Urban Environments Reshape Reproductive Phenology in Plants Across the Tropics

19 Plant phenological responses to global change phenomena like urbanization remain understudied 20 in the tropics, hindering predictions regarding the dynamics of tropical ecosystems amid rapid 21 land use changes. Studies of tropical phenology are limited by complexities, like the limited 22 availability of phenological data, especially in urbanized landscapes. Observations recorded on 23 citizen science platforms can overcome this limitation by providing vast, spatially distributed 24 data. In this study, we utilize iNaturalist data to evaluate plant reproductive phenology in tropical 25 urban vs. rural habitats. We first compare iNaturalist data (111533 records) to herbarium 26 collections (217991 records) in order to validate their use, and we then investigate urban-rural 27 phenology differences within 25-km spatial grids for 238 species. Data from iNaturalist and 28 herbaria yield complementary insights, with the former being uniformly distributed between 29 urban and rural settings, and the latter biased towards rural observations. On average, we found 30 species to have significantly longer reproductive duration (β = 11.79 ± 2.83 SE, t = 4.16, p < 31 10^4), and correspondingly weaker strength of seasonality in urban settings than in nearby rural 32 localities. We also find trait-mediated variation, with seasonal, annual, and herbaceous plants 33 showing more pronounced differences in reproductive duration and seasonality strength. These 34

suggest that urbanization in tropical landscapes might have important implications for 35 plant demography, with potential consequences for community and ecosystem dynamics. Our 36 work also points to the value of integrating insights from natural history collections with data 37 from citizen science platforms for enabling broad-scale insights into ecological dynamics in 38 tropical urban landscapes. 39 40 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 3 41 Key words: 42 Plant phenology, urbanization, tropical ecosystems, reproductive seasonality, citizen science, 43 iNaturalist, herbarium records 44 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 4 1. INTRODUCTION 45 Phenology, the study of the timing of recurring biological events, links organisms to their abiotic 46 environment, including temperature, precipitation, and photoperiod, and underpins ecosystem 47 functioning (Caparros-Santiago et al., 2021; Johansson et al., 2015). In terrestrial systems, the 48 timing of events such as leaf-out, flowering, and fruiting in plants can influence primary 49 productivity, nutrient cycling, and energy flow within ecosystems (Forrest & Miller-Rushing, 50 2010; Gallinat et al., 2021; Moore et al., 2016; Tang et al., 2016), as well as ecological 51 interactions like pollination, herbivory, predation, and competition (Forrest & Miller-Rushing, 52 2010; Johansson et al., 2015; Kharouba et al., 2018). Shifts in plant phenology are among the 53 earliest and most consistent biological responses to global environmental change, serving as key 54 indicators of species’ adaptive capacity and ecosystem resilience (Cleland et al., 2007; Parmesan 55 & Yohe, 2003). Most current understanding of these impacts comes from temperate regions, 56 where seasonal variations in important cues like climate and daylength are well-defined across 57 the year. For example, Menzel et al., (2006)’s landmark study linking phenology to climate 58 warming relied on phenological data from a systematic network across Europe to show that a 59 vast majority of plant species show signals of advanced leafout, flowering, and fruiting, 60 consistent with warmer winter temperatures and earlier spring conditions. Climate-driven shifts 61 in plant phenology have now been identified across temperate regions (Stuble et al., 2021), 62 although the dynamics and drivers of these shifts can vary across continents (Zohner et al., 63 2017). Conversely, much less is known about phenological responses to global environmental 64 change in tropical ecosystems (Piao et al., 2019), where rainfall patterns and uniform 65 temperatures give rise to complex, multi-modal seasonal patterns (Abernethy et al., 2018; 66 Fitchett et al., 2015). 67 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 5 68 Among the most prominent forces for environmental change in the tropics is 69 urbanization, which is associated with changes to local temperature patterns, water availability, 70 and biological diversity (Kabano et al., 2021; McDonald et al., 2020; Ogunbode et al., 2025; 71 Ribeiro et al., 2024). Studies of urban phenological shifts are useful for identifying urban heat 72 islands and evaluating impacts of warming on phenology (Jochner et al., 2013). In temperate 73 cities, urban heat islands and irrigation systems often reduce climatic constraints on plant 74 development, leading to earlier or extended reproductive periods in cities (Li et al., 2019; Neil & 75 Wu, 2006; D. S. Park et al., 2023; Wohlfahrt et al., 2019). Although urbanization has been rapid 76 and widespread throughout the tropics and is of particular concern for tropical biodiversity 77 conservation (McDonald et al., 2020; Simkin et al., 2022), tropical cities are surprisingly 78 understudied in phenological research (Bonebrake et al., 2025), which constrains predictions 79 regarding the resilience of tropical plant dynamics amid rapid climatic and land use changes. 80 Understanding how urbanization influences reproductive phenology in the tropics is therefore a 81 pressing research gap (Kabano et al., 2021; Marcacci et al., 2023). 82 Several challenges complicate studies of tropical phenology. First, unlike in temperate 83 regions, where strong seasonal temperature and photoperiod cues trigger synchronized 84 phenological events, tropical phenology is influenced by a broader suite of environmental cues 85 such as rainfall, irradiance, and local microclimate (Borchert, 1996; Borchert et al., 2002; 86 Jochner et al., 2013; Numata et al., 2022). These cues may operate asynchronously, and their 87 timing, strength, and biological relevance can vary across space, years, and species phenology 88 (Abernethy et al., 2018; J. Y. Park et al., 2019; Sakai, 2001; Singh & Kushwaha, 2005). Second, 89 tropical ecosystems are generally more species-rich than temperate regions, which results in 90 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 6 tropical plant communities encompassing a wider variety of phenological strategies (Abernethy 91 et al., 2018; J. Y. Park et al., 2019; Singh & Kushwaha, 2005). Along with this, plants’ life 92 history and growth form also play a definite role in phenological responses, as annual and 93 perennial species and contrasting growth forms vary in resource allocation strategies and 94 dependency on environmental cues (Borchert, 1996; Sakai, 2001). Finally, the study of tropical 95 phenology has historically been constrained by the limited availability of long-term phenological 96 monitoring plots and field stations (Abernethy et al., 2018; Bush et al., 2017, 2018). Although 97 remote sensing techniques can capture seasonal changes in canopy greenness, they fall short in 98 capturing species- and population-level phenology (Fisher et al., 2006; Zhang et al., 2006) 99 (Fisher et al., 2006; Zhang et al., 2006). 100 One potential approach to overcome these limitations is through using data derived from 101 citizen science platforms, which have the potential to transform phenology research by providing 102 vast, globally distributed datasets of species occurrences. These data are timestamped, 103 georeferenced, and often include photographs that enable the inference of phenological events 104 across unprecedented spatial and temporal scales (Barve et al., 2020; Callaghan et al., 2020; 105 Wolf et al., 2022). For instance, iNaturalist, a widely used citizen science platform, currently 106 hosts over 69 million research grade plant observations. 