Abstract With the introduction and adaptation of non‐native mosquitoes to cities, vector‐borne diseases are emerging concerns in the United States of America. A growing body of literature suggests that mosquito infestations and associated diseases are a greater burden in socioeconomically disadvantaged areas within the heterogeneous urban ecosystem, but the ecological processes behind such spatial variation are not well understood. Urban plant communities that provide detrital resources to larval mosquitoes also vary across fine scales, with most field studies showing less abundant and diverse vegetation in lower income neighborhoods. We propose that socioeconomically aligned variation in plant communities is an important mechanism behind socioeconomically aligned mosquito infestations. However, few studies evaluating mosquito–plant interactions link experimental treatments to fine‐scale variation in urban plant community composition. Related knowledge on the interconnections between socioeconomics, mosquitoes, and vegetation is scattered among disciplines, from sociology to vector biology to plant ecology. We synthesize knowledge from diverse disciplinary sources to better understand urban mosquito ecology. We evaluate and summarize results from studies investigating socioeconomic influences on urban mosquitoes and residential vegetation in the continental United States. We then review studies examining the vegetation influences on larval mosquitoes. Synthesis across these previously disparate components of urban mosquito socioecology emphasizes knowledge gaps, including areas for future research to better integrate these topics. When considered together, the studies highlighted here contribute to a more comprehensive understanding of urban mosquito ecology than any one discipline could alone.

Juvenile salmonid mortality due to infectious haematopoietic necrosis virus (IHNV) can be a major burden on fish hatcheries. We consider possible IHNV transmission routes and specialist-generalist patterns across three regions in the Pacific Northwest of North America: Coastal Washington and Oregon (CWO), Lower Columbia River Basin (LCRB) and Snake River Basin (SRB) to obtain multi-region inference about IHNV transmission and specialization. While individual regions have been studied previously, our consideration of three regions to identify consistent trends or localized patterns is novel. The most consistent patterns we found were that most exposure to IHNV was from migrating adult salmonids and that the IHNV lineage MD specialized in steelhead and rainbow trout. Our results were consistent with LCRB being a source of IHNV and the SRB and CWO being sinks. Results specific to particular regions include the role of local geography on exposure and influence of hatchery complexes on transmission, which highlights the need to understand local hatchery operations before disease ecology can be understood and suitable hatchery management can be planned. Results demonstrate the need for spatially and temporally explicit quantitative approaches to understand disease dynamics and inform management.This article is part of the theme issue 'Managing infectious marine diseases in wild populations'.

Pulsed resources, including mast production by forest trees, often have knock-on effects on consumer populations and their prey, predators, parasites and mutualists. Response by small rodents to fluctuating acorn production in temperate forests is a widespread example. Long-term research in Maine, USA, recently suggested combined effects of a warming climate and forest maturation on acorn production by red oak trees, leading to directional increases in average population density and body mass of white-footed mice. To foster reproducibility in long-term ecological research, we analysed data from our long-term study in southeastern New York, USA, which used similar field methods. Such a comparison allowed us to assess impacts of climate warming and forest growth on the same pulsed resource and responses by the same consumer species over time and at different latitudes. Despite a clear directional increase in mean minimum temperature and considerable growth in the average size and total basal area of trees during our 33-year study, neither acorn production by red oak trees nor abundance of white-footed mice showed directional increases. Similarly, average body mass of the mice did not change through time. Abundance of mice in mid-summer increased with increasing red oak acorn production the prior autumn. Mouse abundance also was higher in warmer years, although the effect of acorn abundance on mice was stronger. We found no evidence that temperature modified the acorn-driven population responses by mice. Long-term studies are notoriously hard to maintain and even harder to replicate between sites. The direct comparison of similar studies between Maine and New York provides an opportunity to assess the generality of mechanistic models linking climate change, mast seeding and consumer responses.

