Temporal Bias Research Articles

Instagram data for validating Nothofagus pumilio distribution mapping in the Southern Andes: A novel ground truthing approach

The availability of valid, non-biased species occurrence data has always been a major challenge for biodiversity research and modelling studies. Data from open-source databases or remote sensing are promising approaches to increase the availability of species occurrence data. However, these data may contain spatial, temporal, and taxonomic biases or require ground truthing. In recent years, social media has received attention for its contribution to species occurrence data sampling and ground truthing approaches. The wide reach of social media platforms allows for rapid and large-scale analyses. Here we introduce a novel Instagram ground truthing approach to validate the occurrence mapping of Nothofagus pumilio across its entire distribution range. The treeline species of the southern Andes has been extensively studied in small-scale studies, but large-scale modelling approaches are largely missing due to limited accessibility to treeline sites resulting in a lack of occurrence data. The content posted on the social media platform Instagram consists of images and videos in which the species N. pumilio and its location can be identified. By searching for suitable posts using hashtags and location tags, we created 1238 Instagram ground truthing points. We compared the performance of our dataset with open-source data from the Global Biodiversity Information Facility (GBIF) through visual, quantitative, and bias analyses, acknowledging that both social media-based and Citizen Science data are subject to sampling and spatial biases due to collection in human-accessible areas. The Instagram ground truthing points were subsequently used to validate remote sensing occurrence data, generated using Sentinel-2 level 2A data and Supervised Classification. The combined approach – Instagram ground truthing and remote sensing – allows for the collection of occurrence data over the entire latitudinal range of N. pumilio, covering approximately 2000 km. The use of social media content provides potentially important contributions to species occurrence data sampling and ground truthing In our study we introduce a novel ground truthing approach for species occurrence data sampling based on Instagram data Instagram ground truthing points, combined with Supervised Classification generate species occurrence data of Nothofagus pumilio over its entire distribution range in the southern Andes The performance of the Instagram ground truthing points is evaluated by comparison with existing data from the GBIF database. Our Instagram ground truthing approach demonstrates a new way of sampling species occurrence data and can be applied to other suitable species and study areas.

Decoding and modifying dynamic attentional bias in gaming disorder.

With the spread of smartphones and computer games, concerns have escalated regarding the rising prevalence of gaming disorder. Patients often display attentional biases, unconsciously turning their attention towards gaming-related stimuli. However, attempts to discover and ameliorate these attentional deficits have yielded inconsistent outcomes, potentially due to the dynamic nature of attentional bias. This study investigated neural mechanisms underlying attentional bias state by combining neuroimaging (functional magnetic resonance imaging -fMRI) with an approach-avoidance task tailored to an individual's gaming preference. We conducted a multivariate pattern analysis of endogenous brain activity in 21 participants with probable gaming disorder. Our analyses revealed that activity patterns in the insula tracked temporal attentional bias states specific to gaming stimuli. A broad network of frontal and parietal regions instead appeared to predict a general temporal attentional bias state. Finally, we conducted a proof-of-concept study for 'just-in-time' attentional bias training through fMRI-decoded neurofeedback of insula activity patterns, named decoded attentional bias training (DecABT). Our preliminary results suggest that DecABT may help to decrease the attractiveness of gaming stimuli via a insula- and precuneus-based neural mechanism. This work provides new evidence for the insula as an endogenous regulator of attentional bias states in gaming disorder and a starting point to develop novel, individualized therapeutic approaches to treat addiction.This article is part of the theme issue 'Neurofeedback: new territories and neurocognitive mechanisms of endogenous neuromodulation'.