4.6 million of these observations come 107 from tropical Asia, a traditionally data-scarce region, and this number is increasing rapidly (Di 108 Cecco et al., 2021). In temperate ecosystems, iNaturalist records have been proven to yield 109 valuable insights into plant phenology. For example, Li et al. (2019) used 22 million images to 110 extract phenology records from the USA and Europe and found that urbanization advances 111 flowering and leaf-out in colder regions. Similarly, (Iwanycki Ahlstrand et al., 2022) compared 112 directed citizen science, herbarium, and iNaturalist records for three spring flowering species in 113 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 7 Denmark, and found that iNaturalist provided the broadest spatial coverage and captured peak 114 flowering well. Moreover, combining citizen science dataset with traditional monitoring 115 networks and herbarium records can help fill spatial and temporal gaps, enhancing the detection 116 of climate-driven phenological changes (Davis et al., 2015; D. S. Park et al., 2023; Willis et al., 117 2017). In sum, despite important limitations such as observer bias, variable sampling effort, and 118 uneven taxonomic coverage, opportunistically collected observation records in citizen science 119 platforms can yield meaningful phenological insights, especially when analyzed with modern 120 circular statistics and hierarchical modeling techniques (Capinha et al., 2024; Lai, 2025; Pabon-121 Moreno et al., 2019; Willig et al., 2024). 122 In this study, we utilize iNaturalist data to evaluate plant reproductive phenology in urban 123 vs. rural habitats across the tropical latitudes. We first evaluated the value of iNaturalist 124 observations for tropical phenological research by comparing phenological estimates derived 125 from these observations to those derived from herbarium specimens collected in tropical 126 latitudes. Next, we asked whether plant reproductive phenology in urban centers differs from 127 reproductive phenology in adjacent rural areas. Finally, we evaluated whether plant growth and 128 reproductive strategies impact phenological responses to urbanization. Our study offers one of 129 the first large-scale, trait-specific evaluations of how urbanization influences reproductive 130 phenology in tropical plants. 131 132 2. METHODS 133 2.1. Data acquisition 134 135 2.1.1 Plant observations from iNaturalist 136 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 8 As a source of our citizen science data, we used plant data collected through iNaturalist (Fig. 1, 137 left panel). To retrieve observations along with their phenology status (presence of flowers, 138 fruits, or both), we accessed data through PhenoBase (Dinnage et al., 2025). PhenoBase, through 139 its machine learning system known as PhenoVision, has processed millions of field photos from 140 iNaturalist and provides them with phenology status and geographic and temporal metadata. 141 PhenoVison detects flowers and fruits using a Vision Transformer (ViT) model fine-tuned with a 142 masked autoencoder pretraining method. This machine learning program was trained on over 1.5 143 million human-annotated iNaturalist images, achieving high validation accuracy (98.5% for 144 flowers, 95% for fruits). Additional calibration steps refined detection thresholds and reduced 145 false positives, ensuring high-confidence phenological observations. As annotations on 146 PhenoBase are limited to observations between 2010 and 2023, we limited our search to those 147 years. We downloaded 217,991 observations representing 296 species that had a minimum of 148 100 observations within the tropics (23.50 N – 23.50 S) during the defined time period. 149 150 2.1.2. Plant observations from herbarium records 151 We retrieved data from the Global Biodiversity Information Facility (GBIF) to compare 152 phenological estimates derived from observations on iNaturalist to those derived from tropical 153 plant specimens deposited into herbaria worldwide (Fig. 1, left panel). We queried GBIF for all 154 plant specimen records from tropical latitudes collected between 2010-2023, with the additional 155 constraints that the record was not marked as having a geospatial issue and had collection 156 coordinates with uncertainty of less than 100m. To expand the geographic and taxonomic scope 157 of this herbarium-derived data, we also retrieved tropical herbarium records from Tropicos, a 158 curated botanical database maintained by the Missouri Botanical Garden; this dataset had not 159 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 9 been retrieved with our previous query as it has unspecified coordinate uncertainties. Together, 160 these queries yielded 111533 observations from 3853 species with a minimum of 10 records 161 collected globally within 2010-2023. Subsequent analyses assume that the presence of a 162 herbarium record indicates that the plant in question was reproductive (i.e., flowering or fruiting) 163 at the time of collection, based on common collection practices that promote collecting plants 164 with reproductive features present (Goëau et al., 2020; Heberling et al., 2019; Willis et al., 2017). 165 166 2.1.3 Urbanization 167 To assign urbanization status for each observation, we used the 2020 version of the Global 168 Human Settlement-Degree of Urbanization (GHS-SMOD) layer, developed by the European 169 Commission’s Joint Research Centre (Fig. 1, left panel) (Florczyk et al., 2019). This layer offers 170 consistent global data on human settlement intensity at a 1 km resolution, based on population 171 density and built-up areas. Each grid cell is classified into one of eight urbanization categories, 172 ranging from dense city centers to largely uninhabited, very low-density zones, allowing a 173 continuous representation of settlement intensity. We simplified these categories into two 174 categories: urban (Grid Codes 21, 22, 23, 30: representing peri-urban to urban cores) and rural 175 (Grid Codes 11–13: covering very low-density rural to rural clusters), and assigned an 176 urbanization class to each plant observation (Melchiorri et al., 2018, 2019; Santillan & Heipke, 177 2023). 178 179 2.2. Data analysis 180 181 2.2.1 Comparison between phenology estimation 182 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 10 To determine if species-level reproductive seasonality from citizen-science data aligns with 183 herbarium-based inferences, we compared observations from iNaturalist and herbaria. For each 184 species in each dataset (iNaturalist and herbarium), day-of-year (DOY) values were converted to 185 radians and summarized using circular statistics in the circular package in R (Lund et al., 2025). 186 We estimated the mean vector length (r), a concentration parameter from 0 to 1 that measures the 187 strength of seasonality (Fig. 1, middle panel) (Morellato et al., 2009) and used this to classify the 188 strength of seasonality for each species (Weak (r < 0.35), Moderate (0.35 ≤ r < 0.70), and Strong 189 (r ≥ 0.70)). We selected this r -based classification because r directly measures effect size, 190 indicating how closely observations cluster around the mean flowering date, regardless of sample 191 size (Alsammani et al., 2023; Staggemeier et al., 2020; Willig et al., 2024). Moreover, when 192 comparing the same species across two large, uneven datasets (GBIF vs. iNaturalist), p-values 193 from circular significance tests can be misleading, as the Rayleigh p-value is highly sample-size-194 dependent (Alsammani et al., 2023; Willig et al., 2024). We then used a confusion matrix to 195 compare the agreement in flowering strength between two datasets. 