storm surge at mouths of main stem rivers (Wahl et al. 2015). This combination led to increased lateral hydrologic connections across coastal watersheds during hurricanes in the Gulf and Atlantic shores.

Dispersal risk assessment

Rarely are species introduced outside of their native range via natural dispersal mechanisms (Hulme 2009). Initial introductions to novel environments typically are via human-induced vectors, such as pet releases, intentional plantings, or unintentionally transported through recreation (e.g., boating, fishing), commerce, and transit (Hulme 2009). Once introduced, some species are adapted to utilize natural biotic and abiotic pathways for further dispersal (Green and Figuerola 2005; VonBank et al. 2018; Pieper et al. 2022). Despite the introduction pathway a species uses, they are just as likely to cause impacts to ecosystems, economies, and human health once they establish a reproducing population.

Our approach to assess dispersal risk was to provide a public list of taxa with aquatic habits found in flood-prone areas and locations of their potential dispersal to neighboring watersheds after flood events. The risk of a species’ introduction, establishment, and ecological and economic impact (i.e., invasiveness) is relative to the values placed on the area and held by the people who inhabit or use that area, while also dependent on the pool of scientific knowledge for that species. For example, the process of assigning risk to a species introduction might take into consideration evidence from multiple studies on a species’ dispersal, survival, fecundity, and effects on trophic levels, nutrient cycling, infrastructure damage, and management costs in introduced ranges (e.g., Ecological Risk Screen Summary; USFWS 2020). The FaST maps could assist in the prioritization of AIS management through the development of watchlists of potentially dispersed species and to target control efforts at flood corridors. Not all species on the maps have equal invasiveness, but 12 were considered noxious or injurious at a national level. Also, some species, such as the Tire Track Eel, Mastacembelus favus Hora 1924, found in Cape Coral, Florida in 2022, are known to take advantage of flood pulses (Siriwattanarat et al. 2022), but supporting evidence of their invasiveness is lacking (USGS 2023c).

Management implications

It is critical for management agencies and their partners to have the ability to detect and respond rapidly to new invasive species to limit their spread and reduce impacts (NISC 2022). Once a species is introduced, the best chance of eradication or containment is as an emerging population (Leung et al. 2002). Early Detection and Rapid Response (EDRR) implementation involves significant coordination and collaboration across state, regional, and national agencies (Reaser et al. 2020) and attention to these systems has come to the forefront in recent years (NISC 2022). FaST maps are an EDRR tool for managers to locate potential flood-prone areas with likely hydrologic connections that could be introduction sites for priority aquatic species. The maps are informative, accessible on the USGS NAS website (https://nas.er.usgs.gov), quickly provide the most up-to-date information after a flood event to resource managers about potential new invasions, and act as an additional tool for EDRR systems. The FaST tool helps meet the need of resource managers to forecast potential AIS movement after significant flooding and storm events by identifying flood corridors that could aid dispersal.

Future considerations

We made two assumptions implicitly estimating non-zero, uniform probabilities in the current model to make FaST maps: (1) water levels at height and proximity thresholds relative to HUC borders are sufficient for HUC-to-HUC water transfer, and (2) non-native aquatic taxa with established populations can disperse along water corridors if provided sufficient water height above ground level. We did not, however, explicitly measure or assign these probabilities, nor did we test these assumptions in this study. Instead, we generalized an unknown, uniform (i.e., “potential”) dispersal risk for all hydrologic connections and inter-basin transfer of non-native aquatic taxa across a flood-impacted area. Future studies could explicitly test these assumptions of dispersal risk to better assign probabilities and uncertainties in their estimations. For example, verification of FaST-identified lateral connection points during and after floods would provide binomial probabilities to likelihood of inter-basin connections, while measuring duration of inter-basin flood connections and the terrain slopes at connection points would give us a mechanistic attribute to our model. For testing our dispersal assumption, experiments of varying stream flows and levels on aquatic plant and animal dispersal distance and direction would yield estimates of spread during flood and storm events per taxonomic unit. Even if all these uncertainties were known, we cannot fully infer the establishment or impact of aquatic taxa in our model, which would give us a more thorough estimate of invasion risk potential and prioritization. Applications like habitat suitability models (Elith et al. 2008) would test an establishment assumption, and contextualized impacts on ecologies and economies would help prioritize importance of taxa in a region.

