Machine learning (ML) can accelerate the extraction of phenological data from herbarium specimens; however, no studies have assessed whether ML-derived phenological data can be used reliably to evaluate ecological patterns. In this study, 709 herbarium specimens representing a widespread annual herb, Streptanthus tortuosus, were scored both manually by human observers and by a mask R-CNN object detection model to (1) evaluate the concordance between ML and manually-derived phenological data and (2) determine whether ML-derived data can be used to reliably assess phenological patterns. The ML model generally underestimated the number of reproductive structures present on each specimen; however, when these counts were used to provide a quantitative estimate of the phenological stage of plants on a given sheet (i.e., the phenological index or PI), the ML and manually-derived PI’s were highly concordant. Moreover, herbarium specimen age had no effect on the estimated PI of a given sheet. Finally, including ML-derived PIs as predictor variables in phenological models produced estimates of the phenological sensitivity of this species to climate, temporal shifts in flowering time, and the rate of phenological progression that are indistinguishable from those produced by models based on data provided by human observers. This study demonstrates that phenological data extracted using machine learning can be used reliably to estimate the phenological stage of herbarium specimens and to detect phenological patterns.
more »
« less
Using photographs and deep neural networks to understand flowering phenology and diversity in mountain meadows
Mountain meadows are an essential part of the alpine–subalpine ecosystem; they provide ecosystem services like pollination and are home to diverse plant communities. Changes in climate affect meadow ecology on multiple levels, for example, by altering growing season dynamics. Tracking the effects of climate change on meadow diversity through the impacts on individual species and overall growing season dynamics is critical to conservation efforts. Here, we explore how to combine crowd‐sourced camera images with machine learning to quantify flowering species richness across a range of elevations in alpine meadows located in Mt. Rainier National Park, Washington, USA. We employed three machine‐learning techniques (Mask R‐CNN, RetinaNet and YOLOv5) to detect wildflower species in images taken during two flowering seasons. We demonstrate that deep learning techniques can detect multiple species, providing information on flowering richness in photographed meadows. The results indicate higher richness just above the tree line for most of the species, which is comparable with patterns found using field studies. We found that the two‐stage detector Mask R‐CNN was more accurate than single‐stage detectors like RetinaNet and YOLO, with the Mask R‐CNN network performing best overall with mean average precision (mAP) of 0.67 followed by RetinaNet (0.5) and YOLO (0.4). We found that across the methods using anchor box variations in multiples of 16 led to enhanced accuracy. We also show that detection is possible even when pictures are interspersed with complex backgrounds and are not in focus. We found differential detection rates depending on species abundance, with additional challenges related to similarity in flower characteristics, labeling errors and occlusion issues. Despite these potential biases and limitations in capturing flowering abundance and location‐specific quantification, accuracy was notable considering the complexity of flower types and picture angles in this dataset. We, therefore, expect that this approach can be used to address many ecological questions that benefit from automated flower detection, including studies of flowering phenology and floral resources, and that this approach can, therefore, complement a wide range of ecological approaches (e.g., field observations, experiments, community science, etc.). In all, our study suggests that ecological metrics like floral richness can be efficiently monitored by combining machine learning with easily accessible publicly curated datasets (e.g., Flickr, iNaturalist).
