New Species Discoveries Refine Global Biodiversity Patterns

ABSTRACT

Global biodiversity patterns underpin ecological theory and conservation planning, yet the extent to which these inferences are shaped by the ongoing accumulation of taxonomic knowledge remains poorly understood. Here we quantified how terrestrial vertebrate discoveries from 1920 to 2020 reshaped estimated global diversity patterns and their inferred environmental drivers. Estimated diversity patterns remained largely stable for birds, whereas reptiles and amphibians showed notable reconfiguration, with 20.3% and 31.7% of their diversity centers (top 5% of total range‐weighted rarity) shifting to new regions, including Australia and Southeast Asia. These shifts were accompanied by changes in inferred environmental drivers, with the apparent influence of temperature declining and precipitation gaining importance for ectothermic groups. Our findings suggest that estimated global diversity patterns and driver relationships for less well‐known taxa may be less reliable than often assumed. Continued field exploration and taxonomic work are therefore essential to close critical biodiversity knowledge gaps.

Analysing 34,036 terrestrial vertebrates (39.6% formally described since 1920), we found that environmental drivers of species richness have remained stable for birds and mammals. For reptiles and amphibians, however, newly discovered species have substantially shifted these drivers, in particular mean annual temperature. This suggests that global biodiversity estimates for poorly studied taxa may be less reliable than previously assumed.

1. Introduction

Understanding global biodiversity patterns and their environmental drivers is crucial in quantifying ecological and evolutionary processes, identifying conservation priorities and informing strategies to mitigate biodiversity losses (Gaston 2000; Myers et al. 2000). Global biodiversity has been extensively mapped for well‐studied groups such as vertebrates (Ceballos and Ehrlich 2006; Jetz et al. 2012; Roll et al. 2017) and plants (Cai et al. 2023; Kreft and Jetz 2007), and increasingly for underrepresented taxa like invertebrates (Orr et al. 2021; Phillips et al. 2019; Pinkert et al. 2025) and microorganisms (Tedersoo et al. 2014). However, these maps often suffer from inherent biases, stemming from uneven geographic coverage and incomplete taxonomic knowledge (Feeley et al. 2017; Meyer et al. 2016). These limitations can bias biodiversity assessment, distort ecological and evolutionary inferences, and lead to suboptimal conservation decisions (Abeli et al. 2022; James et al. 2021). While geographic biases are increasingly recognized and addressed (Menegotto and Rangel 2018; Yang et al. 2013), the quantitative influences of incomplete taxonomic knowledge, which is a fundamental limitation known as the Linnean shortfall (Hortal et al. 2015), on our understanding of global biodiversity remain largely unknown.

All taxonomic groups contain a substantial proportion of undiscovered species, with significant additions of new species in recent decades (Mora et al. 2011; Stork 2018). For example, while over 2000 new vascular plant species are discovered annually (Cheek et al. 2020), it is estimated that as many as 100,000 species remain unknown (Corlett 2016). This trend is also observed in well‐studied taxonomic groups (Liu et al. 2022). Between 1993 and 2008, 408 new mammal species were described, accounting for about 10% of previously known diversity (Burgin et al. 2018; Ceballos and Ehrlich 2009). Moreover, undiscovered and newly discovered species are disproportionately concentrated in biodiversity hotspots and remote areas (Joppa et al. 2011; Moura and Jetz 2021). Incomplete sampling in these often‐inaccessible areas introduces substantial uncertainty into assessments of global biodiversity patterns, particularly for understudied taxonomic groups (Mora et al. 2008). For example, the observed mid‐latitudinal peak in global earthworm diversity likely reflects under‐sampling in the tropics relative to well‐sampled temperate regions (Phillips et al. 2019), a discrepancy increasingly revealed by molecular studies uncovering the hidden earthworm diversity in the tropics (Maggia et al. 2021). Similarly, for ants, centers of richness and rarity are expected to shift significantly with greater sampling efforts (Kass et al. 2022). Given that approximately 80% of Earth's species are yet to be formally described (Mora et al. 2011), it is crucial to understand the distribution of species discoveries and their influences on estimated diversity patterns.

New species can also reshape ecological inferences regarding the relationships between diversity and environmental factors, challenging predictions about how biodiversity responds to global change. For example, the prevalence of newly described species in tropical regions may strengthen species‐energy relationships by highlighting how more individuals of newly discovered species coexist (Gaston 2000). Specifically, a hump‐shaped relationship between diversity and mean annual temperature may be further evidenced due to the disproportionately high diversity in regions with high elevation and relatively cool temperature (Moura and Jetz 2021). Conversely, the environmental heterogeneity hypothesis (Stein et al. 2014) may also gain further importance in explaining global biodiversity patterns as species discoveries continue to be made across tropical montane regions (Ceballos and Ehrlich 2009; Costello et al. 2015). While recent studies have suggested climate as the dominant factor in explaining species richness patterns for different groups (Cai et al. 2023; Mi et al. 2024), the impact of new species on the relative role of different aspects of climate remains poorly understood. Addressing this major knowledge gap is critical for refining extinction risk assessment and informing conservation strategies, especially for data‐deficient taxa (Goodsell et al. 2025).