Risk of human exposure to Lyme disease and other tick-borne zoonoses is affected by interactions between ticks, pathogens, vertebrate hosts for ticks and pathogens, and abiotic conditions. Inherent complexities in these ecological systems present challenges to the development of a predictive understanding useful for reducing exposure risk. Here we evaluate a conceptual model of the Lyme disease system using long-term monitoring (>30 y) at sites within an endemic region of New York State, United States. Contrary to conventional wisdom, we found that the abundance of nymphal black-legged ticks—a key risk factor for humans—was not correlated with the abundance of white-tailed deer, the prior abundance of larval ticks, or seasonal temperature extremes. Instead, the abundance of nymphal ticks was correlated with prior abundance of white-footed mice, which in turn was driven by production of acorns by red oak trees. Infection prevalence with zoonotic pathogens at our relatively intact forest sites was not associated with the abundance of white-footed mice, a species known to transmit infection with high efficiency, but instead appeared to arise from the distribution of larval ticks across the entire community of larval hosts. High densities of mice strongly reduced the tick burdens on individual mice and chipmunks, suggesting intraguild competition for parasites. Lyme disease risk arises from the dynamics of the ecological community, rather than from the dynamics of any single population of hosts or from extreme seasonal weather. These dynamics are complex but predictable, such that understanding them can facilitate improvements in management and education aimed at protecting public health.

Mosquito larvae feed on microbes growing on decomposing organic matter, which in cities is often limited to a finite amount of allochthonous leaf litter in artificial container habitats. Curiously, field surveys in US cities consistently document a mismatch between blocks with the greatest mosquito infestations and blocks where their plant resources are most plentiful, with variation shaped by socioeconomic factors. We gathered leaves characteristic of socioeconomically diverse blocks in Baltimore, Maryland, and Washington, DC, to create lab mesocosms for two common mosquito species, Aedes albopictus (Skuse) and Culex pipiens (Linnaeus). We reared mosquitoes at varying densities to assess the effects of leaf litter composition on survival, development time, and body size, from which we calculated population performance and effects on competition. We also investigated traits of the leaf litter and aquatic environment. Leaf litter containing at least one non-native species, including the mix common on low-income blocks (nonnative trees Ailanthus altissima and Paulownia tomentosa) and the mix common across the region (native tree Juglans nigra and nonnative tree Morus alba) improved mosquito outcomes compared to the native leaf litter common on high-income blocks (native trees Acer rubrum and Ulmus americana) and also alleviated competitive effects on development time, perhaps making coexistence more likely. The three litter types showed significant differences in C:N ratio, decay rate, and tannin/lignin concentration but not in microbial abundance. These results offer a mechanistic explanation for previously observed socioeconomic patterns in urban mosquito populations and point toward novel strategies for mosquito control through vegetation management.

Plants on residential urban properties can provide valuable ecosystem services or produce harmful disservices depending on fine-scale characteristics tied to plant species identity, such as growth habit or native status. The composition of plant identities on a given block is often influenced by socioeconomic factors, leading to variable green space function across a city. We surveyed residential plants on Baltimore, MD, and Washington, D.C. blocks, documenting differences in structure, richness, and community composition along an income gradient and between abandoned and neighboring occupied properties. Both canopy and ground vegetation on low-income residential properties covered less area and were more likely to contain non-native species than their higher-income counterparts, with different tree, vine, and non-native communities present. Abandoned properties had more canopy cover and higher tree richness than occupied neighbors but similar community composition including five common vines, four of which were non-native. These differences have important implications for ecosystem services, and such fine-scale knowledge could better inform the management of green space to benefit urban residents.

Specific host-tick interactions in temperate forest systems influence variation in density and infection prevalence of nymphal blacklegged tick (Ixodes scapularis). The density of infected nymphs (DIN), which is the product of nymphal infection prevalence (NIP) and density of questing nymphs (DON), influences the risk of human exposure to tick-borne pathogens. Questing nymphs (>2000) collected between 2014 and 2022 were screened for 16 zoonotic pathogens. More than a third (38.8%) of nymphs were infected with at least one pathogen, and Borrelia burgdorferi (the agent of Lyme disease) was detected in all six sampling locations and years. Pathogens included Babesia microti (NIP = 21.4%), Bo. burgdorferi (19.3%), Anaplasma phagocytophilum (5.8%), Bo. miyamotoi (1.5%), Powassan virus (<0.01%), and two regionally emergent Rickettsia (10 nymphs sampled in 2016 and 2021). Rates of Ba. microti infection were high relative to prior work, and coinfection with Bo. burgdorferi increased during the study period. White-footed mouse (Peromyscus leucopus) and eastern chipmunk (Tamias striatus) are important zoonotic hosts in temperate forests. We evaluated how variation in host abundance and the number of larval ticks feeding per animal (larval burden) explained and predicted DIN, as well as DON and NIP independently. While both mouse abundance and mouse larval burden were positive predictors of DON in this study, their influence on NIP and DIN for commonly detected pathogens differed. Both mouse and chipmunk larval burden explained significant variation in Bo. burgdorferi DIN, and larval burden on mice specifically improved prediction when observed Bo. burgdorferi DIN was high. Chipmunk larval burden also predicted variance in Bo. burgdorferi NIP and also explained significant variation in Ba. microti DIN. While observed host-tick contact metrics improved predictive skill, underprediction was most evident when observed DIN or NIP was high. These results emphasize how tick-borne zoonoses depend on a distribution of larval meals across a community of hosts. For Bo. burgdorferi in particular, NIP may be stabilized by tick-feeding on a diverse host community. Thus, while large changes in mouse abundance may predict regional changes in DIN, local NIP in particular may be more responsive to shifts in larval feeding activity across the entire community of hosts.