Distinct effects of slow and fast theta tACS in enhancing temporal memory

Abstract Temporal memory plays a crucial role in organizing real-life events into meaningful segments. Previous research suggests that the clustering of temporally related information builds on the neural correlates of associative memory, including theta oscillations. Transcranial alternating current stimulation (tACS) provides a means of modulating theta oscillations within associative memory networks, possibly including hippocampal modulation when targeting parietal cortex. Theta oscillations are not limited to a single frequency range, but instead, exhibit functional specialization, with slow theta (3 Hz) implicated in short-term episodic memory formation and fast theta (8 Hz) involved in spatial navigation. This study aimed to investigate the distinct effects of slow and fast theta stimulation on temporal memory. Participants encoded visual objects paired with frame color while receiving tACS stimulation at 8 Hz, 3 Hz, or sham targeting the left parietal cortex. The frame color would change for every eight objects, establishing a context boundary with each color change. Subsequently, participants performed a timeline task to assess temporal memory performance. Results showed that slow, but not fast, theta stimulation led to an enhancement in temporal accuracy (absolute temporal error) compared to sham tACS, in support of our main hypothesis. Under sham stimulation, participants consistently underestimated the temporal position of items presented further away from boundary, compared to those presented at boundary. This finding resembled temporal compression observed during event segmentation. Interestingly, fast, but not slow, theta stimulation reduced this temporal bias (rated position— actual position). This study represents the first tACS evidence for differential contributions of slow versus fast theta to temporal memory formation in humans. We speculate that parietal theta tACS may modulate the hippocampus and facilitate temporal memory formation.

Addressing multiple facets of bias and uncertainty in continental scale biodiversity databases

The availability of biodiversity databases is expanding at unprecedented rates. Nevertheless, species occurrence data can be intrinsically biased and contain uncertainties that impact the accuracy and reliability of biodiversity estimates. In this study, we developed a reproducible framework to assess three dimensions of bias—taxonomic, spatial, and temporal—as well as temporal uncertainty associated with data collections. We utilized the vegetation plot data located in Europe, from sPlotOpen, an open-access database, as a case study. The metrics proposed for estimating bias include completeness of the species richness for taxonomic bias, Nearest Neighbor Index for spatial bias, and Pielou’s index for temporal bias. Additionally, we introduced a new method based on a negative exponential curve to model the temporal decay in biodiversity data, aiming to quantify temporal uncertainty. Finally, we assessed the sampling bias considering the influence of various spatial variables (i.e, road density, human population count, Natura 2000 network and topographic roughness). We discovered that the facets of bias and the temporal uncertainty varied throughout Europe, as did the different roles played by spatial variables in determining biases. sPlotOpen showed a clustered distribution of the vegetation plots, and an uneven distribution in sampling completeness, year of sampling and temporal uncertainty. The facets of bias were significantly explained mainly by the presence of Natura 2000 network and marginally by the human population count. These results suggest that employing an efficient procedure to examine biases and uncertainties in data collections can enhance data quality and provide more reliable biodiversity estimates.

Searching for Linearity: Reconstructive Processes Reverse Temporal Scrambling in Memory for Movie Scenes

Meaning-making and temporal memory are closely intertwined, yet we still do not know how the overall understanding of complex events affects retrospective temporal judgments. The present study investigated the effect of a manipulation of the temporal linearity of a narrative on the subsequent memory-for-time performance. Participants indicated the time of occurrence of short video clips extracted from a previously encoded movie on a horizontal timeline representing the movie duration. Importantly, a group of participants ( N = 20) watched the original movie, which depicts events occurring in chronological order, whereas another group ( N = 30) watched a scrambled version of the same movie in which the temporal linearity was lost. This procedure allowed us to measure the quantity and direction of the temporal memory bias. The scrambled presentation produced a mild and general impairment of recognition memory compared to the linear presentation. More importantly, it biased temporal judgments as a function of the direction and amount of discrepancy between the story and the viewing time, in accordance with an automatic reshaping of temporal memory caused by a chronological representation of the storyline. This effect could be distinguished from a tendency to move judgments toward the center of the timeline, independently from the specific scrambling arrangement, consistent with the idea that the non-linearity of the story also generally increased the degree of temporal uncertainty. Taken together, our results provide further evidence that temporal memories are automatically reconstructed according to the general meaning of the events.

Self-attention with temporal prior: can we learn more from the arrow of time?

Many diverse phenomena in nature often inherently encode both short- and long-term temporal dependencies, which especially result from the direction of the flow of time. In this respect, we discovered experimental evidence suggesting that interrelations of these events are higher for closer time stamps. However, to be able for attention-based models to learn these regularities in short-term dependencies, it requires large amounts of data, which are often infeasible. This is because, while they are good at learning piece-wise temporal dependencies, attention-based models lack structures that encode biases in time series. As a resolution, we propose a simple and efficient method that enables attention layers to better encode the short-term temporal bias of these data sets by applying learnable, adaptive kernels directly to the attention matrices. We chose various prediction tasks for the experiments using Electronic Health Records (EHR) data sets since they are great examples with underlying long- and short-term temporal dependencies. Our experiments show exceptional classification results compared to best-performing models on most tasks and data sets.