196 197 2.2.2 Using iNaturalist observations to compare rural vs. urban phenology 198 199 Exclusion of multimodal species 200 201 In the tropics, some species can express multimodal reproductive phenology, with multiple 202 flowering peaks per year (Wright et al., 2019), which complicate comparisons of phenology 203 across urban and rural contexts. To account for this, we first used the Hermans-Rasson (HR) test, 204 a robust, nonparametric method for detecting nonuniformity that does not depend on a specific 205 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 11 distribution (Hermans & Rasson, 1985; Landler et al., 2019) and excluded species that express 206 multimodal reproductive phenology from subsequent analyses. First, we used Rayleigh’s test to 207 determine whether observations are significantly clustered around a mean direction (Morellato et 208 al., 2009). Next, we used the Rayleigh test to evaluate the strength of seasonality (aseasonal, 209 when Rayleigh p > 0.05, or seasonal, Rayleigh p < 0.05). Both seasonal and aseasonal species 210 were included in the analysis because the seasonality-strength measure r is relevant for unimodal 211 distributions and interpretable as “weak seasonality” in uniform cases. However, species with 212 multimodal reproductive phenology were excluded, as these violate the assumptions behind the 213 circular statistics used in our analysis (Datta et al., 2025; Morellato et al., 2009; Staggemeier et 214 al., 2020). 215 Modeling reproductive seasonality strength and duration 216 Reproductive seasonality strength (r) was estimated from circular statistics (Fig. 1, right panel). 217 We estimated total reproductive duration by calculating the percentiles along the circular 218 ordering of flowering and fruiting angles and taking the angular distance between the 5th and 219 95th percentiles, and finally converting this angular distance back into days of year. 220 221 Next, to test whether seasonality and duration differ between urban and rural 222 environments, we fitted the following linear mixed-effects models: 223 224 225 226 Here, phenology ijk was either a species i’s estimated $r$ or its estimated reproductive duration 227 in grid cell j and in urban class k. β₁ captures the fixed effect of urbanization as rural and urban 228 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 12 class, and species and grid were treated as random intercepts to account for taxonomic and 229 spatial non-independence. 230 231 Finally, as the circular statistics in our modeling approach can be sensitive to the number 232 of observations being used to generate the statistic (e.g. longer flowering durations or reduced 233 seasonality strength could be estimated simply due to more extensive sampling), we additionally 234 conducted a supplemental analysis in which we subsampled the data to have equal representation 235 within each species-grid-urban class category (Fig. S5).. Briefly, we “rarefied” the data 1000 236 times to ensure equal representation in each urban class within a species-grid cell combination, 237 estimated phenology using the circular statistics above, and fitted the global linear mixed effect 238 model. As the results from this approach were consistent with the results of our “global” model 239 that included all observations (Fig. S5), we focus our main text on results from the global model. 240 241 Modeling with species’ traits 242 To assess how different plant growth and reproductive strategies impact phenological responses 243 to urbanization, we also modeled: (1) seasonality class, including seasonal and aseasonal species, 244 (2) life duration categories, such as annual and perennial species, and (3) life forms, namely 245 herbs, shrubs, and vines separately for smaller subsets of the data. Trait-specific information for 246 all the species was compiled through manual online searches. We only consulted the 247 authoritative botanical sources like the USDA and botanical gardens' websites. We fitted 248 models with interactive effects of urbanization class and plant traits (seasonality, life duration, or 249 life forms). All the models were fitted using lmerTest with Satterthwaite-adjusted t-tests, 250 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 13 weighting observations by sample size per grid (Kuznetsova et al., 2017). All analyses were 251 performed using R version 4.5.1 (R Core Team, 2025). 252 253 254 3. RESULTS 255 3.1 Data overview 256 We had 217,991 unique observations with flowers or fruits or both from iNaturalist and 111,533 257 from herbaria, spanning 2010-2023. Our iNaturalist and herbaria datasets had 296 and 3,835 258 unique species, respectively. They had 229 species in common. 259 Our iNaturalist dataset (Fig. 2, top panel) showed broad tropical coverage across 260 continents, with dense clusters of observations in South and Southeast Asia, East Africa, and 261 Central America, with roughly equal distributions between rural and urban regions (Fig 2a). In 262 contrast, the herbarium data obtained from GBIF (Fig. 2, bottom panel) were spatially sparser, 263 concentrated mainly in Latin America, Northern Australia, West Africa, and parts of Southeast 264 Asia. Very few herbarium records (~3.3%) were collected in urban locations (Fig. 2b), consistent 265 with the traditional botanical survey bias toward less disturbed ecosystems. However, herbarium 266 records spanned 3,835 species in contrast to the 229 species represented in iNaturalist, indicating 267 that iNaturalist offers broader spatial coverage with comparable number of observations in urban 268 and rural areas, while herbaria provide taxonomically rich, rural-biased observations. 269 Similarly, we found iNaturalist observations peaking in April–May (late dry to early wet 270 season in many tropical regions), reflecting heightened flowering visibility and user activity. 271 Similarly, we observed secondary peaks in October–November, suggesting a bimodal 272 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 14 phenological rhythm consistent with intertropical seasonal transitions. Both urban and rural 273 classes show similar seasonal timing, but urban peaks were slightly higher in April, May, and 274 November, reflecting more observer activity and maybe prolonged reproductive activity in cities 275 due to heat islands (Supplementary file, Fig. 1, top panel). Whereas the herbaria dataset with a 276 consistent rural-dominated signal showed a peak collection timing between March and June, and 277 moderate activity throughout the year. The absence of strong urban representation mainly 278 reflects sampling bias rather than biological inactivity (Supplementary file, Fig. 1, bottom panel). 279 280 3.2 Phenological insights from iNaturalist and herbarium-derived 281 Classifications of reproductive seasonality were broadly consistent across the herbarium-derived 282 and iNaturalist datasets. Of the 229 species in this comparison, 118 were identified as weakly 283 seasonal with both iNaturalist and herbarium data, and only 7 were identified as strongly 284 seasonal with both datasets (Table 1, top panel). The most common discrepancy occurred due to 285 species being assigned as having “moderate” seasonality through herbarium data and only 286 “weak” seasonality when phenology was assessed with iNaturalist records. Values of the 287 seasonality metric (r) from herbarium and iNaturalist records were significantly correlated with 288 one another (Spearman’s ρ = 0.31, p = 1.31 × 10⁻⁵). 289 290 3.3 Comparison of reproductive phase duration and reproductive seasonality strength 291 3.3.1. Global comparison 292 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 15 Across 3104 grid × species × urban class combinations, representing 285 unique grids and 238 293 species, reproductive duration ranged from 3 to 364 days. On average across species and grid 294 cells, plant reproductive periods were significantly longer in urban than in rural contexts (Fig. 295 4A; rural flowering period = 195.04 ± 5.06; urban flowering period = 206.83 ± 5.06; p = 3.26e-296 05, Table S1). The strength of seasonality (r) was correspondingly lower in urban than in rural 297 settings (Fig. 4B; rural r = 0.61 ± 0.014; urban r = 0.59 ± 0.014; p = 3.1e-04, Table S2). 