Where human settlements abut or intermix with wildlands, people may encounter animals that host zoonotic pathogens which can spillover to cause human disease. Known as the wildland-urban interface (WUI), this zone occupies around 5% of the Earth's surface and is home to 3.5 billion people. The rapid spread of SARS-CoV-2 has demonstrated the importance of understanding risk factors for disease among an increasingly urbanized population. However, the contribution of the WUI to zoonotic disease risk is poorly understood. Here, we show that low-level host richness occurs throughout most of the global WUI, and 20% of the human WUI population live in zones of particularly high zoonotic potential, where more than 20 host species could occur. Zones of high zoonotic potential are concentrated in low-middle-income countries (LMICs) across equatorial Africa, Brazil, Central America, and Southeast Asia where vulnerability is further elevated by widespread poverty, inadequate housing, and lack of easily accessible healthcare. Three of four people living in WUIs with high host richness (520 million people) are in LMICs. Of this population, 35% (183 million) live in and around cities in West, East, and South Africa. This means that WUI-based populations of LMICs may face the double threat of high zoonotic potential and vulnerability to disease. Our results identify global priorities for monitoring exposure to zoonotic diseases in the rapidly expanding WUI.

Incorporating vegetation into urban landscapes (hereafter, greening) has numerous ecological and social benefits. Not all greening processes are intentional, though, and not all nature conveys benefits to urban residents under conditions of uneven urban development, racial segregation, and unrecognized care work. We describe a framework for integrating multiple lines of evidence to explore the social contexts and socioecological impacts of urban greening. We assemble data (e.g., human surveys, Photovoice, spatial mapping, and ecological protocols) from neighborhoods in Baltimore City, Maryland, that were historically redlined and racially segregated but have subsequently experienced divergent paths of population and wealth accumulation, or continued official marginalization. Incorporating vegetation into cities offers both risk and benefit to local residents, and we demonstrate that the source of both initiative and resources matters. Greening initiatives that did not have local buy-in became local burdens. Although greening "vacant" properties in neighborhoods with population decline might convey city-scale benefits, local residents associated the greening with loss of valued human community, and they were unlikely to use or maintain such imposed or incidental green spaces. Local benefits, including heat amelioration, were not evident in our analysis. Discontent with greening was further associated with low expectations for help with other nature-based disamenities. In cities, reporting disamenities, such as mosquito nuisance, is often the trigger for directing resources toward management. In our study, residents with the greatest exposure to disamenities were least likely to initiate the processes that trigger external management.

AbstractClimate‐induced shifts in mosquito phenology and population structure have important implications for the health of humans and wildlife. The timing and intensity of mosquito interactions with infected and susceptible hosts are a primary determinant of vector‐borne disease dynamics. Like most ectotherms, rates of mosquito development and corresponding phenological patterns are expected to change under shifting climates. However, developing accurate forecasts of mosquito phenology under climate change that can be used to inform management programs remains challenging despite an abundance of available data. As climate change will have variable effects on mosquito demography and phenology across species it is vital that we identify associated traits that may explain the observed variation. Here, we review a suite of modeling approaches that could be applied to generate forecasts of mosquito activity under climate change and evaluate the strengths and weaknesses of the different approaches. We describe four primary life history and physiological traits that can be used to constrain models and demonstrate how this prior information can be harnessed to develop a more general understanding of how mosquito activity will shift under changing climates. Combining a trait‐based approach with appropriate modeling techniques can allow for the development of actionable, flexible, and multi‐scale forecasts of mosquito population dynamics and phenology for diverse stakeholders.