Gymnosperm collecting in China: Past, present, and future

AbstractAssessing sampling biases caused by nonrandom specimen collecting is crucial in systematics, biogeography, and conservation. Nevertheless, research on the collecting biases of vascular plants in China remains limited. Here, we investigated the collecting status of gymnosperms in China using 48 673 herbarium specimens representing 180 Chinese gymnosperm species. The spatial and temporal patterns, collector and determiner biases, and phylogenetic and taxon biases were studied to comprehend the collecting bias of gymnosperms in China during 1900–2021. Meanwhile, we assessed the inventory completeness of gymnosperms to identify collecting hotspots and coldspots. The results showed that gymnosperms predominantly inhabit mountainous areas in China. The historical collecting of gymnosperms in China can be divided into two distinct stages with four peaks of collecting activities. The distribution of collected or identified specimens per individual displays significant skewness, and the collecting of gymnosperms has the issues of overcollecting or undersampling. Inventory completeness remains deficient, and collecting intensity is weak in both northern China and the range of 22°–25° N in the southern region (i.e., central Yunnan, southern Guangxi, and eastern Guangdong). Additionally, observation of the spatial distribution reveals both hotspots and coldspots scattered across mountain regions, lacking clustering. This study revealed a highly polarized collecting pattern for gymnosperms in China. Consequently, we recommend optimizing collecting measures, targeted specimen collecting, and continuous dynamic monitoring of gymnosperms in future collecting efforts. These recommendations hold relevance not only for gymnosperms in China but also for other land plants worldwide.

Functional consequences of spatial, temporal and ligand bias of G protein-coupled receptors.

G protein-coupled receptors (GPCRs) regulate every aspect of kidney function by mediating the effects of various endogenous and exogenous substances. A key concept in GPCR function is biased signalling, whereby certain ligands may selectively activate specific pathways within the receptor's signalling repertoire. For example, different agonists may induce biased signalling by stabilizing distinct active receptor conformations - a concept that is supported by advances in structural biology. However, the processes underlying functional selectivity in receptor signalling are extremely complex, involving differences in subcellular compartmentalization and signalling dynamics. Importantly, the molecular mechanisms of spatiotemporal bias, particularly its connection to ligand binding kinetics, have been detailed for GPCRs critical to kidney function, such as the AT1 angiotensin receptor (AT1R), V2 vasopressin receptor (V2R) and the parathyroid hormone 1 receptor (PTH1R). This expanding insight into the multifaceted nature of biased signalling paves the way for innovative strategies for targeting GPCR functions; the development of novel biased agonists may represent advanced pharmacotherapeutic approaches to the treatment of kidney diseases and related systemic conditions, such as hypertension, diabetes and heart failure.

Revealing biases in insect observations: A comparative analysis between academic and citizen science data.

Citizen Science is a powerful tool for biodiversity research, as it facilitates data recording at large scales that would otherwise be impossible to cover by standard academic research. Despite its benefits, the accuracy of citizen science data remains a subject of concern among scientists, with varying results reported so far. Neither citizen science data nor academic records are immune to biases, which can significantly impact the quality and reliability of observations. Here, using insects in the Iberian Peninsula as a case study, we compare data collected by participatory platforms to those obtained through academic research projects, and assess their taxonomic, spatial, temporal, and environmental biases. Results show a prominent taxonomic bias in both academic and citizen science data, with certain insect orders receiving more attention than others. These taxonomic biases are conserved between different participatory platforms, as well as between groups of users with different levels of contribution performance. The biases captured by leading contributors in participatory platforms mirrored those of sporadic users and academic data. Citizen science data had higher spatial coverage and less spatial clustering than academic data, showing also clearer trends in temporal seasonality. Environmental coverage over time was more stable in citizen science than in academic records. User behaviour, preference, taxonomical expertise, data collection methodologies and external factors may contribute to these biases. This study shows the multifaceted nature of biases present in academic records and citizen science platforms. The insights gained from this analysis emphasize the need for careful consideration of these biases when making use of biodiversity data from different sources. Combining academic and citizen science data enhances our understanding of biodiversity, as their integration offers a more comprehensive perspective than relying solely on either dataset alone, especially since biases in these two types of data are not always the same.

Using citizen science data for predicting the timing of ecological phenomena across regions.

