This dataset describes the potential responses of native freshwater-dependent and terrestrial vertebrates to threats in the North Australian Tropical Savannas. We used Delphi expert elicitation to assess the probability of persistence of functional groups of species under different threat levels, including confidence levels. The probability of persistence is the likelihood that a species within a functional group would persist under a specified threat level (low, medium, high). The elicitation included 36 experts (25 terrestrial, ten aquatic, one both) across seven faunal groups and 21 threats. The data collection comprises 70,947 records of individual expert assessments (66,033 terrestrial and 4,914 aquatic), across 64 functional groups (46 terrestrial, 18 aquatic) and 21 threats (14 terrestrial, three aquatic, four both). Threats include introduced invasive species, livestock grazing, altered fire regimes, altered water flow regimes, and rivers and streams' longitudinal barriers. The dataset can support threat assessments or conservation planning initiatives but is not a substitute for data derived from experimentation, monitoring, and modelling empirical data. We followed a structured modified Delphi elicitation process to estimate the potential responses of species to selected threats. The process involved email-based assessments and discussions. The authors of this study include scientists with experience in conservation planning and elicitation processes and experts with knowledge about the ecology of diverse taxa and threats in northern Australia. The elicitation process was undertaken independently for aquatic and terrestrial species but followed the same steps and procedures to elicit, analyse, and summarise expert knowledge. The assessment comprised three stages (pre-elicitation, elicitation, and post-elicitation) based on a 4-point elicitation procedure. During the elicitation stage, experts were asked to provide values for each functional group-threat combination, including: (1) probability of persistence (0 to 1) under three levels of each of the threats (best guess); (2) lower and (3) upper bounds, representing the lowest and the highest plausible values of probability of persistence; and (4) level of confidence (between 50% and 100%) that the true value fell within the bounds. The probability of persistence ranged between 0 and 1 and was defined as the likelihood that a species within a functional group would persist under a specified threat level (low, medium, high). We defined persistence as the maintenance of populations of species at levels sufficient to perform their ecological function over 20 years, which we considered a reasonable timeframe for reviewing the effectiveness of conservation actions. The probability of persistence was estimated independently for each threat. We considered the possibility of assessing interactions but decided against it due to knowledge gaps and many possible and complex combinations, especially for interactions involving more than two factors. We assessed the potential responses of species to 21 threats for which data on spatial distribution are readily available across northern Australia, including introduced invasive species, livestock grazing, altered fire regimes, altered water flow regime, and longitudinal barriers to rivers and streams. We assessed 18 specific threats for terrestrial species: introduced invasive animals (8 species) and plants (8 types), livestock grazing, and fire. For aquatic species, we focused on eight threats: aquatic invasive plants (2 species), invasive animals (3 species, also assessed for terrestrial species), livestock grazing, altered flow regime (e.g., water extraction, regulation for agriculture or hydropower), and longitudinal barriers (e.g., dams, weirs, culverts). These threats were selected for their relevance and range of negative impacts on aquatic and terrestrial ecosystems across northern Australia. For the assessment of terrestrial species, we identified 46 functional groups in four taxonomic classes: four amphibian groups from 60 species; 19 reptile groups from 210 species; 12 bird groups from 286 species; and 11 mammal groups from 102 species. For aquatic species, we categorised 138 species from three different aquatic taxa (44 fishes, eight turtles, 86 waterbirds) into 18 functional groups (6 groups for each taxon). We worked with species experts to classify species into functional groups that would likely have similar responses to the threats we examined. Functional groups were based on traits that experts considered relevant in terms of how different taxonomic groups (e.g., amphibians, birds, fish, mammals) could respond to threats, including size, trophic guild, activity (e.g., diurnal, nocturnal), and habitat use (e.g., aquatic, arboreal, fossorial, saxicoline, terrestrial). The main component of this data collection is comprised of 70,947 records of individual expert assessments (66,033 terrestrial and 4,914 aquatic), across 64 functional groups (46 terrestrial, 18 aquatic) and 21 threats (14 terrestrial, three aquatic, four both); records are in two Excel spreadsheets, one for terrestrial species (Data Table 1) and one for aquatic species (Data Table 2). Each record includes the estimated probability of persistence (0 to 1) of a given functional group under a specified threat level (low, medium, high) and the confidence associated with that value (51 to 99%), as well as the standardised persistence (Persistence fit) at 80% confidence level (Confidence req.), based on linear extrapolation. Each full record also includes the following fields: anonymised expert (Expert ID), taxonomic category (e.g., birds, mammals), functional group (e.g., B01 to B12 for birds), threat (e.g., fire, grazing); threat level (1, 2, 3), and type of estimate (upper, best, lower). The second component of the data collection contains summary statistics (Data Table 3) derived from the data records, which includes 25 individual Excel worksheets, one for each threat; it comprises 954 records (828 terrestrial, 126 aquatic), each corresponding to a functional group-threat combination. Summary tables for threats that were assessed for both terrestrial and aquatic species are included as separate worksheets. Each record in the tables consists of the following data: taxonomic group, functional group, sample size (N), and summary statistics for estimated persistence under each threat level (mean, min, first quartile, median, third quartile, and maximum); these are coded by combining threat level and statistic (e.g., L1_AVG = mean persistence for low threat level). Software/equipment used to create/collect the data: All data was collated, prepared and analysed using MS Excel software. Software/equipment used to manipulate/analyse the data: All descriptive and statistical analyses were conducted in R (R Development Core Team, version 4.0.2).