The FaST maps in this study were created using different HUC levels across maps (e.g., 8-digit HUCs in three maps in 2017, 12-digit HUCs in one map in 2018, and 10-digit HUCs in the rest). We assumed that, for the same storm event and impacted area, HUC levels would be highly correlated with connectivity between HUCs, where 8-digit HUCs would have less connectivity than 10-digit HUCs, which in turn would have less connectivity than 12-digit HUCs. This assumption was based on the HUC hierarchy where the density of HUCs increases as their individual scale or area decreases. While the decision of which HUC level to use in each FaST map was mostly opportunistic, we occasionally looked at different HUC levels during map development to assess if most connections would be longitudinal (downstream) and thus would increase production time with little or no change in lateral or novel connections. Still, FaST maps could benefit from a more thorough analysis of this assumption by comparing results from maps created at differing HUC levels for the same storm event and impacted area.

The ability of the FaST maps to assess flood dispersal risk is dependent on background or pre-storm surveys of non-native aquatic species prior to flooding. Many state agencies actively and routinely survey public waters for non-native and invasive species, although data gaps persist due to funding and staffing limitations, sampling bias towards priority species and recreational waters, and fieldwork delays from weather events. A systematic sampling of non-native aquatic species in flood prone areas pre- and post-flood would provide verification of our dispersal risk assessment. The reciprocal situation likewise occurs where background species surveys are conducted annually but evidence of flood-impacted areas are not recorded due to absent or inactive streamgages, storm tide sensors, and high-water marks. Instead, flood risk could be estimated using the topography near streams and flood-prone areas when real-time measurements of flood conditions are lacking. Low elevation coupled with proximity to hydrologic features in the landscape such as wetlands could be associated with high flood risk. Given the uncertainty in elevation data derived from lidar, especially around vegetated areas, we may need to research other methods beyond using water height to predict the extent of flood inundation such as stream discharge, remote sensing, and atmospheric model data (Smith 1997; Brakenridge and Anderson 2006; Teng et al. 2017).

The dispersal risk of aquatic invasive species is not undervalued by invasion ecologists and resource managers based on the amount of risk assessments identifying human-mediated pathways for species spread (Mazzotti and Briggs-Gonzalez 2015; Dean et al. 2024b), but the evidence for natural dispersion risk is relatively limited compared to human-mediated spread. We lack systematic studies of the successful establishment and spread rate of these flood-adapted species after dispersal events. The potential for species to spread via flood waters may be indicative of their behaviors and life history. Future research in developing FaST maps will prioritize species prone to take advantage of high flow conditions, such as Amazonian fishes (Junk et al. 1989), or adapted to utilize buoyancy and fragmentation, such as floating plants. We should also consider the impacts from human-induced climate change on flood frequency and if aquatic invasive species are dispersing during floods at higher rates due to increased opportunities and corridors (Lopez et al. 2022). Interdisciplinary collaborations and advances in automated water level measurements may give us the early detection capabilities needed to halt the spread of impactful species after major floods at a national level.

Conclusions

We have provided a simple model for testing assumptions of aquatic plant and animal migration by flood waters, but we offer a unique perspective that differs from most local scale, single system models by focusing our model at a landscape scale across multiple waterbody types and geographic terrains. The FaST maps provided by the USGS offer a comprehensible display of the flood dispersal model to the public, while the resources for development are derived from public repositories that are only improving (e.g., USGS 3DHP; https://www.usgs.gov/3d-hydrography-program). Beyond the implications for understanding flood dispersal dynamics of aquatic species, the hope with these maps is to provide AIS-concerned parties with a tool for locating potential novel introductions and to build upon an increasing toolset for early detection and rapid response efforts. We aim to continue improving FaST maps with the considerations laid out above while maintaining open access to the public.