The scarcity of long-term observational data has limited the use of statistical or machine-learning techniques for predicting intraannual ecological variation. However, time-stamped citizen-science observation records, supported by media data such as photographs, are increasingly available. In the present article, we present a novel framework based on the concept of relative phenological niche, using machine-learning algorithms to model observation records as a temporal sample of environmental conditions in which the represented ecological phenomenon occurs. Our approach accurately predicts the temporal dynamics of ecological events across large geographical scales and is robust to temporal bias in recording effort. These results highlight the vast potential of citizen-science observation data to predict ecological phenomena across space, including in near real time. The framework is also easily applicable for ecologists and practitioners already using machine-learning and statistics-based predictive approaches.

Estimating sampling biases in citizen science datasets

The rise of citizen science (also called community science) has led to vast quantities of species observation data collected by members of the public. Citizen science data tend to be unevenly distributed across space and time, but the treatment of sampling bias varies between studies, and interactions between different biases are often overlooked. We present a method for conceptualizing and estimating spatial and temporal sampling biases, and interactions between them. We use this method to estimate sampling biases in an example ornithological citizen science dataset from eBird in Brisbane City, Australia. We then explore the effects of these sampling biases on subsequent model inference of population trends, using both a simulation study and an application of the same trend models to the Brisbane eBird dataset. We find varying levels of sampling bias in the Brisbane eBird dataset across temporal and spatial scales, and evidence for interactions between biases. Several of the sampling biases we identified differ from those described in the literature for other datasets, with protected areas being undersampled in the city, and only limited seasonal sampling bias. We demonstrate variable performance of trend models under different sampling bias scenarios, with more complex biases being associated with typically poorer trend estimates. Sampling biases are important to consider when analysing ecological datasets, and analysts can use this method to ensure that any biologically relevant sampling biases are detected and given due consideration during analysis. With appropriate model specification, the effects of sampling biases can be reduced to yield reliable information about biodiversity.

Late Cretaceous ammonoids show that drivers of diversification are regionally heterogeneous

Palaeontologists have long sought to explain the diversification of individual clades to whole biotas at global scales. Advances in our understanding of the spatial distribution of the fossil record through geological time, however, has demonstrated that global trends in biodiversity were a mosaic of regionally heterogeneous diversification processes. Drivers of diversification must presumably have also displayed regional variation to produce the spatial disparities observed in past taxonomic richness. Here, we analyse the fossil record of ammonoids, pelagic shelled cephalopods, through the Late Cretaceous, characterised by some palaeontologists as an interval of biotic decline prior to their total extinction at the Cretaceous-Paleogene boundary. We regionally subdivide this record to eliminate the impacts of spatial sampling biases and infer regional origination and extinction rates corrected for temporal sampling biases using Bayesian methods. We then model these rates using biotic and abiotic drivers commonly inferred to influence diversification. Ammonoid diversification dynamics and responses to this common set of diversity drivers were regionally heterogeneous, do not support ecological decline, and demonstrate that their global diversification signal is influenced by spatial disparities in sampling effort. These results call into question the feasibility of seeking drivers of diversity at global scales in the fossil record.

Simultaneous Spatiotemporal Bias Compensation and Data Fusion for Asynchronous Multisensor Systems

Bias estimation of sensors is an essential prerequisite for accurate data fusion. Neglect of temporal bias in general real systems prevents the existing algorithms from successful application. In this paper, both spatial and temporal biases in asynchronous multisensor systems are investigated and two novel methods for simultaneous spatiotemporal bias compensation and data fusion are presented. The general situation that the sensors sample at different times with different and varying periods is explored, and unknown time delays may exist between the time stamps and the true measurement times. Due to the time delays, the time stamp interval of the measurements from different sensors may be different from their true measurement interval, and the unknown difference between them is considered as the temporal bias and augmented into the state vector to be estimated. Multisensor measurements are collected in batch processing or sequential processing schemes to estimate the augmented state vector, results in two spatiotemporal bias compensation methods. In both processing schemes, the measurements are formulated as functions of both target states and spatiotemporal biases according to the time difference between the measurements and the states to be estimated. The Unscented Kalman Filter is used to handle the nonlinearity of the measurements and produce spatiotemporal bias and target state estimates simultaneously. The posterior Cramer-Rao lower bound (PCRLB) for spatiotemporal bias and state estimation is presented and simulations are conducted to demonstrate the effectiveness of the proposed methods.