AbstractLyme disease, the most prevalent tick‐borne disease in North America, is caused by the bacterium Borrelia burgdorferi, and in the eastern and central United States, it is spread to humans by the black‐legged tick (Ixodes scapularis). Due to the complex, multiyear and multihost life cycle of this species, a matrix modeling approach is needed to effectively estimate subseasonal, multistage survival and transition dynamics in order to better understand and predict when population growth is high. Of the three questing tick life stages (larvae, nymphs, and adults), nymphs are most often associated with transmitting the bacteria to humans, and previous work suggests a mix of abiotic and biotic drivers are associated with nymph abundance. However, understanding tick population growth requires understanding mortality and transition probabilities for each stage and each stage may be individually and uniquely impacted by climate and host availability. A larval tick, for example, may experience warming temperatures differently than nymph or adults, because they are present on the landscape at different times. Here, we describe and validate a model that accounts for field sampling design and evaluates abiotic (temperature, relative humidity, precipitation) and biotic (host abundance) drivers of variation in tick population growth. To account for the drivers of subseasonal and interannual variability in demography, phenology, and population density, we built stage‐structured population models that account for variability in meteorology and host population abundance throughout the full tick lifecycle. Our model is fit and validated with 11 years of tick and host data from the northeastern United States. In this context, we found that a four‐stage model that includes unique transitions to and from a dormant, overwintering nymph state outperforms a model that only includes the three questing stages, and that incorporating the abundance of the predominant host species, Peromyscus leucopus, and weather variables improved predictions and model fit. Additionally, the model accurately predicted all three questing stages at sites different than where they were calibrated, showing that this model structure is generally transferable. Overall, this model lays a foundation for the real‐time iterative forecasting of tick populations needed to effectively protect public health.

The expansion of the tiger mosquito, a vector that can transmit diseases such as dengue, chikungunya, and Zika virus, poses a growing threat to global health. This study focuses on the entomological surveillance of Kastellorizo, a remote Greek island affected by its expansion. This research employs a multifaceted approach, combining KAP survey (knowledge, attitude, practices), mosquito collection using adult traps and human landing catches, and morphological and molecular identification methods. Results from questionnaires reveal community awareness and preparedness gaps, emphasizing the need for targeted education. Mosquito collections confirm the presence of the Aedes albopictus, Aedes cretinus, and Culex pipiens mosquitoes, highlighting the importance of surveillance. This study underscores the significance of community engagement in entomological efforts and proposes a citizen science initiative for sustained monitoring. Overall, this research provides essential insights for developing effective mosquito control programs in remote island settings, thereby emphasizing the importance of adopting a One Health approach to mitigate the spread of vector-borne diseases.

Tree canopy cover is a critical component of the urban environment that supports ecosystem services at multiple spatial and temporal scales. Increasing tree canopy across a matrix of public and private land is challenging. As such, municipalities often plant trees along streets in public rights-of-way where there are fewer barriers to establishment, and composition and biomass of street trees are inextricably linked to human decisions, management, and care. In this study, we investigated the contributions of street trees to the broader urban forest, inclusive of tree canopy distributed across both public and private parcels in Baltimore, MD, USA. We assess how species composition, biodiversity, and biomass of street trees specifically augment the urban forest at local and citywide scales. Furthermore, we evaluate how street tree contributions to the urban forest vary with social and demographic characteristics of local residential communities. Our analyses demonstrate that street trees significantly enhanced citywide metrics of the urban forests' richness and tree biomass, adding an average six unique species per site. However, street tree contributions did not ameliorate low tree canopy locations, and more street tree biomass was generally aligned with higher urban forest cover. Furthermore, species richness, abundance, and biomass added by street trees were all positively related to local household income and population density. Our results corroborate previous findings that wealthier urban neighborhoods often have greater tree abundance and canopy cover and, additionally, suggest that investment in municipally managed street trees may be reinforcing inequities in distribution and function of the urban forest. This suggests a need for greater attention to where and why street tree plantings occur, what species are selected, and how planted tree survival is maintained by and for residents in different neighborhoods.