Here, we used the IUCN Red List of Threatened Species (IUCN 2024), which includes up‐to‐date taxonomic and geographic information covering a broad range of taxa, to quantify how newly discovered terrestrial vertebrates alter spatial patterns of species richness and rarity, and modify the inferred environmental drivers that shape such patterns. We focused on terrestrial vertebrates because they are the primary targets of conservation efforts and represent a gradient of discovery completeness (Burgin et al. 2018; Cox et al. 2022; Roll et al. 2017). While we recognize that new species can arise from a combination of novel discoveries and taxonomic changes such as splitting and lumping (Isaac 2004; Lessa et al. 2024), our study focused on the species that are new to the scientific record. To prioritize the new species over the reclassification of previously known taxa, we identified the discovery date as the year of first formal description, using the authority year associated with each species' scientific name. Moreover, we used the IUCN Red List because it provides a relatively conservative species list that prioritizes widespread taxonomic consensus when recognizing new species (IUCN 2024). We addressed three key questions: (1) Where are newly discovered species concentrated? (2) How do these new species refine our understanding of global diversity patterns? Specifically, do they shift latitudinal diversity gradients and the distribution of diversity centers? (3) How do new species modify the inferred relationships between species diversity and environmental drivers, with implications for predicting the consequences of global change?

2. Materials and Methods

2.1. Data Compilation

We compiled species range maps from multiple sources for terrestrial vertebrates. For birds, we used data from the Handbook of the Birds of the World and BirdLife International version 7 (BirdLife International and Handbook of the Birds of the World 2022). For mammals and amphibians, we used the International Union for Conservation of Nature Red List of Threatened Species (IUCN 2024). We sourced reptile species range maps from Roll et al. (2017). We filtered the range polygons for all taxa to include only extant, native, or reintroduced species. Following the guideline of the IUCN mapping standards, we included species with a presence code of 1–3 (extant, probably extant, possibly extant), an origin code from 1 to 2 (native, reintroduced) and a seasonal code from 1 to 3 (resident, breeding, non‐breeding). We excluded records representing uncertain, extinct, introduced, or transient occurrences, specifically those with a presence code of 4–7 (possibly extinct, extinct, presence uncertain, expected additional range), an origin code from 3 to 6 (introduced, vagrant, origin uncertain, assisted colonization), and a seasonal code from 4 to 5 (passage, seasonal occurrence uncertain). For reptiles, we included all available range maps, as they all represent extant and native species. Our dataset comprised range maps for 10,856 bird, 5586 mammal, 9880 reptile, and 7714 amphibian species, covering 97.1% of known bird (HBW and BirdLife International 2025), 82.9% of known mammal (Burgin et al. 2018), 78.6% of known reptile (Uetz et al. 2021) and 85.0% of known amphibian species (Frost 2024).

2.2. Taxonomy

We standardized species nomenclature and taxonomic authorities for all taxa to ensure consistency. For birds, we followed the taxonomy of the Handbook of the Birds of the World and BirdLife International version 7 (BirdLife International and Handbook of the Birds of the World 2022). For mammals and amphibians, we used the IUCN Red List (IUCN 2024), supplemented with The Mammal Diversity Database (Burgin et al. 2018) and Amphibian Species of the World 6.2 (Frost 2024) where IUCN data were incomplete. For reptiles, we first referenced The Reptile Database (Uetz et al. 2021) checklist (version 2 March 2024). We then resolved any discrepancies by cross‐referencing against The Reptile Database synonym checklist (version April 2023) and synonyms listed in the IUCN Red List (IUCN 2024). We conducted additional literature searches to complete taxonomic authority information for reptile species when necessary.

2.3. Diversity Patterns

To analyse global diversity patterns, we first aggregated species range maps onto a global hexagonal grid. We used the ‘dggrid’ package (Barnes and Sahr 2017) to generate a global hexagonal grid with a cell resolution of 7774 km². This grid is finer than the 1° square grid commonly used in large spatial scale studies. The hexagonal grid can minimize geographic distortions in area and distance, making it suitable for macroecological studies (Sahr et al. 2003). We then overlaid the grid with global land cover data, excluding any cells that the remaining land area was less than half of the original cell size. Within each remaining cell, we determined the presence and number of species by intersecting with species ranges.

To explore temporal dynamics of global vertebrate diversity gradients in relation to new species, we estimated diversity patterns at 10‐year intervals between 1920 and 2020. Each interval represented the cumulative number of species that had been described by the end of the particular period. We considered each species to be new in the year of its first formal description, identified by the authority year associated with each species' scientific name. The latest description date in our dataset was 2019 for birds, 2020 for mammals and amphibians, and 2015 for reptiles. For convenience, we hereafter refer to 2020 as the upper time limit for all groups.

To account for the potential effects of biogeographic history (Rowan et al. 2020), we assigned each hexagonal grid cell to one of six zoogeographic regions: Australian, Neotropical, Nearctic, Palearctic, Ethiopian, and Oriental. This delineation followed the work of Holt et al. (2013) and Ficetola et al. (2017). In total, our final analyses included 9716 bird, 5245 mammal, 8646 reptile, and 7267 amphibian species.