298 3.3.2. Comparison across seasonal and aseasonal species 299 For species with seasonal reproduction (i.e., those with Raleigh p < 0.05, Datta et al., 2025), 300 reproductive periods were three weeks longer in urban than in rural settings (rural reproductive 301 duration = 177.57 days ± 5.56; urban reproductive duration = 198.78 ± 5.35; p = 5.12 e-15; Fig. 302 5A and Table S1). Species with aseasonal reproduction had on average longer reproductive 303 periods; these species also experienced extended reproductive periods in urban than rural 304 settings, although this effect was weaker than for seasonal species (rural reproductive duration = 305 213.65 days ± 5.42; urban reproductive duration = 222.57 ± 5.35; p = .03; Fig. 5B and Table S1). 306 Similar results were also detected when measuring seasonality based on r rather than flowering 307 duration (Table S2 and Fig. S2). 308 3.3.3. Comparison across annual and perennial species 309 For annual species, reproductive periods were nearly a month longer in urban than in rural 310 localities (rural reproductive duration = 171.29 days ± 8.4; urban reproductive duration = 201.12 311 ± 8.15; p = 1.4 × 10^ (-8); Fig. 6A and Table S1). In contrast, reproductive periods were about a 312 week longer in urban than in rural settings for perennial species (rural reproductive duration = 313 205.33 days ± 5.5; urban reproductive duration = 213.5 ± 5.46; p = 1.4 × 10^-8; Fig. 6A and 314 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 16 Table S1). Similarly, annual plant species had considerably lower seasonality (r) in urban than in 315 rural settings, but urbanization was associated with only a modest reduction in seasonality for 316 perennial species (Fig. S3 and Table S2). 317 3.3.4. Comparison across herbs, shrubs, and vines 318 Reproductive periods were significantly longer in urban than in rural localities for herbaceous 319 plants and for vines (Herbaceous: rural reproductive period = 187.51 ± 6.61, urban reproductive 320 period = 211.9 ± 6.47, p = 5.07e-10; Vines: rural reproductive period = 196.79±10.52, urban 321 reproductive period = 212.25 ± 10.23, p = 3.49e-4, see Figs. 7A and 7C). However, no such 322 effect of urbanization on reproductive period was found in shrubs (rural reproductive period = 323 209.43±6.89, urban reproductive period = 207.85 ± 6.76, Fig. 7B, n.s.)). Similar patterns were 324 detected with the strength of seasonality r, which was lower in urban than in rural localities for 325 herbs and vines but not for shrubs (Fig. S4 and Table S2). 326 327 4. DISCUSSION 328 329 Urbanization is among the most important drivers of environmental change across the tropics, 330 yet its influence on the timing of plant reproductive phenology remains poorly studied at broad 331 spatial and taxonomic scales. By integrating multi-continental citizen science observations with 332 curated herbarium records, our study reveals two central findings. First, despite fundamental 333 differences in sampling design, citizen science and herbarium datasets converge on consistent 334 estimates of species-level reproductive seasonality strength at local grid scales, supporting the 335 use of citizen science-based observations for phenology-related studies in the tropics. Second, 336 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 17 across thousands of species-grid combinations, urban plant populations exhibit longer 337 reproductive duration and reduced seasonality strength relative to rural populations, although the 338 magnitude and direction of this effect vary across species and functional groups. Together, these 339

demonstrate that urbanization leaves a detectable and biologically meaningful imprint on 340 tropical reproductive phenology, while also highlighting the value of combining opportunistic 341 and curated data streams to understand phenological responses to global change in the tropics. 342 4.1. Using citizen science data for studying tropical urban phenology 343 A vital contribution of our study lies in demonstrating the efficiency of citizen science 344 observations in answering phenological questions that would be difficult to address using 345 herbarium data alone. Although herbarium collections are taxonomically rich and provide vital 346 phenological insights (Karthikeyan et al., 2025; Ordoñez et al., 2025), these observations tend to 347 be from less developed (or rural) locations (Fig. 2b), which limits their utility for studying urban 348 ecology. In contrast, iNaturalist observations are distributed over dense urban centers, peri-urban 349 mosaics, and rural inland habitats, facilitating urban phenology studies. As a result, the use of 350 iNaturalist data in peer-reviewed studies has increased rapidly, with the data coming from 128 351 countries representing 638 taxonomic families (Mason et al., 2025). This extensive coverage by 352 iNaturalist illustrates the expanding reach of citizen science, even in biodiversity-rich but 353 traditionally under-sampled tropical regions (Barve et al., 2020; Iwanycki Ahlstrand et al., 2022). 354 At the same time, the dataset from herbaria, though biased toward rural observations, 355 offer taxonomically rich, high-quality records and remain equally important in phenology and 356 climate change studies. This result (3835 unique species from herbarium records compared to 357 296 from iNaturalist) aligns with previous demonstrations that herbarium records capture high 358 taxonomic and functional diversity (Eckert et al., 2024), which could make them reliable data 359 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 18 sources for studying phenological signatures of diverse plant taxa (Iwanycki Ahlstrand et al., 360 2022; Willis et al., 2017). Our comparison of reproductive seasonality between data sources 361 further reveals that many species classified as moderately and strongly seasonal based on 362 herbaria records are referred as weakly seasonal in iNaturalist observations. This discrepancy in 363 seasonality classification exists likely due to differences in sampling strategies. Herbarium 364 specimens are often collected during peak phenophases via directed short field campaigns, 365 whereas citizen science platforms like iNaturalist observations are results of opportunistic, 366 repeated encounters across seasons. These contrasts between the herbaria and citizen science 367 datasets affirm their complimentary strength, as combining these datasets bridges spatial and 368 ecological gaps in tropical monitoring by integrating real-time, human-centered observations 369 with long-term scientific records (Ramirez-Parada et al., 2024; Williamson et al., 2025). It also 370 supports previous claims regarding the potential value of dedicated citizen science efforts in 371 tropical areas, and particularly in tropical cities, to fill crucial data gaps in support of long term 372 research and monitoring (e.g., SeasonWatch, Ramaswami et al., 2021). 