Much remains unknown about variation in pathogen transmission across the geographic range of a free-ranging fish or animal species and about the influence of movement (associated with husbandry practices or animal behavior) on pathogen transmission. Salmonid hatcheries are an ideal system in which to study these processes. Salmonid hatcheries are managed for endangered species recovery, supplementation of threatened or at-risk fish stocks, support of fisheries, and ecosystem stability. Infectious hematopoietic necrosis virus (IHNV) is a rhabdovirus of significant concern to salmon aquaculture. Landscape IHNV transmission dynamics previously had been estimated only for salmonid hatcheries in the Lower Columbia River Basin (LCRB). The objectives of this study were to estimate IHNV transmission dynamics in a unique geographic region, the Snake River Basin (SRB), and to quantitatively estimate the effect of model coproduction on inference because previous assessments of coproduction have been qualitative. In contrast to the LCRB, the SRB has hatchery complexes consisting of a main hatchery and ≥1 satellite facility. Knowledge about hatchery complexes was held by a subset of project researchers but would not have been available to project modelers without coproduction. Project modelers generated and tested multiple versions of Bayesian susceptible-exposedinfected models to realistically represent the SRB and estimate the effect of coproduction. Models estimated the frequency of transmission routes, route-specific infection probabilities, and infection probabilities for combinations of salmonid hosts and IHNV lineages. Model results indicated that in the SRB, avoiding exposure to IHNV-positive adult salmonids is the most important action to prevent juvenile infections. Migrating adult salmonids exposed juvenile cohort-sites most frequently, and the infection probability was greatest following exposure to migrating adults. Without coproduction, the frequency of exposure by migrating adults would have been overestimated by 70 cohort-sites, and the infection probability following exposure to migrating adults would have been underestimated by∼0.09. The coproduced model had less uncertainty in the infection probability if no transmission route could be identified (Bayesian credible interval (BCI) width = 0.12) compared to the model without coproduction (BCI width = 0.34). Evidence for virus lineage MD specialization on steelhead and rainbow trout (both Oncorhynchus mykiss) was apparent without model coproduction. In the SRB, we found a greater probability of virus lineage UC infection in Chinook salmon (Oncorhynchus tshawytscha) compared to in O. mykiss, whereas in the LCRB, UC more clearly exhibited a generalist approach. Coproduction influenced estimates that depended on transmission routes, which operated differently at main hatcheries and satellite sites within hatchery complexes. Hatchery complexes are found outside of the SRB and are not specific to salmonid hatcheries alone. There is great potential for coproduction and modeling spatial contact networks to advance understanding about infectious disease transmission in complex production systems and surrounding free-ranging animal populations.

Research in both ecology and AI strives for predictive understanding of complex systems, where nonlinearities arise from multidimensional interactions and feedbacks across multiple scales. After a century of independent, asynchronous advances in computational and ecological research, we foresee a critical need for intentional synergy to meet current societal challenges against the backdrop of global change. These challenges include understanding the unpredictability of systems-level phenomena and resilience dynamics on a rapidly changing planet. Here, we spotlight both the promise and the urgency of a convergence research paradigm between ecology and AI. Ecological systems are a challenge to fully and holistically model, even using the most prominent AI technique today: deep neural networks. Moreover, ecological systems have emergent and resilient behaviors that may inspire new, robust AI architectures and methodologies. We share examples of how challenges in ecological systems modeling would benefit from advances in AI techniques that are themselves inspired by the systems they seek to model. Both fields have inspired each other, albeit indirectly, in an evolution toward this convergence. We emphasize the need for more purposeful synergy to accelerate the understanding of ecological resilience whilst building the resilience currently lacking in modern AI systems, which have been shown to fail at times because of poor generalization in different contexts. Persistent epistemic barriers would benefit from attention in both disciplines. The implications of a successful convergence go beyond advancing ecological disciplines or achieving an artificial general intelligence-they are critical for both persisting and thriving in an uncertain future.

Predicting how the range dynamics of migratory species will respond to climate change requires a mechanistic understanding of the factors that operate across the annual cycle to control the distribution and abundance of a species. Here, we use multiple lines of evidence to reveal that environmental conditions during the nonbreeding season influence range dynamics across the life cycle of a migratory songbird, the American redstart ( Setophaga ruticilla ). Using long-term data from the nonbreeding grounds and breeding origins estimated from stable hydrogen isotopes in tail feathers, we found that the relationship between annual survival and migration distance is mediated by precipitation, but only during dry years. A long-term drying trend throughout the Caribbean is associated with higher mortality for individuals from the northern portion of the species’ breeding range, resulting in an approximate 500 km southward shift in breeding origins of this Jamaican population over the past 30 y. This shift in connectivity is mirrored by changes in the redstart’s breeding distribution and abundance. These results demonstrate that the climatic effects on demographic processes originating during the tropical nonbreeding season are actively shaping range dynamics in a migratory bird.