2.4. Environmental Variables

To identify the environmental relationships for vertebrate diversity, we selected variables reflecting contemporary climate, environmental heterogeneity, and historical climate stability, which are widely recognized as primary predictors of global biodiversity patterns (Field et al. 2009; Fine 2015). We obtained contemporary climate data from the Climatologies at high resolution for the earth's land surface areas (CHELSA v1.2) dataset at a 30 arc‐sec resolution (Karger et al. 2017). We included mean annual temperature (bio1), annual precipitation (bio12) and mean monthly precipitation of the warmest quarter (bio18) to represent the water‐energy dynamics that constrain productivity (Hawkins et al. 2003). We further selected variables describing variation in temperature and precipitation, including annual range of temperature (bio7) and precipitation seasonality (bio15), as high climatic variation can constrain species richness by excluding narrow‐niche species that are unable to tolerate seasonal physiological stress (Hurlbert and Haskell 2003; Ochoa‐Ochoa et al. 2019). To minimize multicollinearity, we performed a Variance Inflation Factor (VIF) analysis of a generalized linear model of species richness. We excluded variables exhibiting high collinearity (VIF > 5: bio4, bio6 and bio17) from further analyses. Environmental heterogeneity was represented by elevation range, calculated as the difference between maximum and minimum altitude from the Global multi‐resolution terrain elevation data 2010 (Danielson and Gesch 2011). This metric serves as a proxy for niche availability and the presence of climate refugia (Stein et al. 2014). As historical climate stability can leave a signature on contemporary diversity by affecting extinction and speciation (Araújo et al. 2008; Rowan et al. 2020), we calculated temperature and precipitation anomalies as the difference between contemporary conditions and those of the Last Glacial Maximum (~21,000 years ago), derived from the CHELSA‐TraCE21k (Karger et al. 2023).

2.5. Analyses

2.5.1. Estimation of Diversity Metrics

We evaluated vertebrate diversity patterns using two diversity metrics: species richness and total range‐weighted rarity (Lennon et al. 2004). We calculated species richness as the total number of species present in each hexagonal grid cell. To emphasize the conservation value of narrow‐ranging species, we also calculated total range‐weighted rarity. We derived this metric by summing the rarity value of each species within a given cell, where rarity of a species was defined as the inverse of the number of cells it occupied. This approach is frequently used to prioritize conservation efforts for species with restricted ranges (Cox et al. 2022; Kass et al. 2022; Pinkert et al. 2025). Considering the inherent L‐shaped distribution of species ranges in our dataset, where a majority of species are narrowly distributed, we adjusted the rarity calculation to prevent the disproportionate influence of highly restricted species. We therefore added a constant offset to the denominator, that is the number of occupied cells, in the rarity calculation. This constant value was set as the 75th percentile of the overall species ranges for each vertebrate group. This method can balance the influence of narrowly distributed species, avoiding the creation of too many high‐rarity ‘islands’ while still emphasizing their conservation importance.

2.5.2. Identifying Diversity Centers

To evaluate the temporal dynamics of biodiversity centers over the past century, we identified the top 5% and 10% of hexagonal grid cells for both species richness and total range‐weighted rarity. These cells, which represent the highest diversity, were considered as diversity centers. These centers were identified for time periods 1920 and 2020 in our analysis. We then tracked the spatial shifts in these centers by comparing their locations across different time periods.

2.5.3. Spatial Models

To explore the shifts in the relationships between species diversity and environmental variables over time, we performed linear regressions of species richness against environmental variables. These explanatory variables included contemporary climate (mean annual temperature, temperature range, annual precipitation, precipitation seasonality, and precipitation of the warmest quarter), environmental heterogeneity (elevation range) and past climate change (temperature and precipitation anomalies since the Last Glacial Maximum). We natural log‐transformed species richness and square root‐transformed environmental variables to reduce skewness. For mean annual temperature and precipitation anomalies since LGM, which are left‐ or right‐skewed, we first subtracted each grid cell's value from the maximum or minimum, respectively, to yield non‐negative values before applying the square‐root transformation. We then multiplied the transformed mean annual temperature by −1, so that larger values corresponded to warmer temperatures. We then standardized all variables to zero mean and unit standard deviation for comparison across models.

Given the detection of significant spatial autocorrelation in our initial linear models, we used simultaneous autoregressive regressions (SAR) with the error type, implemented with the ‘spatialreg’ R package (Bivand and Wong 2018). To define the spatial neighbour structure for the SAR models, we used a distance band approach. We determined the optimal distance by iteratively testing a distance from 100 to 3000 km in an increment of 200 km. We selected the best distance that yielded the lowest Akaike information criterion (AIC), highest Nagelkerke's pseudo R ², and an insignificant Moran's I test for residual spatial autocorrelation. With the appropriate distance, spatial autocorrelation was significantly accounted for. As biogeographic history might vary among regions (Rowan et al. 2020), we repeated the spatial model analysis independently for each biogeographic realm.

To test for the difference in the effect of environmental factors on species richness between 1920 and 2020, we utilized a $z$‐test. As species richness of these two time periods are from the same locations and violate the assumption of independence, we incorporated the covariance between the coefficient estimates in our $z$‐test. To correctly estimate the coefficient covariance, we first extracted the residuals from each spatial model. These residuals were then transformed using the spatial error parameter and a weight matrix. This allowed us to compute the coefficient covariance. The $z$‐test was formulated as follows:

$$z=\frac{{\beta }_1-{\beta }_2}{\sqrt{\mathrm{SE}{\left({\beta }_1\right)}^2+\mathrm{SE}{\left({\beta }_2\right)}^2-2\times \mathrm{Cov}\left({\beta }_1,{\beta }_2\right)}}$$

Here, ${\beta }_1$ and ${\beta }_2$ are the coefficients of the environmental factor from two spatial models, $\mathrm{SE}\left(\beta \right)$ represent the standard error for each coefficient, $\mathit{Cov}\left({\beta }_1,{\beta }_2\right)$ represent the estimated covariance between the two coefficients. A p value < 0.05 from the $z$‐test indicates a significant difference between the two coefficients across the two time periods.

To further understand how new species may influence inferences about species richness‐environment relationships, we conducted two additional analyses. Because the relationships between species richness and temperature or precipitation are often hump‐shaped rather than monotonic (Gaston 2000; Kreft and Jetz 2007), we re‐fitted SAR error models with additional quadratic terms for mean annual temperature and annual precipitation. We further estimated the temperature or precipitation values associated with peak species richness using the quadratic models as: $-{\beta }_1/\left(2{\beta }_2\right)$, where ${\beta }_1$ and ${\beta }_2$ are the estimated linear and quadratic coefficients of temperature or precipitation, respectively.

