The capacity of forest soils to sequester carbon (C) is susceptible to shifts in the litter quantity and chemistry. Using measurements from the third decade of a long-term Detrital Input and Removal Treatment (DIRT) experiment at Harvard Forest (MA, USA), extending previously published work from 20 years, we examined how litter inputs shape soil organic matter (SOM) chemistry. Elemental analysis, targeted compound analysis, solid-state 13C nuclear magnetic resonance (NMR) spectroscopy, and microbial biomass and community composition measurements were used. Despite doubled litter inputs over 30 years, no net soil C accumulation occurred, and the SOM decomposition stage was similar to the control, suggesting continuous microbial processing of added inputs. The exclusion of litter, roots, or both led to lower soil C and more advanced SOM decomposition in mineral soils. Shifts in microbial community composition, particularly an increase in Gram (+) to Gram (-) bacteria under exclusion treatments, point to microbial reorganization in response to altered substrate availability. This long-term study underscores the limited potential for long-term soil C sequestration due to sustained microbial decomposition and the role of continuous plant inputs in shaping SOM chemisty under changing detrital regimes in a temperate forest.

Accurate, high-resolution mapping of Vegetation Optical Depth (VOD) is critical for assessing forest biomass, ecosystem health yet current satellite products lack the spatial detail needed to capture fine-scale heterogeneity. Ground-based GNSS Transmissometry (GNSS-T) offers meter-scale VOD mea surements but is limited in spatial coverage due to operational constraints. This paper presents a novel data fusion framework that combines mobile GNSS-T observations with airborne LiDAR to produce large-scale VOD maps at 5 m × 5 m resolution blocks. The method leverages a Cone-Based Point Distribution Imaging (CPDI) approach, which transforms LiDAR point clouds within GNSS-aligned conical sampling zones into 28 × 30 × 3 angu lar–height histograms, encoding vegetation structure relevant to microwave attenuation. These CPDI images are used to train a convolutional neural network (CNN) to predict block-level VOD values, which are then mapped across entire LiDAR domain over a 400 m × 400 m region. The framework was validated using data from the SMAPVEX22 campaign at Harvard Forest, MA and Millbrook, NY, achieving strong agreement between predicted and measured VOD (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.93 training, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.74 testing) and demonstrating consistent correlations with in-situ basal area (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.72) and biomass (R2 = 0.46) at the regional level. The results show that integrating GNSS-T and LiDAR enables scalable, high-resolution VOD mapping across heteroge neous forested landscapes, offering new capabilities for drought monitoring, forest management, and wildfire risk assessment.

Abstract. Dry deposition is a major loss pathway for reactive nitrogen species from the atmospheric boundary layer. We evaluate isolated components of the parameterization for species-specific gaseous dry-deposition velocity Vd(x) for HNO3 and NO2 from the GEOS-Chem chemical transport model by running a stand-alone version of Vd code in single-point mode to enable a more direct comparison to field observations. Improved measurement–model agreement results mainly from (i) updates to the calculation of molecular diffusivities and (ii) the representation of ground surface NO2 hydrolysis in the formulation of non-stomatal uptake. We evaluate the parameterization for non-stomatal dry deposition of NO2 by comparing to eddy-covariance-inferred nocturnal Vd(NO2) over Harvard Forest. We address a large low bias (−80 %) in simulated nocturnal Vd(NO2) by representing NO2 heterogeneous hydrolysis on deposition surfaces, paying attention to chemical flux divergence, soil NO emission, and canopy surface area effects. Finally, we evaluate the updated oxidized reactive nitrogen (NOy) dry-deposition parameterization by comparing to eddy-covariance-inferred Vd(NOy) over Harvard Forest, finding that a modest nocturnal low bias (−19 %) remains in simulated Vd(NOy) due to the compensating effects of updates to the calculation of molecular diffusivities (28 % reduction in nocturnal Vd(NOy)) and the representation of NO2 heterogeneous hydrolysis (25 % increase in nocturnal Vd(NOy)). These developments are a first step towards a tractable representation of NO2 hydrolysis in a dry-deposition scheme and have important implications for the near-surface NO2 lifetime through a mechanism involving HONO emission.