The relationship between (a) the structure and composition of the landscape around an individual's home and (b) environmental perceptions and health outcomes has been well demonstrated (eg the value of vegetation cover to well-being). Few studies, however, have examined how multiple landscape features (eg vegetation and water cover) relate to perceptions of multiple environmental problems (eg air or water quality) and whether those relationships hold over time. We utilized a long-term dataset of geolocated telephone surveys in Baltimore, Maryland, to identify relationships between residents' perceptions of environmental problems and nearby landcover. Residents of neighborhoods with more vegetation or located closer to water were less likely to perceive environmental problems. Water quality was one exception to this trend, in that people were more likely to perceive water-quality problems when nearby water cover was greater. These trends endured over time, suggesting that these relationships are stable and therefore useful for informing policy aimed at minimizing perceived environmental problems.

Climatic conditions are widely thought to govern the distribution and abundance of ectoparasites, such as the blacklegged tick (Ixodes scapularis), vector of the agents of Lyme disease and other emerging human pathogens. However, translating physiological tolerances to distributional limits or mortality is challenging. Ticks may be able to avoid or tolerate unsuitable conditions, and what is lethal to one life history stage may not extend to others. Thus, even after decades of research, there are clear gaps in our knowledge about how climatic conditions determine tick distributions or patterns of abundance. We present results from a 3-year study combining daily hazard models and data from field experiments at three sites spanning much of I. scapularis' current latitudinal distribution. We examine three predominant hypotheses regarding how temperature and vapor pressure deficits affect stage-specific survival and transition success and consider how these results influence population growth and distribution. We found that larvae are sensitive to temperature and vapor pressure deficits, whereas mortality of nymphs and adults is consistent with depletion of energy reserves. Consistent with prior work, we found that overwinter survival was high and successful stage transitions (e.g., fed nymphs molting to adults) were sensitive to temperature. Collectively, results from this comprehensive, multiyear, multistage field study suggest that population growth of I. scapularis is less limited by restrictive climatic conditions than has been broadly assumed, although influences on larval survival may slow tick population growth and establishment in some desiccating conditions. Further studies should integrate climate effects on stage-specific survival to better understand these effects on population dynamics and range expansion in a changing climate.

Near-term ecological forecasts provide resource managers advance notice of changes in ecosystem services, such as fisheries stocks, timber yields, or water quality. Importantly, ecological forecasts can identify where there is uncertainty in the forecasting system, which is necessary to improve forecast skill and guide interpretation of forecast results. Uncertainty partitioning identifies the relative contributions to total forecast variance introduced by different sources, including specification of the model structure, errors in driver data, and estimation of current states (initial conditions). Uncertainty partitioning could be particularly useful in improving forecasts of highly variable cyanobacterial densities, which are difficult to predict and present a persistent challenge for lake managers. As cyanobacteria can produce toxic and unsightly surface scums, advance warning when cyanobacterial densities are increasing could help managers mitigate water quality issues. Here, we fit 13 Bayesian state-space models to evaluate different hypotheses about cyanobacterial densities in a low nutrient lake that experiences sporadic surface scums of the toxin-producing cyanobacterium, Gloeotrichia echinulata. We used data from several summers of weekly cyanobacteria samples to identify dominant sources of uncertainty for near-term (1- to 4-week) forecasts of G. echinulata densities. Water temperature was an important predictor of cyanobacterial densities during model fitting and at the 4-week forecast horizon. However, no physical covariates improved model performance over a simple model including the previous week's densities in 1-week-ahead forecasts. Even the best fit models exhibited large variance in forecasted cyanobacterial densities and did not capture rare peak occurrences, indicating that significant explanatory variables when fitting models to historical data are not always effective for forecasting. Uncertainty partitioning revealed that model process specification and initial conditions dominated forecast uncertainty. These findings indicate that long-term studies of different cyanobacterial life stages and movement in the water column as well as measurements of drivers relevant to different life stages could improve model process representation of cyanobacteria abundance. In addition, improved observation protocols could better define initial conditions and reduce spatial misalignment of environmental data and cyanobacteria observations. Our results emphasize the importance of ecological forecasting principles and uncertainty partitioning to refine and understand predictive capacity across ecosystems.