3. Results

3.1. Geographic Shifts in Diversity Patterns

We found that species described between 1920 and 2020 cumulatively constituted 39.6% (13,491 of 34,036 species) of total terrestrial vertebrate species, representing 7.7% (832 of 10,856) of birds, 31.7% (1774 of 5586) of mammals, 50.7% (5,012 of 9,880) of reptiles, and 76.1% (5,873 of 7,714) of amphibians (Table 1). Notably, 17.3% of reptiles and 33.0% of amphibians have been discovered since 2000 (Table 1). New species were strongly concentrated in the tropics, decreasing sharply towards the poles (Figure 1). Therefore, species discoveries of vertebrates, especially for reptiles and amphibians, accentuated the unimodality of latitudinal diversity gradients (Figure 1).

TABLE 1.

Number and proportion of discovered terrestrial vertebrate species between 1920 and 2020. Proportion of discovered species within each time period is shown in parentheses.

Time period	Birds	Mammals	Reptiles	Amphibians
1920–1940	408 (3.8%)	466 (8.3%)	755 (7.6%)	629 (8.2%)
1940–1960	134 (1.2%)	204 (3.7%)	525 (5.3%)	500 (6.5%)
1960–1980	87 (0.8%)	241 (4.3%)	845 (8.6%)	941 (12.2%)
1980–2000	111 (1%)	357 (6.4%)	1179 (11.9%)	1256 (16.3%)
2000–2020	92 (0.8%)	506 (9.1%)	1708 (17.3%)	2,547 (33%)
1920–2020	832 (7.7%)	1774 (31.7%)	5012 (50.7%)	5,873 (76.1%)

FIGURE 1.

New species discoveries in the tropics steepen the latitudinal diversity gradient. Maps show the diversity of new species discovered between 1920 and 2020 for (a) birds, (b) mammals, (c) reptiles and (d) amphibians. The plots illustrate the corresponding shift in the latitudinal diversity gradient (LDG) for total species number in 1920 (red) and 2020 (blue). All maps use an identical colour scale. In the plots, solid circles represent mean species number per 5‐degree latitudinal band, and the shaded area indicates the standard deviation (SD) around the mean. Silhouettes are from PhyloPic (credits: a, Andy Wilson; b, Leonardo Ancillotto; c, Kailah Thorn; d, Vijay Karthick).

Across our 95‐km‐resolution grid, the number of species discoveries showed a significant relationship with species number in 1920 across all vertebrate groups (p < 0.001 for all four groups based on our generalized additive models; Figure 2). Birds and amphibians showed a monotonically increasing number of new species, whereas discoveries of new mammal species have declined in the most species‐rich regions since 1920. Discoveries of reptiles peaked at a moderate level of historical richness and then plateaued. In particular, new mammal species were exceptionally low in the forests and savannas of Lake Victoria Basin in East Africa (a mammal biodiversity center), while we found a disproportionately high number of discoveries in Papuan montane forests (Figure 2c), implying shifting centers of new mammalian species over the past century. Australian savannas and shrublands exhibited an unexpectedly high number of new reptile species (Figure 2e), increasing Australia's known diversity by 2.5‐fold (from 376 species in 1920 to 933 species in 2020; Figure S1).

FIGURE 2.

Shifts in estimated diversity patterns following new species discoveries. (a, c, e, g), The relationship between species richness in 1920 and subsequent discoveries (1920–2020) for birds, mammals, reptiles, and amphibians. Each point represents the species richness of a grid cell, with point colour indicating the level of total range‐weighted rarity, consistent with the right panels. Solid line represents a generalized additive model spline. Annotated labels highlight regions: Papua New Guinea (PG); Victorica Basin (VB); Australia (AU). (b, d, f, h), Maps show the shift in diversity centers of total range‐weighted rarity between 1920 and 2020. Dark shades indicate regions in the top 5% of rarity, medium shades the top 5%–10%, and light shades and grey indicate regions outside the top 10%. Shifted diversity centers are highlighted in colour: Emergent centers (regions that emerged as diversity centers after new species discoveries, blue), historical centers (regions initially identified as diversity centers but have since diminished, brown) and stable centers (regions that remained stable across discoveries, green). Inset bar plots show the proportion of shifted diversity centers for each group. Silhouettes are from PhyloPic (credits: a, b, Andy Wilson; c, d, Leonardo Ancillotto; e, f, Kailah Thorn; g, h, Vijay Karthick).

To assess how diversity centers with high biogeographic uniqueness shift with increasing taxonomic knowledge, we identified regions with the top 5% total range‐weighted rarity (emphasizing range‐restricted species) and classified them into three categories: emergent diversity centers (newly identified after increasing taxonomic knowledge), historical diversity centers (initially identified but no longer significant following species discoveries), and stable diversity centers (persistent across all discovery scenarios). Newly described bird species had negligible effects on overall diversity patterns (Figure 2b). However, we observed a significant shift in diversity centers for mammals, with 12.1% of their total area replaced by emergent centers, a result of shrinkage in the Equatorial Afrotropics and expansion in the Central Andes, Madagascar, and Indonesian islands (Figure 2d; Table S1). The spatial change in diversity centers was even more pronounced for reptiles and amphibians, with 20.3% and 31.7% of their respective center areas being replaced (Table S1). For reptiles, emergent centers appeared predominantly in the northern and western Australian savannas and shrublands (Figure 2f). For amphibians, key emergent centers included the Andean montane forests and Southeast Asian tropical forests (Figure 2h). These trends for mammals, reptiles and amphibians remain striking across different time periods and diversity metrics, with notable shifts observed when considering only the past 20 years (4.3% for mammals, 5.0% for reptiles, 13.6% for amphibians) or species richness centers (11.2% for mammals, 17.4% for reptiles, 27.3% for amphibians; Figures S2 and S3; Table S1).