Multi-decadal datasets from systematic surveys of birds are rarely published, despite their potential to yield important information about local changes in the environment over time. I compare bird surveys at Harvard Forest in central Massachusetts, USA, between 1993 and 2016 with two unpublished annotated bird checklists from the area (1948 and 1970), along with data from other long-term bird survey sites in New England, and regional community science data. While a handful of breeding species at the Harvard Forest are common each year, species turnover has been constant, with roughly a third of bird species found one survey year and not the other. I list 18 species that have apparently colonized Harvard Forest as breeding species after 1948, and 16 that have declined since then, including two to the point of extirpation. I also note several examples of phylogenetic replacements in the recent vs. historical avifauna. More colonizing species than declining species are at their northern range limit rather than their southern limit in southern New England, and more migratory species appear to be declining than colonizing. More colonizing species favor forest interior than declining ones, and nearly all declining species are associated with edge/early-successional habitats or are considered habitat generalists. The majority of species to have colonized Harvard Forest since 1948 show positive statewide population trends, and none showed a loss in the number of breeding bird atlas blocks occupied between the 1970s and the 2000s. I identify three declining or extirpated species, Ruffed Grouse, Eastern Whip-poor-will, and Olive-sided Flycatcher, as particularly deserving of conservation attention. Long-term, site-scale monitoring is essential to detect the impact of local forest management techniques, which at Harvard Forest has included the creation of experimental clear-cuts, the loss of conifer plantations, and the continued maturation of the mixed hardwood forest.

Climate change, land disturbance, and atmospheric chemistry have substantially impacted northeastern hardwood forests. However, it is challenging to quantify the exacerbating or mitigating interactions among these disturbances on carbon (C), nitrogen (N), and water cycling in forest ecosystems. To evaluate these effects, we applied the PnET-CN-daily model to simulate the historical patterns of C, N, and water cycling at Harvard Forest in central Massachusetts, United States. The model was run with a reconstructed historical climate and air chemistry scenario, and results were compared with field measurements at Harvard Forest for calibration. The calibrated model was then run with a series of hypothetical scenarios to decompose the impacts of individual environmental drivers on C, N, and water cycling of the forest ecosystem. Model simulations suggest that increases in atmospheric carbon dioxide (CO2) concentrations, changes in climate, and decreases in atmospheric N deposition have contributed to historical changes in the plant C cycle. Elevated CO2 concentrations have been the dominant factor, though these effects have diminished with increasing concentrations. The combination of elevated CO2 and a warmer climate has led to increased plant growth, resulting in higher plant N storage but a decline in the soil N pool. However, elevated atmospheric N deposition has mitigated this decline in soil N and also suppressed soil decomposition. Climate has been a key driver of recent changes in the water cycle, with increased air temperatures leading to higher transpiration rates. Despite this change, soil water content at Harvard Forest remained relatively constant over the simulation period because of increasing water-use efficiency associated with increasing CO2 concentration, indicating that plant growth is not limited by water at Harvard Forest. Future investigations should use modeling approaches to project the functional responses of the forest ecosystem to the interacting effects of future climate scenarios and contrasting air quality regulations.

Abstract Accurately estimating carbon fluxes in terrestrial biosphere models (TBMs) is of great significance for Earth system science. Using satellite observations to optimize TBMs is an effective approach to achieve this goal. Radiative transfer is the physical linkage between TBMs and satellite observations. Therefore, a sophisticated radiative transfer model in TBMs, which explicitly links satellite‐observable surface reflectance to biophysical and biochemical processes within the vegetation canopy, and allows directly using remote sensing data to effectively constrain or optimize TBMs. Here, we developed the terrestrial ecosystem carbon cycle simulator (TECs) with a strategic design that incorporates an advanced radiative transfer model (RTM) based on the spectral invariant theory. This model simultaneously simulates carbon fluxes and high‐resolution spectral signals across optical to thermal wavelengths under any specified sun‐sensor geometry. We calibrated and tested TECs simulations at the Harvard Forest (HARV) National Ecological Observation Network site using a range of ecological and satellite data. After calibrating parameters, TECs accurately simulates net ecosystem exchange (NEE) (hourly: R 2 = 0.80, mean absolute error (MAE) = 1.85 μmol/m 2 /s; daily: R 2 = 0.71, MAE = 1.25 μmol/m 2 /s), hyperspectral reflectance ( R 2 : 0.85, MAE: 0.04), and land surface temperature (LST) ( R 2 : 0.85, MAE: 3.04°C). These results demonstrate that TECs is a promising tool for enhancing terrestrial carbon flux modeling using next‐generation hyperspectral observations. TECs lays a strong foundation for future integration of hyperspectral data and models to improve carbon flux predictions.