373 374 4.2. Urbanization impacts on plant reproductive phenology 375 Analyzing nearly 3100 combinations of grid cells, species, and urban classes, we 376 observed that urban populations had longer reproductive duration and lower seasonality strength 377 than rural observations. Our observed patterns across the tropics mirror findings from temperate 378 cities, where urban heat islands, increased atmospheric CO2, regulated irrigation, and altered 379 light regimes prolong resource availability. These signatures of urbanization enable plants to 380 flower earlier, flower longer, or produce multiple flowering cycles within a year (Fujiwara et al., 381 2025; Neil & Wu, 2006; D. S. Park et al., 2023; Sexton et al., 2023; Wohlfahrt et al., 2019). The 382 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 19 similarity in phenological signal between temperate and tropic responses to urbanization may 383 seem surprising, given the widely documented complexity and diversity of tropical phenology 384 strategy. Here, the convergence between tropical and temperate urban phenology responses may 385 reflect common urban mechanisms overriding otherwise distinct climate drivers. Indeed, our 386 detection of a consistent signal of urbanization underscores its potentially powerful role as a 387 driver of phenological shifts among diverse taxa. 388 Despite the clear urbanization effect in our global model (Table S1, S2), we also saw 389 pronounced interspecific variation in both the direction and magnitude of phenological 390 responses. Trait-level analyses (Figs.5-7) reveal varied phenological responses between life 391 cycles and plant functional types. Seasonal and annual species, particularly with herbs and vines, 392 exhibited stronger urban-associated increases in reproductive duration and reduction in 393 seasonality strength. Longer reproductive periods for annual plants in urban settings might be 394 due to their being able to exploit prolonged favorable conditions (Fujiwara et al., 2025), in 395 contrast to perennial species, which are more likely to retain conservative phenological schedules 396 (Marcacci et al., 2023; Sexton et al., 2023; Stanley & Ashman, 2025). Although relatively few 397 studies have examined variation in phenological responses across functional groups (herbs, 398 shrubs, vines) to urbanization, recent results from a review paper suggest that the composition of 399 plant species and their functional group can lead to diverse phenology due to differences in 400 growth characteristics and adaptability (Sun et al., 2026). As temperature is found to be a 401 primary driver of flowering (Sun et al., 2026; Zhou et al., 2023), and temperature tends to vary 402 between urban-rural settings (Ogunbode et al., 2025; Ribeiro et al., 2024), these variation in 403 temperature and distinctiveness of plant species across functional group could explain variation 404 in phenological responses across functional groups between urban-rural environments. 405 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 20 Together, the results suggest a clear effect of urbanization on tropical reproductive 406 phenology through an extended reproducing duration in urban settings. Prolonged reproductive 407 phases could increase overlap among co-flowering species, modify plant-pollinator interactions, 408 alter mutualistic networks, and intensify interspecific pollen transfer, with consequences for 409 reproductive success and competitive dynamics (Dzul-Cauich & Munguía-Rosas, 2025; 410 Marcacci et al., 2023; Sexton et al., 2023). At the broader scale, lengthened reproductive seasons 411 may have ecosystem-level consequences by altering carbon uptake, surface energy balance, and 412 feedbacks between vegetation and urban microclimates (Williamson et al., 2025; Wohlfahrt et 413 al., 2019). In rapidly urbanizing tropical regions, where biodiversity is high and ecological 414 interactions are tightly linked to phenological timing, such shifts could propagate through 415 mutualistic networks and ecosystem processes, meriting further study. 416 417 4.3. Limitations 418 Our work provides important insights into the implications of urbanization for plant phenology 419 across the tropics, but a few important caveats should shape the interpretation of our results. Our 420 analyses are implemented with urbanization classifications at a 1x1 km resolution within 25x25 421 km grid cells, assuming limited impacts of fine-scale environmental heterogeneity on the 422 phenology within urban-rural landscapes. Our urban classification likewise simplifies a 423 multidimensional gradient that includes variation in land use, management intensity, and 424 microclimate. On-the-ground studies of phenology and demography across urban-rural gradients 425 can help disentangle these multidimensional signatures of urbanization. Additionally, both 426 citizen-science and herbarium datasets remain influenced by observer behavior and collection 427 practices, particularly in urban settings where observations are more frequent and temporally 428 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 21 biased. Although we control for potential biases towards higher density of sampling in urban 429 settings (Fig. S5), it remains possible that more extensive temporal coverage of data contribute to 430 a broader estimate of flowering duration or weaker seasonality strength in urban environments. 431 Finally, our comparison of phenology estimates from iNaturalist, and herbarium specimens 432 assumed that all herbaria specimens were reproductive at the time of collection. Although this 433 assumption is consistent with most collection practices (Goëau et al., 2020; Heberling & Isaac, 434 2017), it may have led to our overestimating reproductive periods. Making herbarium specimens 435 more valuable for phenological research will require continued advancements that draw on 436 human and machine-based annotations of herbaria observations (Grady et al., 2025). 437 438 4.4. Conclusions 439 This study represents one of the first large-scale applications of iNaturalist to investigate plant 440 phenology across urban contexts in tropical ecosystems. By coupling these records with 441 herbarium observations, we demonstrate that opportunistic, crowd-sourced imagery can yield 442 ecologically consistent and complementary insights into reproductive periodicity. The approach 443 establishes a scalable framework for monitoring phenological resilience to urbanization and 444 climate change, particularly in data-poor tropical regions where systematic observations are 445 limited. Our findings suggest that abiotic features associated with urbanization result in extended 446 reproductive periods than in nearby rural localities, with potential downstream impacts on 447 population, community, and ecosystem dynamics. As digital archives and machine-learning-448 assisted phenophase annotation advance, integrating citizen-science, herbarium, and remote-449 sensing data will enable dynamic, high-resolution phenological mapping across space and time. 450 Such efforts can transform tropical ecology from a historically data-scarce discipline into one 451 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 22 driven by continuous, community-powered observation networks, improving forecasts of 452 biodiversity responses to global change. 453 454 5. Author Contributions 455 Rohit Raj Jha, Anita Simha, Richard Ekeng Ita, Rachana Rao, and Gaurav Kandlikar conceived 456 the idea. All authors contributed to data acquisition. Data curation was done by Rohit Raj Jha 457 and Gaurav Kandlikar. Rohit Raj Jha completed the formal analysis with Gaurav Kandlikar. 458 Rohit Raj Jha wrote the paper with inputs from Gaurav Kandlikar, Daijiang Li, and Anita Simha, 459 and all authors reviewed the paper. 460 461

462 We thank the contributors to iNaturalist and Herbarium observations and the authors of various 463 open-source packages that make this computational work possible. We acknowledge the 464 Louisiana State University startup funds to Gaurav Kandlikar and the National Science 465 Foundation grant DBI2223508 to Daijiang Li. We thank Karthik Thrikkaderri for his insights 466 during data acquisitions and comments on the manuscript. 467 468 Ethics statement 469 None 470 Conflicts of Interest 471 The authors declare no conflicts of interest. 472 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 23 473 Data Availability Statement 474 The data and codes associated with this manuscript are archived in Zenodo 475 (https://doi.org/10.5281/zenodo.18252047). 476 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 24