Uncovering milestone papers: A network diffusion and game theory approach

Methods to rank documents in large-scale citation data are increasingly assessed in terms of their ability to identify small sets of expert-selected papers. Here, we propose an algorithm for the accurate identification of milestone papers from citation networks. The algorithm combines an influence propagation process with game theory concepts. It outperforms state-of-the-art metrics in the identification of milestone papers in aggregate citation network data, while potentially mitigating the ranking's temporal bias compared with metrics that have similar milestone identification performance. The proposed method sheds light on the interplay between ranking accuracy and temporal bias.

Human-like Learning in Temporally Structured Environments

Natural environments have correlations at a wide range of timescales. Human cognition is tuned to this temporal structure, as seen by power laws of learning and memory, and by spacing effects whereby the intervals between repeated training data affect how long knowledge is retained. Machine learning is instead dominated by batch iid training or else relatively simple nonstationarity assumptions such as random walks or discrete task sequences. The main contributions of our work are: (1) We develop a Bayesian model formalizing the brain's inductive bias for temporal structure and show our model accounts for key features of human learning and memory. (2) We translate the model into a new gradient-based optimization technique for neural networks that endows them with human-like temporal inductive bias and improves their performance in realistic nonstationary tasks. Our technical approach is founded on Bayesian inference over 1/f noise, a statistical signature of many natural environments with long-range, power law correlations. We derive a new closed-form solution to this problem by treating the state of the environment as a sum of processes on different timescales and applying an extended Kalman filter to learn all timescales jointly. We then derive a variational approximation of this model for training neural networks, which can be used as a drop-in replacement for standard optimizers in arbitrary architectures. Our optimizer decomposes each weight in the network as a sum of subweights with different learning and decay rates and tracks their joint uncertainty. Thus knowledge becomes distributed across timescales, enabling rapid adaptation to task changes while retaining long-term knowledge and avoiding catastrophic interference. Simulations show improved performance in environments with realistic multiscale nonstationarity. Finally, we present simulations showing our model gives essentially parameter-free fits of learning, forgetting, and spacing effects in human data. We then explore the analogue of human spacing effects in a deep net trained in a structured environment where tasks recur at different rates and compare the model's behavioral properties to those of people.

Measuring temporal bias in sequential numerosity comparison

While several methods have been proposed to assess the influence of continuous visual cues in parallel numerosity estimation, the impact of temporal magnitudes on sequential numerosity judgments has been largely ignored. To overcome this issue, we extend a recently proposed framework that makes it possible to separate the contribution of numerical and non-numerical information in numerosity comparison by introducing a novel stimulus space designed for sequential tasks. Our method systematically varies the temporal magnitudes embedded into event sequences through the orthogonal manipulation of numerosity and two latent factors, which we designate as “duration” and “temporal spacing”. This allows us to measure the contribution of finer-grained temporal features on numerosity judgments in several sensory modalities. We validate the proposed method on two different experiments in both visual and auditory modalities: results show that adult participants discriminated sequences primarily by relying on numerosity, with similar acuity in the visual and auditory modality. However, participants were similarly influenced by non-numerical cues, such as the total duration of the stimuli, suggesting that temporal cues can significantly bias numerical processing. Our findings highlight the need to carefully consider the continuous properties of numerical stimuli in a sequential mode of presentation as well, with particular relevance in multimodal and cross-modal investigations. We provide the complete code for creating sequential stimuli and analyzing participants’ responses.

Consistency of Aerosol Optical Properties between MODIS Satellite Retrievals and AERONET over a 14-Year Period in Central–East Europe