Abstract Ecosystem models offer a rigorous way to formalize scientific theories and are critical to evaluating complex interactions among ecological and biogeochemical processes. In addition to simulation and prediction, ecosystem models are a valuable tool for testing hypotheses about mechanisms and empirical findings because they reveal critical internal processes that are difficult to observe directly. However, many ecosystem models are difficult to manage and apply by scientists because of complex model structures, lack of consistent documentation, and low‐level programming implementation. Here, we present the ‘ pnetr ’ R package, which is designed to provide an easy‐to‐manage ecosystem modelling framework and detailed documentation in both model structure and programming. The framework implements a family of widely used PnET (net photosynthesis, evapotranspiration) ecosystem models, which are relatively parsimonious but capture essential biogeochemical cycles of water, carbon and nitrogen. We chose the R programming language because it is familiar to many ecologists and has abundant statistical modelling resources. We showcase examples of model simulations and test the effects of phenology on carbon assimilation and wood production using data measured by the Environmental Measurement Station (EMS) eddy‐covariance flux tower at Harvard Forest, MA. We hope ‘ pnetr ’ can facilitate further development of ecological theory and increase the accessibility of ecosystem modelling and ecological forecasting.

Diurnal and seasonal dynamics of regional CO2 drawdown at Harvard Forest: Integrating remote sensing and modeling perspectives.

We report seven highquality draft genomes of Ralstonia spp. isolated from the Harvard Forest Long-Term Warming Experimental plots: four de novo hybrid assemblies and three de novo long-read assemblies. The genomes have a minimum estimated completeness of 92.6% and an average GC content of 63.45%.

In this essay, I explore the scalar and rhetorical dimensions of care. On one hand, I argue, care can be calibrated to different scales, and caring at one scale often involves neglect or harm at another. On the other hand, I claim that decisions to care at one scale rather than another emerge in particular rhetorical ecologies within which certain things are made to matter more than others. Scale and rhetoric productively, if also frustratingly, complicate our understanding of care. While the notion of scale forces us to front the unsettling fact that acts of care at one scale often require acts of violence at another, the notion of rhetoric invites us to face the fact that decisions to care for this rather than that – at this scale rather than that one – are inevitably forged amid a rhetorical ecology that moves us to care in some ways and not others. To demonstrate the utility of these insights, I dwell at length on the rhetorical ecology in which Harvard Forest researchers arrived at and articulated their decision to ‘do nothing’ to save an imperiled native species. This article is part of the special issue ‘Re-creating Care as Mattering Practices’.

Abstract As a warmer climate enables an increase in atmospheric humidity, extreme precipitation events have become more frequent in the Northeastern United States. Understanding the impact of evolving precipitation patterns is critical to understanding water cycling in temperate forests and moisture coupling between the atmosphere and land surface. Although the role of soil moisture in evapotranspiration has been extensively studied, few have analyzed the role of soil texture in determining ecosystem‐atmosphere feedbacks. In this study, we utilized long term data associated with ecosystem water fluxes to deduce the strength of land‐atmosphere coupling at Harvard Forest, Petersham, MA, USA. We found a 1.5% increase in heavy precipitation contribution per decade where high‐intensity events compose upwards of 42% of total yearly precipitation in 2023. Intensifying precipitation trends were found in conjunction with a long‐term soil drying at the Harvard Forest despite no significant increase in evapotranspiration over 32 years. This suggests that soil water holding capacity is a key mediating variable controlling the supply of water to ecosystems and the atmosphere. We found that these land surface changes directly impacted the lifted condensation level (LCL) height over Harvard Forest which was found to be increasing at a rate of 6.62 m per year while atmospheric boundary layer (ABL) heights have fallen at a modest rate of 1.76 m per year. This has amplified dry feedbacks between the land surface and the atmosphere such that 80% of observed summers ending in a water deficit also had an anomalously low soil water content in the spring.