477 Abernethy, K., Bush, E. R., Forget, P., Mendoza, I., & Morellato, L. P. C. (2018). Current issues 478 in tropical phenology: A synthesis. Biotropica, 50(3), 477–482. 479 Alsammani, A., Stacey, W. C., & Gliske, S. V. (2023). Estimation of circular statistics in the 480 presence of measurement bias. IEEE Journal of Biomedical and Health Informatics, 481 28(2), 1089–1100. 482 Barve, V. V., Brenskelle, L., Li, D., Stucky, B. J., Barve, N. V., Hantak, M. M., McLean, B. S., 483 Paluh, D. J., Oswald, J. A., Belitz, M. W., Folk, R. A., & Guralnick, R. P. (2020). Methods 484 for broad-scale plant phenology assessments using citizen scientists’ photographs. 485 Applications in Plant Sciences, 8(1), e11315. https://doi.org/10.1002/aps3.11315 486 Bonebrake, T. C., Tsang, T. P. N., Yu, N., Wang, Y., Ledger, M. J., Tilley, H. B., Yau, E. Y. H., 487 Andersson, A. A., Boyle, M. J. W., Lee, K. W. K., Li, Q., Ling, Y. F., Dongmo, M. A. K., 488 Güçlü, C., Dingle, C., & Ashton, L. A. (2025). Tropical cities as windows into the 489 ecosystems of our present and future. Biotropica, 57(1), e13369. 490 https://doi.org/10.1111/btp.13369 491 Borchert, R. (1996). Phenology and flowering periodicity of Neotropical dry forest species: 492 Evidence from herbarium collections. Journal of Tropical Ecology, 12(1), 65–80. 493 Borchert, R., Rivera, G., & Hagnauer, W. (2002). Modification of Vegetative Phenology in a 494 Tropical Semi-deciduous Forest by Abnormal Drought and Rain. Biotropica, 34(1), 27–495 39. https://doi.org/10.1111/j.1744-7429.2002.tb00239.x 496 Bush, E. R., Abernethy, K. A., Jeffery, K., Tutin, C., White, L., Dimoto, E., Dikangadissi, J., 497 Jump, A. S., & Bunnefeld, N. (2017). Fourier analysis to detect phenological cycles using 498 long-term tropical field data and simulations. Methods in Ecology and Evolution, 8(5), 499 530–540. 500 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 25 Bush, E. R., Bunnefeld, N., Dimoto, E., Dikangadissi, J., Jeffery, K., Tutin, C., White, L., & 501 Abernethy, K. A. (2018). Towards effective monitoring of tropical phenology: Maximizing 502 returns and reducing uncertainty in long-term studies. Biotropica, 50(3), 455–464. 503 Callaghan, C. T., Ozeroff, I., Hitchcock, C., & Chandler, M. (2020). Capitalizing on opportunistic 504 citizen science data to monitor urban biodiversity: A multi-taxa framework. Biological 505 Conservation, 251, 108753. https://doi.org/10.1016/j.biocon.2020.108753 506 Caparros-Santiago, J. A., Rodriguez-Galiano, V., & Dash, J. (2021). Land surface phenology as 507 indicator of global terrestrial ecosystem dynamics: A systematic review. ISPRS Journal 508 of Photogrammetry and Remote Sensing, 171, 330–347. 509 Capinha, C., Ceia-Hasse, A., de-Miguel, S., Vila-Viçosa, C., Porto, M., Jarić, I., Tiago, P., 510 Fernández, N., Valdez, J., & McCallum, I. (2024). Using citizen science data for 511 predicting the timing of ecological phenomena across regions. BioScience, 74(6), 383–512 392. 513 Cleland, E. E., Chuine, I., Menzel, A., Mooney, H. A., & Schwartz, M. D. (2007). Shifting plant 514 phenology in response to global change. Trends in Ecology & Evolution, 22(7), 357–365. 515 Datta, A., Banerjee, S., Naniwadekar, R., Thapa, K., Rathore, A., Thapa, K., Brah, T., Nabum, 516 T., Mogar, N., Shukla, U., Sidhu, S., & Borawake, N. (2025). Patterns of Leaf, Flower, 517 and Fruit Phenology and Environmental Relationships in a Seasonal Tropical Forest in 518 the Indian Eastern Himalaya. Biotropica, 57(3), e70030. 519 https://doi.org/10.1111/btp.70030 520 Davis, C. C., Willis, C. G., Connolly, B., Kelly, C., & Ellison, A. M. (2015). Herbarium records are 521 reliable sources of phenological change driven by climate and provide novel insights into 522 species’ phenological cueing mechanisms. American Journal of Botany, 102(10), 1599–523 1609. 524 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 26 Di Cecco, G. J., Barve, V., Belitz, M. W., Stucky, B. J., Guralnick, R. P., & Hurlbert, A. H. (2021). 525 Observing the observers: How participants contribute data to iNaturalist and implications 526 for biodiversity science. BioScience, 71(11), 1179–1188. 527 Dinnage, R., Grady, E., Neal, N., Deck, J., Denny, E., Walls, R., Seltzer, C., Guralnick, R., & Li, 528 D. (2025). PhenoVision: A framework for automating and delivering research-ready plant 529 phenology data from field images. Methods in Ecology and Evolution. 530 Dzul-Cauich, H. F., & Munguía-Rosas, M. A. (2025). Effects of Urbanization on Flowering 531 Phenology, Pollination, and Reproductive Success in the Chiropterophilous Tropical 532 Tree Ceiba pentandra. Plants, 14(11), 1575. 533 Eckert, I., Bruneau, A., Metsger, D. A., Joly, S., Dickinson, T., & Pollock, L. J. (2024). Herbarium 534 collections remain essential in the age of community science. Nature Communications, 535 15(1), 7586. 536 Fisher, J. I., Mustard, J. F., & Vadeboncoeur, M. A. (2006). Green leaf phenology at Landsat 537 resolution: Scaling from the field to the satellite. Remote Sensing of Environment, 538 100(2), 265–279. https://doi.org/10.1016/j.rse.2005.10.022 539 Fitchett, J. M., Grab, S. W., & Thompson, D. I. (2015). Plant phenology and climate change: 540 Progress in methodological approaches and application. Progress in Physical 541 Geography, 39(4), 460–482. 542 Florczyk, A. J., Corbane, C., Ehrlich, D., Freire, S., Kemper, T., Maffenini, L., Melchiorri, M., 543 Pesaresi, M., Politis, P., & Schiavina, M. (2019). GHSL data package 2019. 544 Luxembourg, Eur, 29788(10.2760), 290498. 545 Forrest, J., & Miller-Rushing, A. J. (2010). Toward a synthetic understanding of the role of 546 phenology in ecology and evolution. Philosophical Transactions of the Royal Society B: 547 Biological Sciences, 365(1555), 3101–3112. 548 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 27 Fujiwara, H., Yamaguchi, H., Nakata, K., & Katsuhara, K. R. (2025). Urbanised landscape and 549 microhabitat differences can influence flowering phenology and synchrony in an annual 550 herb. Journal of Applied Ecology, 62(11), 3115–3127. 551 Gallinat, A. S., Ellwood, E. R., Heberling, J. M., Miller-Rushing, A. J., Pearse, W. D., & Primack, 552 R. B. (2021). Macrophenology: Insights into the broad-scale patterns, drivers, and 553 consequences of phenology. American Journal of Botany, 108(11), 2112–2126. 554 Goëau, H., Mora-Fallas, A., Champ, J., Love, N. L. R., Mazer, S. J., Mata-Montero, E., Joly, A., 555 & Bonnet, P. (2020). A new fine-grained method for automated visual analysis of 556 herbarium specimens: A case study for phenological data extraction. Applications in 557 Plant Sciences, 8(6), e11368. 558 Grady, E. L., LaFrance, R., Li, D., Dinnage, R., Denny, E. G., Deck, J., & Guralnick, R. P. 559 (2025). Petal to the metal: The slow road to automating large-scale phenology labeling 560 for herbarium specimens. bioRxiv, 2025–12. 561 Heberling, J. M., & Isaac, B. L. (2017). Herbarium specimens as exaptations. American Journal 562 of Botany, 104(7), 963–965. 563 Heberling, J. M., McDonough MacKenzie, C., Fridley, J. D., Kalisz, S., & Primack, R. B. (2019). 564 Phenological mismatch with trees reduces wildflower carbon budgets. Ecology Letters, 565 22(4), 616–623. https://doi.org/10.1111/ele.13224 566 Hermans, M., & Rasson, J. (1985). A new Sobolev test for uniformity on the circle. Biometrika, 567 72(3), 698–702. 568 Iwanycki Ahlstrand, N., Primack, R. B., & Tøttrup, A. P. (2022). A comparison of herbarium and 569 citizen science phenology datasets for detecting response of flowering time to climate 570 change in Denmark. International Journal of Biometeorology, 66(5), 849–862. 571 Jochner, S., Alves-Eigenheer, M., Menzel, A., & Morellato, L. P. C. (2013). Using phenology to 572 assess urban heat islands in tropical and temperate regions. International Journal of 573 Climatology, 33(15). 574 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 28 Johansson, J., Nilsson, J.-Å., & Jonzén, N. (2015). Phenological change and ecological 575 interactions: An introduction. Oikos, 124(1), 1–3. 576 Kabano, P., Harris, A., & Lindley, S. (2021). Sensitivity of canopy phenology to local urban 577 environmental characteristics in a tropical city. Ecosystems, 24(5), 1110–1124. 