3.2. New Species Alter Inferred Drivers

Our analysis revealed contrasting temporal trends in the relationships between species richness and environment among vertebrate groups and environmental variables from 1920 to 2020 (Figure 3; Table S2). While the relationships for birds and mammals remained consistent, those for reptile and amphibian species exhibited notable temporal changes (Figure 3; Figures S4 and S5). Specifically, although mean annual temperature remained a significant predictor of richness, its effect size declined significantly between 1920 and 2020 for reptiles (z = 2.01, p = 0.04) and amphibians (z = 3.70, p < 0.001). Moreover, when considering a hump‐shaped relationship, the quadratic coefficient of temperature effects on richness became more negative, indicating a decrease in temperature with the peak of ectothermic species richness (Figures S4 and S5).

FIGURE 3.

Trends in estimated relationships between environment and vertebrate diversity at the global scale. Standardized coefficients from simultaneous autoregressive models illustrate the change in species richness relationships for birds, mammals, reptiles, and amphibians from 1920 to 2020. The shaded grey area represents the 95% confidence interval. Confidence interval not overlapping the zero dashed line indicates a significant environment‐richness relationship within a time period. Temporal trend in these relationships was tested using a modified $z$‐test. Asterisks indicate a significant difference in environmental variable coefficients between 1920 and 2020: p < 0.001 (***), and 0.01 < p < 0.05 (*). Mean.Ann.Temp., mean annual temperature; Temp.Range, annual range of temperature; Ann.Prec., annual precipitation; Prec.Seasonality, precipitation seasonality; Prec.Warmest.Quarter, mean monthly precipitation of the warmest quarter; Elev.Range, elevation range; Anom.Temp.LGM, temperature anomaly since the Last Glacial Maximum; Anom.Prec.LGM, precipitation anomaly since the Last Glacial Maximum. Silhouettes are from PhyloPic (credits: Andy Wilson, bird; Leonardo Ancillotto, mammal; Kailah Thorn, reptile; Vijay Karthick, amphibian).

Similarly, although annual precipitation maintained a significant association with reptile richness, the magnitude of this effect showed a clear decline (z = 1.58, p = 0.11). The estimated precipitation with the peak of reptile richness also declined with the increase of the quadratic coefficient (Figures S4 and S5). In contrast, the contribution of precipitation‐related factors remained significant for amphibian diversity and generally increased with new species discoveries, particularly the influence of precipitation seasonality (z = 3.45, p = 0.001). The significant association with elevation range strengthened for both reptiles (z = −3.77, p < 0.001) and amphibians (z = −7.80, p < 0.001). While the influence of temperature and precipitation anomalies since the Last Glacial Maximum remained largely consistent, precipitation anomalies showed a significant decline for reptile diversity pattern (z = −2.10, p = 0.04; Figure 3, Table S2).

As diversity‐environment relationships can vary among biogeographic realms due to their divergent histories (Rowan et al. 2020), we also analysed these relationships at the realm level. For all vertebrate groups, environment‐richness relationships remained stable across time periods within the Ethiopian, Nearctic, and Palaearctic realms (Figure 4). However, in the Australian, Neotropical, and Oriental realms, shifts in relationships between species richness and temperature and precipitation for reptiles and amphibians generally mirrored global trends (Figure 4, Table S3).

FIGURE 4.

Contrasting environmental driver estimates for vertebrate diversity across biogeographic realms. Standardized coefficients from simultaneous autoregressive models represent predictors of species richness in 1920 (circles) and 2020 (triangles). Panels show results for birds (left), mammals (middle left), reptiles (middle right) and amphibians (right). Biogeographic realm map is shown in gradient of known vertebrate diversity in 2020. Error bar indicates 95% confidence intervals. Error bar not overlapping the zero dashed line indicates a significant environment‐richness relationship within a time period. Difference in these relationships between two time periods was tested using a modified $z$‐test. Asterisks indicate a significant difference in environmental variable coefficients between 1920 and 2020: p < 0.001 (***), 0.001 < p < 0.01 (**) and 0.01 < p < 0.05 (*). Mean.Ann.Temp., mean annual temperature; Temp.Range, annual range of temperature; Ann.Prec., annual precipitation; Prec.Seasonality, precipitation seasonality; Prec.Warmest.Quarter, mean monthly precipitation of the warmest quarter; Elev.Range, elevation range; Anom.Temp.LGM, temperature anomaly since the Last Glacial Maximum; Anom.Prec.LGM, precipitation anomaly since the Last Glacial Maximum. Silhouettes are from PhyloPic (credits: Andy Wilson, bird; Leonardo Ancillotto, mammal; Kailah Thorn, reptile; Vijay Karthick, amphibian).