Aerosols influence Earth’s climate by interacting with radiation and clouds. Remote sensing techniques aim to enhance our understanding of aerosol forcing using ground-based and satellite retrievals. Despite technological advancements, challenges persist in reducing uncertainties in satellite remote sensing. Our study examines retrieval biases in MODIS sensors on Terra and Aqua satellites compared to AERONET ground-based measurements. We assess their performance and the correlation with the AERONET aerosol optical depth (AOD) using 14 years of data (2010–2023) from 29 AERONET stations across 10 Central–East European countries. The results indicate discrepancies between MODIS Terra and Aqua retrievals: Terra overestimates the AOD at 16 AERONET stations, while Aqua underestimates the AOD at 21 stations. The examination of temporal biases in the AOD using the calculated estimated error (ER) between AERONET and MODIS retrievals reveals a notable seasonality in coincident retrievals. Both sensors show higher positive AOD biases against AERONET in spring and summer compared to fall and winter, with few ER values for Aqua indicating poor agreement with AERONET. Seasonal variations in correlation strength were noted, with significant improvements from winter to summer (from R2 of 0.58 in winter to R2 of 0.76 in summer for MODIS Terra and from R2 of 0.53 in winter to R2 of 0.74 in summer for MODIS Aqua). Over the fourteen-year period, monthly mean aerosol AOD trends indicate a decrease of −0.00027 from AERONET retrievals and negative monthly mean trends of the AOD from collocated MODIS Terra and Aqua retrievals of −0.00023 and −0.00025, respectively. An aerosol classification analysis showed that mixed aerosols comprised over 30% of the total aerosol composition, while polluted aerosols accounted for more than 22%, and continental aerosols contributed between 22% and 24%. The remaining 20% consists of biomass-burning, dust, and marine aerosols. Based on the aerosol classification method, we computed the bias between the AERONET AE and MODIS AE, which showed higher AE values for AERONET retrievals for a mixture of aerosols and biomass burning, while for marine aerosols, the MODIS AE was larger and for dust the results were inconclusive.

Memory capacity as the core mechanism of the development of space–time interferences in children

This study investigated the development of spatiotemporal perceptual interactions in 5-to-7 years old children. Participants reproduced the temporal and spatial interval between sequentially presented visual stimuli. The time and spacing between stimuli were experimentally manipulated. In addition, cognitive capacities were assessed using neuropsychological tests. Results revealed that starting at 5 years old, children exhibited spatial biases in their time estimations and temporal biases in their spatial estimations, pointing at space–time interference. In line with developmental improvement of temporal and spatial abilities, these spatiotemporal biases decreased with age. Importantly, short-term memory capacity was a predictor of space–time interference pointing to shared cognitive mechanisms between time and space processing. Our results support the symmetrical hypothesis that proposes a common neurocognitive mechanism for processing time and space.

Moths are less attracted to light traps than they used to be

As evidence of global insect declines continues to mount, insect conservationists are becoming increasingly interested in modeling the demographic history of at-risk species from long-term survey data. However, certain entomological survey methods may be susceptible to temporal biases that will complicate these efforts. Entomological light traps, in particular, may catch fewer insects today than they once did due solely to increases in anthropogenic light pollution. Here we investigate this possibility by comparing the demographic histories of corn earworm moths (Helicoverpa zea) estimated from pairs of blacklight and pheromone traps monitored at the same farms. We find a stark decline in blacklight trap efficacy over 25 years of monitoring in Delaware, USA, mirrored over 10 years of monitoring in New Jersey, USA. While the precise causes of this decline remain a subject for discussion, the practical consequences are clear: insect conservationists cannot fully rely on long-term trends from entomological light traps.

Herbarium data accurately predict the timing and duration of population‐level flowering displays

Forecasting the impacts of changing climate on the phenology of plant populations is essential for anticipating and managing potential ecological disruptions to biotic communities. Herbarium specimens enable assessments of plant phenology across broad spatiotemporal scales. However, specimens are collected opportunistically, and it is unclear whether their collection dates – used as proxies of phenological stages – are closest to the onset, peak, or termination of a phenophase, or whether sampled individuals represent early, average, or late occurrences in their populations. Despite this, no studies have assessed whether these uncertainties limit the utility of herbarium specimens for estimating the onset and termination of a phenophase. Using simulated data mimicking such uncertainties, we evaluated the accuracy with which the onset and termination of population‐level phenological displays (in this case, of flowering) can be predicted from natural‐history collections data (controlling for biases in collector behavior), and how the duration, variability, and responsiveness to climate of the flowering period of a species and temporal collection biases influence model accuracy. Estimates of population‐level onset and termination were highly accurate for a wide range of simulated species' attributes, but accuracy declined among species with longer individual‐level flowering duration and when there were temporal biases in sample collection, as is common among the earliest and latest‐flowering species. The amount of data required to model population‐level phenological displays is not impractical to obtain; model accuracy declined by less than 1 day as sample sizes rose from 300 to 1000 specimens. Our analyses of simulated data indicate that, absent pervasive biases in collection and if the climate conditions that affect phenological timing are correctly identified, specimen data can predict the onset, termination, and duration of a population's flowering period with similar accuracy to estimates of median flowering time that are commonplace in the literature.