578 Karthikeyan, A., Karthik, V., & Chandrasekaran, S. (2025). Flowering out of sync: Climate 579 change alters the reproductive phenology of Terminalia paniculata in the Western Ghats 580 of India. Plants, People, Planet. 581 Kharouba, H. M., Ehrlén, J., Gelman, A., Bolmgren, K., Allen, J. M., Travers, S. E., & Wolkovich, 582 E. M. (2018). Global shifts in the phenological synchrony of species interactions over 583 recent decades. Proceedings of the National Academy of Sciences, 115(20), 5211–584 5216. 585 Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. (2017). lmerTest package: Tests in linear 586 mixed effects models. Journal of Statistical Software, 82, 1–26. 587 Lai, H. R. (2025). Model-based ordination for phenological studies: From controlling sampling 588 bias to inferring temporal associations. Methods in Ecology and Evolution. 589 Landler, L., Ruxton, G. D., & Malkemper, E. P. (2019). The Hermans–Rasson test as a powerful 590 alternative to the Rayleigh test for circular statistics in biology. BMC Ecology, 19(1), 30. 591 Li, D., Stucky, B. J., Deck, J., Baiser, B., & Guralnick, R. P. (2019). The effect of urbanization on 592 plant phenology depends on regional temperature. Nature Ecology & Evolution, 3(12), 593 1661–1667. 594 Lund, U., Agostinelli, C., Arai, H., Gagliardi, A., García-Portugués, E., Giunchi, D., Irisson, J.-O., 595 Pocernich, M., & Rotolo, F. (2025). circular: Circular Statistics (Version 0.5-2) [Computer 596 software]. https://cran.r-project.org/web/packages/circular/index.html 597 Marcacci, G., Westphal, C., Rao, V. S., Kumar S, S., Tharini, K., Belavadi, V. V., Nölke, N., 598 Tscharntke, T., & Grass, I. (2023). Urbanization alters the spatiotemporal dynamics of 599 plant–pollinator networks in a tropical megacity. Ecology Letters, 26(11), 1951–1962. 600 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 29 Mason, B. M., Mesaglio, T., Barratt Heitmann, J., Chandler, M., Chowdhury, S., Gorta, S. B. Z., 601 Grattarola, F., Groom, Q., Hitchcock, C., & Hoskins, L. (2025). iNaturalist accelerates 602 biodiversity research. BioScience, 75(11), 953–965. 603 McDonald, R. I., Mansur, A. V., Ascensão, F., Colbert, M., Crossman, K., Elmqvist, T., 604 Gonzalez, A., Güneralp, B., Haase, D., & Hamann, M. (2020). Research gaps in 605 knowledge of the impact of urban growth on biodiversity. Nature Sustainability, 3(1), 16–606 24. 607 Melchiorri, M., Florczyk, A. J., Freire, S., Schiavina, M., Pesaresi, M., & Kemper, T. (2018). 608 Unveiling 25 years of planetary urbanization with remote sensing: Perspectives from the 609 global human settlement layer. Remote Sensing, 10(5), 768. 610 Melchiorri, M., Pesaresi, M., Florczyk, A. J., Corbane, C., & Kemper, T. (2019). Principles and 611 applications of the global human settlement layer as baseline for the land use efficiency 612 indicator—SDG 11.3. 1. ISPRS International Journal of Geo-Information, 8(2), 96. 613 Menzel, A., Sparks, T. H., Estrella, N., Koch, E., Aasa, A., Ahas, R., Alm-Kübler, K., Bissolli, P., 614 Braslavská, O., Briede, A., Chmielewski, F. M., Crepinsek, Z., Curnel, Y., Dahl, Å., 615 Defila, C., Donnelly, A., Filella, Y., Jatczak, K., Måge, F., … Zust, A. (2006). European 616 phenological response to climate change matches the warming pattern. Global Change 617 Biology, 12(10), 1969–1976. https://doi.org/10.1111/j.1365-2486.2006.01193.x 618 Moore, C. E., Brown, T., Keenan, T. F., Duursma, R. A., Van Dijk, A. I., Beringer, J., Culvenor, 619 D., Evans, B., Huete, A., & Hutley, L. B. (2016). Reviews and syntheses: Australian 620 vegetation phenology: New insights from satellite remote sensing and digital repeat 621 photography. Biogeosciences, 13(17), 5085–5102. 622 Morellato, L. P. C., Alberti, L., & Hudson, I. L. (2009). Applications of circular statistics in plant 623 phenology: A case studies approach. In Phenological research: Methods for 624 environmental and climate change analysis (pp. 339–359). Springer. 625 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 30 Neil, K., & Wu, J. (2006). Effects of urbanization on plant flowering phenology: A review. Urban 626 Ecosystems, 9(3), 243–257. 627 Numata, S., Yamaguchi, K., Shimizu, M., Sakurai, G., Morimoto, A., Alias, N., Noor Azman, N. 628 Z., Hosaka, T., & Satake, A. (2022). Impacts of climate change on reproductive 629 phenology in tropical rainforests of Southeast Asia. Communications Biology, 5(1), 311. 630 Ogunbode, T. O., Oyebamiji, V. O., Sanni, D. O., Akinwale, E. O., & Akinluyi, F. O. (2025). 631 Environmental impacts of urban growth and land use changes in tropical cities. Frontiers 632 in Sustainable Cities, 6, 1481932. 633 Ordoñez, J., Tovar, C., Walker, B., Wheeler, J., Ayala-Ruano, S., Aguirre-Carvajal, K., 634 McMahon, S., & Cuesta, F. (2025). Phenological patterns of tropical mountain forest 635 trees across the neotropics: Evidence from herbarium specimens. Proceedings B, 636 292(2041), 20242748. 637 Pabon-Moreno, D. E., Musavi, T., Migliavacca, M., Reichstein, M., Römermann, C., & Mahecha, 638 M. D. (2019). Ecosystem physio-phenology revealed using circular statistics. 639 Biogeosciences Discussions, 2019, 1–29. 640 Park, D. S., Lyra, G. M., Ellison, A. M., Maruyama, R. K. B., dos Reis Torquato, D., Asprino, R. 641 C., Cook, B. I., & Davis, C. C. (2023). Herbarium records provide reliable phenology 642 estimates in the understudied tropics. Journal of Ecology, 111(2), 327–337. 643 https://doi.org/10.1111/1365-2745.14047 644 Park, J. Y., Muller-Landau, H. C., Lichstein, J. W., Rifai, S. W., Dandois, J. P., & Bohlman, S. A. 645 (2019). Quantifying leaf phenology of individual trees and species in a tropical forest 646 using unmanned aerial vehicle (UAV) images. Remote Sensing, 11(13), 1534. 647 Parmesan, C., & Yohe, G. (2003). A globally coherent fingerprint of climate change impacts 648 across natural systems. Nature, 421(6918), 37–42. 649 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 31 Piao, S., Liu, Q., Chen, A., Janssens, I. A., Fu, Y., Dai, J., Liu, L., Lian, X., Shen, M., & Zhu, X. 650 (2019). Plant phenology and global climate change: Current progresses and challenges. 651 Global Change Biology, 25(6), 1922–1940. 652 R Core Team. (2025). _R: A Language and Environment for Statistical Computing_ [Computer 653 software]. R Foundation for Statistical Computing, Vienna, Austria. 655 Ramaswami, G., Sidhu, S., & Quader, S. (2021). Using citizen science to build baseline data on 656 tropical tree phenology. Current Science, 121(11), 1409–1416. 657 Ramirez-Parada, T. H., Park, I. W., Record, S., Davis, C. C., Ellison, A. M., & Mazer, S. J. 658 (2024). Plasticity and not adaptation is the primary source of temperature-mediated 659 variation in flowering phenology in North America. Nature Ecology & Evolution, 8(3), 660 467–476. https://doi.org/10.1038/s41559-023-02304-5 661 Ribeiro, M. P., Menezes, G. P., Figueiredo, G. K., de Mello, K., & Valente, R. A. (2024). Impacts 662 of urban landscape pattern changes on land surface temperature in Southeast Brazil. 663 Remote Sensing Applications: Society and Environment, 33, 101142. 664 Sakai, S. (2001). Phenological diversity in tropical forests. Population Ecology, 43(1), 77–86. 665 Santillan, J., & Heipke, C. (2023). Using GHSL to Analyze Urbanization and Land-Use 666 Efficiency in the Philippines from 1975-2020: Trends and Implications for Sustainable 667 Development. 668 Sexton, A. N., Benton, S., & Emery, S. M. (2023). Urbanization and plant diversity influence 669 different aspects of floral phenology. Urban Ecosystems, 26(2), 517–524. 670 Simkin, R. D., Seto, K. C., McDonald, R. I., & Jetz, W. (2022). Biodiversity impacts and 671 conservation implications of urban land expansion projected to 2050. Proceedings of the 672 National Academy of Sciences, 119(12), e2117297119. 673 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 32 Singh, K., & Kushwaha, C. (2005). Paradox of leaf phenology: Shorea robusta is a semi-674 evergreen species in tropical dry deciduous forests in India. Current Science, 1820–675 1824. 