Soil microbial communities are pivotal to plant health and nutrient acquisition. It is becoming increasingly clear that many interactions, both among and between microbes and plants, are governed by small bioactive molecules or "secondary metabolites" that can aid in communication, competition, and nutrient uptake. Yet, secondary metabolite biogeography - who makes what, where, and why-is in its infancy. Further, secondary metabolite biosynthesis genes are often silent or weakly expressed under standard laboratory conditions, making it incredibly difficult to study these small molecules. To begin to address these dual challenges, we focused on redox-active metabolites (RAMs), a specific class of small molecules, and took advantage of recent findings that many RAMs aid in acquiring phosphorus and that their production is frequently stimulated by stress for this macronutrient. We developed a screen for RAM-producing bacteria that leverages phosphorus limitation to stimulate metabolite biosynthesis and uses a colorimetric (ferrozine) iron-reduction assay to identify redox activity. We isolated 557 root-associated bacteria from grasses collected at sites across the United States (Santa Rita Experimental Range [AZ], Konza Prairie Biological Station [KS], and Harvard Forest [MA]) and from commercial tomato plants and screened them for RAM production. We identified 128 soil isolates of at least 19 genera across Proteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes that produced RAMs under phosphorus stress. Our work reveals that the production of RAMs under phosphorus stress is common across diverse soil bacteria and provides an approach to screen for these small molecules rapidly.IMPORTANCEBy secreting secondary metabolites, bacteria at the plant root can defend against diseases and help acquire essential nutrients. However, the genes that synthesize secondary metabolites are typically inactive or are weakly expressed under standard laboratory conditions. This fact makes it difficult to study these small molecules and hinders the discovery of novel small molecules that may play crucial roles in agricultural and biomedical settings. Here, we focus on redox-active metabolites (RAMs), a class of secondary metabolites that can help bacteria solubilize phosphorus and are often produced when phosphorus is limited. We developed a screen that rapidly identifies RAM-producing bacteria by utilizing a colorimetric iron-reduction assay in combination with phosphorus limitation to stimulate biosynthesis. The screen reveals that RAM-producing bacteria are far more prevalent in soil than previously appreciated and that this approach can be used to identify RAM producers.

Forest canopy height (FCH) is a vital indicator for assessing forest health and ecosystem service capacity. Over the past two decades, full-waveform (FW) LiDAR has been widely employed for estimating forest biophysical variables due to its high precision in measuring vertical forest structures. However, the impact of terrain undulations on forest parameter estimation remains challenging. To address this issue, this study proposes a bisection approximation decomposition (BAD) method for processing GEDI L1B data and FCH estimation. The BAD method analyzes the energy composition of simplified echo signals and determines the fitting parameters by integrating overall signal energy, the differences in unresolved signals, and the similarity of inter-forest signal characteristics. FCH is subsequently estimated based on waveform peak positions. By dynamically adjusting segmentation points and Gaussian fitting parameters, the BAD method achieved precise separation of mixed canopy and ground signals, substantially enhancing the physical realism and applicability of decomposition results. The effectiveness and robustness of the BAD method for FCH estimation were evaluated using 2049 footprints across varying slope conditions in the Harvard Forest region of Petersham, Massachusetts. The results demonstrated that digital terrain models (DTMs) extracted using the GEDI data and the BAD method exhibited high consistency with the DTMs derived using airborne laser scanning (ALS) data (coefficient of determination R2 &gt; 0.99). Compared with traditional Gaussian decomposition (GD), wavelet decomposition (WD), and deconvolution decomposition (DD) methods, the BAD method showed significant advantages in FCH estimation, achieved the smallest relative root mean square error (rRMSE) of 17.19% and greatest mean estimation accuracy of 84.57%, and reduced the rRMSE by 10.74%, 21.49%, and 28.93% compared to GD, WD, and DD methods, respectively. Moreover, the BAD method exhibited a significantly stronger correlation with ALS-derived canopy height mode data than the relative height metrics from GEDI L2A products (r = 0.84, p &lt; 0.01). The robustness and adaptability of the BAD method to complex terrain conditions provide great potential for forest parameters using GEDI data.