4. Discussion

4.1. Geographic Shifts in Diversity Patterns

The ongoing accumulation of taxonomic knowledge, especially new species, has profound implications for our understanding of global biodiversity, especially given the taxonomic and spatial biases inherent in these discoveries (Hortal et al. 2015; Meyer et al. 2016; Turton‐Hughes et al. 2024). Our findings demonstrate that this process can systematically refine estimates of global diversity patterns, especially for understudied taxa (e.g., reptiles and amphibians) and less‐explored regions such as tropical mountains. New species are strongly concentrated in the tropics, reinforcing and in some cases steepening the latitudinal diversity gradient (Figure 1). Indeed, the tropics not only host a high diversity of life (Pillay et al. 2022), but are also home to the majority of the world's new and undiscovered species (Raven et al. 2020), such as vertebrates (Ceballos and Ehrlich 2009; Moura and Jetz 2021), ants (Kass et al. 2022), and plants (Joppa et al. 2011). However, counter to the expectation that high known diversity would correlate with a high number of species discoveries (Joppa et al. 2011), our analysis revealed an opposite trend in some important areas (e.g., mammals in the Equatorial Afrotropics), resulting in a decline in their relative importance as diversity centers. Instead, we identified several regions as emergent diversity centers, characterized by a disproportionately high number of newly discovered species (Figure 2). This finding, which highlights newly identified diversity centers in the North Australian savanna and West Australian shrubland (for reptiles) and in the Amazonian and Southeast Asian forests (for amphibians), aligns with the concept of ‘hidden hotspots’ (Xu et al. 2024) (underrecognized biodiverse regions due to incomplete knowledge) and ‘shifted hotspots’ (Flanagan et al. 2024) (where the relative importance of regions as diversity hotspots changes once better biodiversity knowledge alter the global ranking) and extends their existence to a broader spatial and taxonomic scales. The identification of emergent centers not only refines global biodiversity patterns, but also highlights previously underrecognized regions that now require emergent conservation attention.

Our study identified the critical role of historical sampling effort in shaping the temporal change of diversity patterns. We found that biodiversity patterns for well‐studied groups (e.g., birds) and well‐sampled regions (e.g., Palaearctic biogeographic realm) remained consistent despite ongoing species discoveries and taxonomic changes. This observation is likely a direct consequence of high historical sampling efforts in these areas (Boakes et al. 2010; Orr et al. 2021), which have yielded a limited number of newly described species in recent years. Consequently, existing diversity patterns for these groups and regions are largely consistent (Faurby and Sandel 2025). In contrast, the temporal shift in amphibian diversity patterns in the Brazilian rainforest and Andean cloud forest is largely due to these regions having been historically under‐sampled (Guerra et al. 2020). This finding highlights the importance of continued exploration and taxonomic efforts in under‐sampled regions and taxonomic groups.

4.2. New Species Alter Inferred Drivers

New species discoveries challenge our understanding of the environmental drivers of biodiversity patterns. We found a significant decline in the importance of mean annual temperature in explaining reptile and amphibian species richness. This is likely because new discoveries are disproportionately skewed towards relatively cool habitats. For example, the Strabomantidae family, which constitutes an exceptionally high proportion of newly described amphibian species (Figures S6 and S7), is often found in cool and humid habitats, such as the historically underexplored Andean cloud and montane forests (Guerra et al. 2020). The discovery of numerous lizard species in deciduous forests also challenges the assumption that members of this group prefer hot habitats (Cox et al. 2022). The ongoing discovery of species in relatively cool habitats suggests that the monotonically positive association between temperature and richness for ectotherms may be weaker than previously estimated.

Furthermore, new species have shifted the observed temperature at which species richness peaks towards cooler conditions. Reptiles and amphibians are already identified as among the most vulnerable vertebrates to global warming (Harfoot et al. 2021; Murali et al. 2023; Pottier et al. 2025). We found that regions of high vulnerability coincide with emergent diversity centers, where newly described species are concentrated. These new species might face elevated extinction risks as they are often narrow‐ranged endemics. More critically, these underexplored regions likely harbour a substantial number of species yet to be formally described (Moura and Jetz 2021). These undiscovered species likely face threats similar to known species, potentially going extinct before being formally documented (Costello et al. 2013). Consequently, diversity‐energy relationships inferred from historically incomplete species inventories may insufficiently capture richness in these regions, suggesting that projections of future biodiversity loss could be conservative.

Precipitation‐related factors have become increasingly important in inferring amphibian diversity, as most new amphibian species occur in tropical regions with relatively high rainfall (Wu et al. 2024; Xu et al. 2024). Given increasing climate extreme events and induced dryness (Pottier et al. 2025; Wu et al. 2024), conservation planning for amphibians in these high‐rainfall regions would benefit from accounting for potentially undiscovered species. Conversely, the strength of inferred association between annual precipitation and reptile diversity has declined over time, driven mainly by discoveries of arid‐tolerant lineages (e.g., lizards) in savanna regions of Australia and the Cerrado (Cox et al. 2022; Flanagan et al. 2024). The increasing importance of environmental heterogeneity in shaping reptile and amphibian diversity further supports the need for ongoing field surveys in mountain and valley areas (Moura and Jetz 2021), particularly in the Neotropical and Oriental realms. Therefore, our results highlight the need to consider the altered contributions of environmental factors caused by species discoveries when implementing biodiversity models and projections (e.g., species distribution models).

New species may also influence our understanding of the biotic drivers of diversity patterns. For instance, biotic interactions are hypothesized to intensify towards the tropics, promoting species richness through niche partitioning and coevolutionary diversification (Fine 2015; Schemske et al. 2009). Because newly discovered species often have narrow ranges and are concentrated in tropical regions, they likely occupy specialized niches that are critical for coexistence, reinforcing the importance of these interactions. Yet new species and their interactions remain poorly documented (Hortal et al. 2015), leaving their specific contribution to the effect of species interactions on diversity patterns unclear.