676 Staggemeier, V. G., Camargo, M. G. G., Diniz-Filho, J. A. F., Freckleton, R., Jardim, L., & 677 Morellato, L. P. C. (2020). The circular nature of recurrent life cycle events: A test 678 comparing tropical and temperate phenology. Journal of Ecology, 108(2), 393–404. 679 Stanley, A. M., & Ashman, T. (2025). Urbanization Alters Phenology, Mating System Allocation, 680 and Life History of Impatiens capensis (Balsaminaceae) via Trait-Specific Plasticity and 681 Genetic Differentiation. Ecology and Evolution, 15(6), e71583. 682 Stuble, K. L., Bennion, L. D., & Kuebbing, S. E. (2021). Plant phenological responses to 683 experimental warming—A synthesis. Global Change Biology, 27(17), 4110–4124. 684 Sun, J., Lv, W., Wang, S., Iler, A. M., Meng, F., Li, B., Zhou, Y., Lv, J., Yuan, F., & Luo, C. 685 (2026). Functional group and aridity regulate impacts of climate change on plant 686 phenology: A meta-analysis. Nature Communications. 687 Tang, J., Körner, C., Muraoka, H., Piao, S., Shen, M., Thackeray, S. J., & Yang, X. (2016). 688 Emerging opportunities and challenges in phenology: A review. Ecosphere, 7(8), 689 e01436. 690 Williamson, D. R., Prestø, T., Westergaard, K. B., Trascau, B. M., Vange, V., Hassel, K., Koch, 691 W., & Speed, J. D. (2025). Long-term trends in global flowering phenology. New 692 Phytologist. 693 Willig, M. R., Rojas-Sandoval, J., & Presley, S. J. (2024). Phenological patterns in ecology: 694 Problems using circular statistics and solutions based on simulations. Methods in 695 Ecology and Evolution, 15(5), 868–885. 696 Willis, C. G., Ellwood, E. R., Primack, R. B., Davis, C. C., Pearson, K. D., Gallinat, A. S., Yost, J. 697 M., Nelson, G., Mazer, S. J., & Rossington, N. L. (2017). Old plants, new tricks: 698 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 33 Phenological research using herbarium specimens. Trends in Ecology & Evolution, 699 32(7), 531–546. 700 Wohlfahrt, G., Tomelleri, E., & Hammerle, A. (2019). The urban imprint on plant phenology. 701 Nature Ecology & Evolution, 3(12), 1668–1674. 702 Wolf, S., Mahecha, M. D., Sabatini, F. M., Wirth, C., Bruelheide, H., Kattge, J., Moreno 703 Martínez, Á., Mora, K., & Kattenborn, T. (2022). Citizen science plant observations 704 encode global trait patterns. Nature Ecology & Evolution, 6(12), 1850–1859. 705 https://doi.org/10.1038/s41559-022-01904-x 706 Wright, S. J., Calderón, O., & Muller-Landau, H. C. (2019). A phenology model for tropical 707 species that flower multiple times each year. Ecological Research, 34(1), 20–29. 708 https://doi.org/10.1111/1440-1703.1017 709 Zhang, X., Friedl, M. A., & Schaaf, C. B. (2006). Global vegetation phenology from Moderate 710 Resolution Imaging Spectroradiometer (MODIS): Evaluation of global patterns and 711 comparison with in situ measurements. Journal of Geophysical Research: 712 Biogeosciences, 111(G4). 713 Zhou, H., Min, X., Chen, J., Lu, C., Huang, Y., Zhang, Z., & Liu, H. (2023). Climate warming 714 interacts with other global change drivers to influence plant phenology: A meta-analysis 715 of experimental studies. Ecology Letters, 26(8), 1370–1381. 716 Zohner, C. M., Benito, B. M., Fridley, J. D., Svenning, J., & Renner, S. S. (2017). Spring 717 predictability explains different leaf-out strategies in the woody floras of North America, 718 Europe and East Asia. Ecology Letters, 20(4), 452–460. 719 720 721 722 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 34 723 724 725 726 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 35 727 Figure 1. Overview of data source, phenological metrics, and analytical framework used to 728 assess urban-rural differences in tropical plant reproductive phenology 729 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 36 730 731 732 Figure 2. Spatial distribution of observations included from iNaturalist (top row) and herbaria 733 (bottom row) across tropical regions. Colored points represent urban (red) and rural (blue) 734 observations. iNaturalist data exhibit broad spatial coverage with strong urban clustering, 735 whereas herbaria data show a rural-dominated sampling pattern. Inset bars represent total counts 736 of rural and urban observations and are scaled consistently across both figure panels. 737 738 739 740 741 742 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 37 743 Figure 3. Circular flowering phenology for Cyanthillium cinereum (Bottom right) developed 744 using iNaturalist observations from 2010 to 2023, covering a single grid (Bottom left). Rural 745 iNaturalist records show a more concentrated, seasonal flowering peak, whereas urban 746 observations display a broader, less seasonal distribution. r indicates the corresponding values for 747 reproductive seasonality strength, and n is the number of observations in each urban class. 748 Specimen credit: Lava Chen, 2026 749 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 38 750 Figure 4. Urban–rural differences in reproductive duration (A) and seasonality strength (B) 751 across all species x grids x urban class combinations. Each colored point represents one species 752 in a grid cell, in either rural (blue) or urban (red) localities; thin grey lines connect urban and 753 rural estimates. The white points and lines represent model-estimated marginal means for rural 754 and urban localities. This result of longer reproductive periods in urban than in rural conditions 755 was supported in a supplemental analysis that controlled for potential effects of uneven urban vs. 756 rural sampling intensities within grids. 757 758 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 39 759 Figure 5. Paired urban–rural reproductive durations for aseasonal and seasonal species found in 760 both rural and urban grid cells. A: Species categorized as aseasonal (with no peak) showed a 761 longer flowering duration in urban settings. B: Species with seasonality (single-peak) in 762 flowering also displayed a similar pattern with longer duration of flowering in urban regions. 763 Thin gray lines denote raw paired data; black points and lines represent model-estimated 764 marginal means. 765 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 40 766 Figure 6. Paired urban–rural durations for annual and perennial species found in both rural and 767 urban grid cells. A: Annual species displaying a significant longer reproductive phase in urban 768 regions. B: Perennial species showing similar length of reproductive phase for rural and urban 769 observations. Thin gray lines denote raw paired data; black points and lines represent model-770 estimated marginal means. 771 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 41 772 Figure 7. Paired urban–rural reproductive durations for herbs, shrubs, and vines found in both 773 rural and urban grid cells. A: Herbs displaying a significant longer reproductive phase in urban 774 regions. B: Shrubs showing no difference in reproductive duration between rural and urban 775 observations. C: Vines with a significant difference in reproductive duration between rural and 776 urban observations. Thin gray lines denote raw paired data; black points and lines represent 777 model-estimated marginal means. 778 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint 42 Table 1. Confusion matrix comparing reproductive seasonality classifications based on data from 779 iNaturalist (rows) and herbaria (columns). The top half shows species-level comparisons, 780 including only taxa present in both datasets. The bottom half shows grid-level data based on 781 shared species × grid combinations across iNaturalist and herbaria. The diagonal cells' value, 782 denoted in bold shows agreement between datasets. 783 784 iNat / Herbaria Weak Moderate Strong Species-level seasonality classification Weak 118 53 3 Moderate 17 24 2 Strong 0 3 6 785 786 787 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 30, 2026. ; https://doi.org/10.64898/2026.01.28.702306doi: bioRxiv preprint