Across biomes, soil biodiversity promotes ecosystem functions. However, whether this relationship will be maintained within ecosystems under climate change is uncertain. Here, using two long-term soil warming experiments, we investigated how warming affects the relationship between ecosystem functions and bacterial diversity across seasons, soil horizons, and warming duration. Soils were sampled from these warming experiments located at the Harvard Forest Long-Term Ecological Research (LTER) site, where soils had been heated +5°C above ambient for 13 or 28 years at the time of sampling. We assessed seven measurements representative of different ecosystem functions and nutrient pools. We also surveyed bacterial community diversity. We found that ecosystem function was significantly affected by season, with autumn samples having a higher intercept than summer samples in our model, suggesting a higher overall baseline of ecosystem function in the fall. The effect of warming on bacterial diversity was similarly affected by season, where warming in the summer was associated with decreased bacterial evenness in the organic horizon. Despite the decreased bacterial evenness in the warmed plots, we found that the relationship between ecosystem function and bacterial diversity was unaffected by warming or warming duration. Our findings highlight that season is a consistent driver of ecosystem function as well as a modulator of climate change effects on bacterial community evenness.

Across biomes, soil biodiversity promotes ecosystem functions. However, whether this relationship will be maintained within ecosystems under climate change is uncertain. Here, using two long-term soil warming experiments, we investigated how warming affects the relationship between ecosystem functions and bacterial diversity across seasons, soil horizons, and warming duration. Soils were sampled from these warming experiments located at the Harvard Forest Long-Term Ecological Research (LTER) site, where soils had been heated +5°C above ambient for 13 or 28 years at the time of sampling. We assessed seven measurements representative of different ecosystem functions and nutrient pools. We also surveyed bacterial community diversity. We found that ecosystem function was significantly affected by season, with autumn samples having a higher intercept than summer samples in our model, suggesting a higher overall baseline of ecosystem function in the fall. The effect of warming on bacterial diversity was similarly affected by season, where warming in the summer was associated with decreased bacterial evenness in the organic horizon. Despite the decreased bacterial evenness in the warmed plots, we found that the relationship between ecosystem function and bacterial diversity was unaffected by warming or warming duration. Our findings highlight that season is a consistent driver of ecosystem function as well as a modulator of climate change effects on bacterial community evenness.

Abstract Understanding tree transpiration variability is vital for assessing ecosystem water‐use efficiency and forest health amid climate change, yet most landscape‐level measurements do not differentiate individual trees. Using canopy temperature data from thermal cameras, we estimated the transpiration rates of individual trees at Harvard Forest and Niwot Ridge. PT‐JPL model was used to derive latent heat flux from thermal images at the canopy‐level, showing strong agreement with tower measurements (R2 = 0.70–0.96 at Niwot, 0.59–0.78 at Harvard at half‐hourly to monthly scales) and daily RMSE of 33.5 W/m2 (Niwot) and 52.8 W/m2 (Harvard). Tree‐level analysis revealed species‐specific responses to drought, with lodgepole pine exhibiting greater tolerance than Engelmann spruce at Niwot and red oak showing heightened resistance than red maple at Harvard. These findings show how ecophysiological differences between species result in varying responses to drought and demonstrate that these responses can be characterized by deriving transpiration from crown temperature measurements.

Climate change may alter soil microbial communities and soil organic matter (SOM) composition. Soil carbon (C) cycling takes place over multiple time scales; therefore, long-term studies are essential to better understand the factors influencing C storage and help predict responses to climate change. To investigate this further, soils that were heated by 5 °C above ambient soil temperatures for 18 years were collected from the Barre Woods Soil Warming Study at the Harvard Forest Long-term Ecological Research site. This site consists of large 30 × 30 m plots (control or heated) where entire root systems are exposed to sustained warming conditions. Measurements included soil C and nitrogen concentrations, microbial biomass, and SOM chemistry using gas chromatography–mass spectrometry and solid-state 13C nuclear magnetic resonance spectroscopy. These complementary techniques provide a holistic overview of all SOM components and a comprehensive understanding of SOM composition at the molecular-level. Our results showed that soil C concentrations were not significantly altered with warming; however, various molecular-level alterations to SOM chemistry were observed. We found evidence for both enhanced SOM decomposition and increased above-ground plant inputs with long-term warming. We also noted shifts in microbial community composition while microbial biomass remained largely unchanged. These findings suggest that prolonged warming induced increased availability of preferred substrates, leading to shifts in the microbial community and SOM biogeochemistry. The observed increase in gram-positive bacteria indicated changes in substrate availability as gram-positive bacteria are often associated with the decomposition of complex organic matter, while gram-negative bacteria preferentially break down simpler organic compounds altering SOM composition over time. Our results also highlight that additional plant inputs do not effectively offset chronic warming-induced SOM decomposition in temperate forests.