4.3. Implications

Our findings have important implications for global biodiversity conservation. Specifically, we found that a high proportion of diversity centers for mammals, reptiles, and amphibians have been replaced over the past century by newly identified centers in previously overlooked regions, yet these new centers are often threatened by deforestation and associated biodiversity loss (Pillay et al. 2024). As a result, we may also be underestimating hidden biodiversity loss in these regions (Liu et al. 2022). In addition, since our study focused on vertebrates, which are relatively well‐studied, global diversity patterns and ecological inferences derived from incomplete taxonomic knowledge may be even less reliable for poorly studied taxa. Indeed, studies on global patterns of earthworms and fungi, which are less‐studied groups, have been criticized for their incomplete taxonomical surveys (James et al. 2021; Phillips et al. 2019). Therefore, it becomes more critical to incorporate unknown species into macroecological studies to effectively assess and guide conservation across different taxonomic groups (Goodsell et al. 2025), highlighting the need to improve biodiversity monitoring and close knowledge gaps, as emphasized by the Kunming‐Montreal Global Biodiversity Framework. To effectively integrate new species discoveries into conservation management, we advocate developing new methods to predict the taxonomic and spatial distributions of undiscovered species (e.g., Moura and Jetz 2021), especially in tropical and montane areas.

While mean annual temperature is a primary driver of biodiversity responses to climate change (Pottier et al. 2025), our results suggest that temperature alone is insufficient to accurately predict species losses, particularly for amphibians for which water availability acts as a critical physiological constraint. While temperature change is more consistent and predictable than precipitation (Wu et al. 2024), integrating water availability constraints with temperature is a necessary step for designing effective conservation strategies.

We are aware that the observed temporal changes in global biodiversity patterns may not exclusively reflect novel species discoveries, but are also influenced by taxonomic changes such as species lumping and splitting (Lessa et al. 2024). This is particularly relevant for well‐studied groups such as birds and mammals (Faurby and Sandel 2025). While taxonomic splitting may not always increase local species richness, it often increases total range‐weighted rarity, as newly recognized species typically have smaller ranges than the taxon from which they were split (Flanagan et al. 2024). This shift in estimated rarity patterns influences conservation prioritization and highlights the necessity of continued investment in taxonomy. While our analyses do not fully distinguish between novel discoveries and taxonomic changes, this limitation highlights that taxonomic incompleteness introduces significant and predictable uncertainty into global biodiversity patterns. Furthermore, while our study represents a substantial proportion of described terrestrial vertebrates (78%–97% across taxa), it does not capture the full extent of known biodiversity, let alone true diversity. The species absent from our dataset are disproportionately newly described and range‐restricted, suggesting that our estimated shifts in global diversity patterns may still underrepresent what a more complete dataset would reveal.

5. Conclusion

Our work provides a dynamic perspective on global biodiversity patterns, showing how the accumulation of new species notably refines estimates of macroecological patterns and relationships. While broad‐scale patterns remain relatively robust for well‐studied taxa, we find that both the geography of diversity centers and the inferred importance of key environmental variables are more sensitive to the addition of new species in less well‐known groups, especially reptiles and amphibians in tropical regions. These findings indicate that current estimates of global diversity patterns and processes are subject to systematic uncertainty due to incomplete taxonomic knowledge. Models and projections based on incomplete taxonomic knowledge may underestimate diversity and mischaracterize its environmental constraints, especially in tropical and topographically complex regions where many species remain undescribed or poorly resolved. This has important implications for biodiversity assessment and conservation planning under global change. We therefore echo Wilson's (2017) call for ‘more boots on the ground’ to accelerate taxonomic research both in the field and in natural history collections, to close critical biodiversity knowledge gaps.

Author Contributions

Jiajia Liu conceived the initial idea of this work. Yangqing Luo and Jiajia Liu designed the study. Yangqing Luo compiled and curated the data. Yangqing Luo analysed and visualized the data. Yangqing Luo wrote the manuscript with substantial contributions from David Lindenmayer, Jens‐Christian Svenning, Wen‐Yong Guo, Weiguo Du, Zhijun Ma and Jiajia Liu. All authors commented on the draft.

Funding

Jiajia Liu and Yangqing Luo were supported by the National Key Research and Development Program of China (2022YFF0802400) and the National Natural Science Foundation of China (32401318). Jens‐Christian Svenning was supported by the Center for Ecological Dynamics in a Novel Biosphere (ECONOVO), funded by the Danish National Research Foundation (grant DNRF173).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: New species discoveries across countries. Known species numbers in 1920 (blue) and species discoveries from 1920 to 2020 (red) across top 10 countries with highest species number for (a) birds, (b) mammals, (c) reptiles and (d) amphibians.

Figure S2: Shift in diversity centers of total range‐weighted rarity for vertebrates from 2000 to 2020. Maps showing the shift in diversity centers of rarity between 2000 and 2020 for (a) birds, (b) mammals, (c) reptiles and (d) amphibians. Dark shades indicate regions in the top 5% of rarity, medium shades the top 5%–10% and light shades and grey indicate regions outside the top 10%. Shifted diversity centers are highlighted in colour: emergent centers (regions that emerged as diversity centers after new species discoveries, blue), historical centers (regions initially identified as diversity centers but have since diminished, brown) and stable centers (regions that remained stable across discoveries, green). Inset bar plots show the proportion of shifted diversity centers for each group.

Figure S3: Shift in species richness centers for vertebrates from 1920 to 2020. Maps showing the shift in diversity centers of species richness between 1920 and 2020 for (a) birds, (b) mammals, (c) reptiles and (d) amphibians. Dark shades indicate regions in the top 5% of rarity, medium shades the top 5%–10%, and light shades and grey indicate regions outside the top 10%. Shifted diversity centers are highlighted in colour: emergent centers (regions that emerged as diversity centers after new species discoveries, blue), historical centers (regions initially identified as diversity centers but have since diminished, brown) and stable centers (regions that remained stable across discoveries, green). Inset bar plots show the proportion of shifted diversity centers for each group.

Figure S4: Shifts in estimated relationships between environment and vertebrate diversity after accounting for non‐linear effects of temperature and precipitation on species richness. Standardized coefficients from simultaneous autoregressive models represent predictors of species richness in 1920 (circles) and 2020 (triangles). Panels show results for birds (left), mammals (middle left), reptiles (middle right) and amphibians (right). Error bar indicates 95% confidence intervals. Mean.Ann.Temp. and (Mean.Ann.Temp.)², the linear and quadratic terms of mean annual temperature (bio1); Temp.Range, annual range of temperature (bio7); Ann.Prec. and (Ann.Prec.)², the linear and quadratic terms of annual precipitation (bio12); Prec.Seasonality, precipitation seasonality (bio15); Prec.Warmest.Quarter, mean monthly precipitation of the warmest quarter (bio18); Elev.Range, elevation range; Anom.Temp.LGM, temperature anomaly since Last Glacial Maximum; Anom.Prec.LGM, precipitation anomaly since the Last Glacial Maximum.

Figure S5: Shifts in estimated non‐linear relationships between vertebrate species richness and climate. Partial residual plots illustrating the effects of mean annual temperature and annual precipitation on species richness between 1920 (blue) and 2020 (orange) for (a, b) birds, (c, d) mammals, (e, f) reptiles and (g, h) amphibians. Dashed line represents the estimated optimal temperature or precipitation between 1920 and 2020, defined here as temperature or precipitation values associated with the highest richness derived from the linear and quadratic terms. Coloured band indicates 95% confidence interval. Mean annual temperature and annual precipitation were back‐transformed to the original scale to facilitate the biological interpretation. The y‐axis represents the natural log‐transformed species richness attributed to temperature or precipitation alone in the models.

Figure S6: Temporal trends in taxonomic composition. The bar plots show changes in the proportion of species from major taxonomic groups within the total number of known species for each decade from 1920 to 2020 for (a) reptiles and (b) amphibians. The categories of taxonomic groups for reptiles and amphibians follow Moura and Jetz (2021).

Figure S7: Differences in taxonomic composition at the family level between 1920 and 2020. The scatter plots show the differences in the proportion of known species number for each family relative to the total species number of (a) reptiles and (b) amphibians. This illustrates how species discoveries have altered the taxonomic composition of these groups, particularly reptile infraorder Gekkota and amphibian superfamilies Hyloidea and Ranoidea. Annotated labels highlight families with distinct differences (difference > 0.01) or represent high proportion in 1920 (proportion > 0.05).

Table S1: Proportion of shifts in terrestrial vertebrate diversity centers, 1920–2020 and 2000–2020. The table shows the proportion of diversity centers (top 5% of area) that have shifted over two time periods for terrestrial vertebrates. The main values correspond to centers of rarity, and the values in parentheses correspond to centers of species richness. Emergent centers, regions that emerged as diversity centers after new species discoveries; historical centers, regions initially identified as diversity centers but have since diminished; stable centers, regions that remained stable across discoveries.

Table S2: Coefficients of the relationships between species richness and environments. The table presents standardized coefficients from simultaneous autoregressive models that quantify the relationship between species richness and environmental variables in 1920 and 2020. Asterisks indicate the significance in environmental variable coefficients: p < 0.001 (***), 0.001 < p < 0.01 (**) and 0.01 < p < 0.05 (*). The significance of the difference in coefficients for each variable between the two time periods was assessed using a Z test, with the resulting Z values and p values presented.

Table S3: Coefficients of the relationships between species richness and environments for each realm. The table presents standardized coefficients from simultaneous autoregressive models that quantify the relationship between species richness and environmental variables for each realm in 1920 and 2020. Asterisks indicate the significance in environmental variable coefficients: p < 0.001 (***), 0.001 < p < 0.01 (**) and 0.01 < p < 0.05 (*). The significance of the difference in coefficients for each predictor between the two time periods was assessed using a Z test, with the resulting Z values and p values presented.

Data Availability Statement

Global biological and environment data used in our study are open source. Species range maps are availability from several public repositories: birds from the Handbook of the Birds of the World and BirdLife International version 7 (https://datazone.birdlife.org); mammals and amphibians from the IUCN Red List of Threatened Species version 2023‐12 (https://www.iucnredlist.org/); and reptiles from Roll et al. (2017) (https://doi.org/10.1038/s41559‐017‐0332‐2). Taxonomic data are available as follows: birds from the Handbook of the Birds of the World and BirdLife International version 7 (https://datazone.birdlife.org); mammals from the IUCN Red List of Threatened Species (https://www.iucnredlist.org/) and The Mammal Diversity Database (https://www.mammaldiversity.org/); reptiles from The Reptile Database (https://reptile‐database.org/); amphibians from the IUCN Red List of Threatened Species (https://www.iucnredlist.org/) and Amphibian Species of the World 6.2 (https://amphibiansoftheworld.amnh.org/). Environment data are available at CHELSA (https://chelsa‐climate.org/) and GMTED2010 (http://pubs.er.usgs.gov/publication/ofr20111073). The processed data and code necessary to replicate the analyses are available on Dryad (https://doi.org/10.5061/dryad.66t1g1kdj).