Abstract Spatial mapping of biodiversity is crucial to investigate spatial variations in natural communities. Several indices have been proposed in the literature to represent biodiversity as a single statistic. However, these indices only provide information on individual dimensions of biodiversity, thus failing to grasp its complexity comprehensively. Consequently, relying solely on these single indices can lead to misleading conclusions about the actual state of biodiversity. In this work, we focus on biodiversity profiles, which provide a more flexible framework to express biodiversity through nonnegative and convex curves, which can be analyzed by means of functional data analysis. By treating the whole curves as single entities, we propose to achieve a functional zoning of the region of interest by means of a penalized model‐based clustering procedure. This provides a spatial clustering of the biodiversity profiles, which is useful for policy‐makers both for conserving and managing natural resources and revealing patterns of interest. Our approach is evaluated using a simulation study and discussed through the analysis of the Harvard Forest Data, which provides information on the spatial distribution of woody stems within a plot of the Harvard Forest.

Imaging spectroscopy (hyperspectral sensing) is a proven tool for mapping and monitoring the spatial distribution of vegetation species composition. However, there exists a gap when it comes to the availability of high-resolution spatial and spectral imagery for accurate tree species mapping, particularly in complex forest environments, despite the continuous advancements in operational remote sensing and field sensor technologies. Here, we aim to bridge this gap by enhancing our fundamental understanding of imaging spectrometers via complex simulated environments. We used DIRSIG, a physics-based, first-principles simulation approach to model canopy-level reflectance for 3D plant models and species-level leaf reflectance in a synthetic forest scene. We simulated a realistic scene, based on the same species composition, found at Harvard Forest, MA (USA). Our simulation approach allowed us to better understand the interplay between instrument parameters and landscape characteristics, and facilitated comprehensive traceability of error budgets. To enhance our understanding of the impact of sensor design on classification performance, we simulated image samples at different spatial, spectral, and scale resolutions (by modifying the pixel pitch and the total number of pixels in the sensor array, i.e., the focal plane dimension) of the imaging sensor and assessed the performance of a deep learning-based convolutional neural network (CNN) and a traditional machine learning classifier, support vector machines (SVMs), to classify vegetation species. Overall, across all resolutions and species mixtures, the highest classification accuracy varied widely from 50 to 84%, and the number of genus-level species classes identified ranged from 2 to 17, among 24 classes. Harnessing this simulation approach has provided us valuable insights into sensor configurations and the optimization of data collection methodologies to improve the interpretation of spectral signatures for accurate tree species mapping in forest scenes. Note that we used species classification as a proxy for a host of imaging spectroscopy applications. However, this approach can be extended to other ecological scenarios, such as in evaluating the changing ecosystem composition, detecting invasive species, or observing the effects of climate change on ecosystem diversity.

Plant traits and insect herbivory have been highly studied within the modern record but only to a limited extent within the paleontological. Preservation influences what can be measured within the fossil record, but modern methods are also not compatible with paleobotanical methods. To remedy this knowledge gap, a comparable framework was created here using modern and paleobotanical methods, allowing for future comparisons within the fossil record. Insect feeding damage on selected tree species at Harvard Forest, the Smithsonian Environmental Research Center, and La Selva were characterized using the damage type system prevalent within paleobotanical studies and compared with leaf traits. Linear models and random forest analyses tested the influence of leaf traits on total, specialized, gall, and mine frequency and diversity. Structural traits like leaf dry mass per area and palatability traits, including lignin and phosphorus concentrations, are important variables affecting gall and mine damage. The significance and strength of trait-herbivory relationships varied across forest types, which is likely driven by differences in local insect populations. This work addresses the persistent gap between modern and paleoecological studies focusing on the influence of leaf traits on insect herbivory. This is important as modern climate change alters our understanding of plant-insect interactions, providing a need for contextualizing these relationships within evolutionary time. The fossil record provides information on terrestrial response to past climatic events and, thus, should be implemented when considering how to preserve biodiversity under current and future global change.