more » « less- NSF-PAR ID:
- 10490713
- Publisher / Repository:
- Wiley Blackwell (John Wiley & Sons)
- Date Published:
- Journal Name:
- Remote Sensing in Ecology and Conservation
- ISSN:
- 2056-3485
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Phenology––the timing of life-history events––is a key trait for understanding responses of organisms to climate. The digitization and online mobilization of herbarium specimens is rapidly advancing our understanding of plant phenological response to climate and climatic change. The current common practice of manually harvesting data from individual specimens greatly restricts our ability to scale data collection to entire collections. Recent investigations have demonstrated that machine-learning models can facilitate data collection from herbarium specimens. However, present attempts have focused largely on simplistic binary coding of reproductive phenology (e.g., flowering or not). Here, we use crowd-sourced phenological data of numbers of buds, flowers, and fruits of more than 3000 specimens of six common wildflower species of the eastern United States (Anemone canadensis, A. hepatica, A. quinquefolia, Trillium erectum, T. grandiflorum, and T. undulatum} to train a model using Mask R-CNN to segment and count phenological features. A single global model was able to automate the binary coding of reproductive stage with greater than 90% accuracy. Segmenting and counting features were also successful, but accuracy varied with phenological stage and taxon. Counting buds was significantly more accurate than flowers or fruits. Moreover, botanical experts provided more reliable data than either crowd-sourcers or our Mask R-CNN model, highlighting the importance of high-quality human training data. Finally, we also demonstrated the transferability of our model to automated phenophase detection and counting of the three Trillium species, which have large and conspicuously-shaped reproductive organs. These results highlight the promise of our two-phase crowd-sourcing and machine-learning pipeline to segment and count reproductive features of herbarium specimens, providing high-quality data with which to study responses of plants to ongoing climatic change.more » « less
-
more » « less
Abstract We use a quantitative model of photosynthesis to explore leaf‐level limitations to plant growth in an alpine tundra ecosystem that is expected to have longer, warmer, and drier growing seasons. The model is parameterized with abiotic and leaf trait data that is characteristic of two dominant plant communities in the alpine tundra and specifically at the Niwot Ridge Long Term Ecological Research Site: the dry and wet meadows. Model results produce realistic estimates of photosynthesis, nitrogen‐use efficiency, water‐use efficiency, and other gas exchange processes in the alpine tundra. Model simulations suggest that dry and wet meadow plant species do not significantly respond to changes in the volumetric soil moisture content but are sensitive to variation in foliar nitrogen content. In addition, model simulations indicate that dry and wet meadow species have different maximum rates of assimilation (normalized for leaf nitrogen content) because of differences in leaf temperature. These differences arise from the interaction of plant height and the abiotic environment characteristic of each plant community. The leaf temperature of dry meadow species is higher than wet meadow species and close to the optimal temperature for photosynthesis under current conditions. As a result, 2°C higher air temperatures in the future will likely lead to declines in dry meadow species’ carbon assimilation. On the other hand, a longer and warmer growing season could increase nitrogen availability and assimilation rates in both plant communities. Nonetheless, a temperature increase of 4°C may lower rates of assimilation in both dry and wet meadow plant communities because of higher, and suboptimal, leaf temperatures.
-
Background Landscape composition is known to affect both beneficial insect and pest communities on crop fields. Landscape composition therefore can impact ecosystem (dis)services provided by insects to crops. Though landscape effects on ecosystem service providers have been studied in large-scale agriculture in temperate regions, there is a lack of representation of tropical smallholder agriculture within this field of study, especially in sub-Sahara Africa. Legume crops can provide important food security and soil improvement benefits to vulnerable agriculturalists. However, legumes are dependent on pollinating insects, particularly bees (Hymenoptera: Apiformes) for production and are vulnerable to pests. We selected 10 pigeon pea (Fabaceae: Cajunus cajan (L.)) fields in Malawi with varying proportions of semi-natural habitat and agricultural area within a 1 km radius to study: (1) how the proportion of semi-natural habitat and agricultural area affects the abundance and richness of bees and abundance of florivorous blister beetles (Coleoptera: Melloidae ), (2) if the proportion of flowers damaged and fruit set difference between open and bagged flowers are correlated with the proportion of semi-natural habitat or agricultural area and (3) if pigeon pea fruit set difference between open and bagged flowers in these landscapes was constrained by pest damage or improved by bee visitation. Methods We performed three, ten-minute, 15 m, transects per field to assess blister beetle abundance and bee abundance and richness. Bees were captured and identified to (morpho)species. We assessed the proportion of flowers damaged by beetles during the flowering period. We performed a pollinator and pest exclusion experiment on 15 plants per field to assess whether fruit set was pollinator limited or constrained by pests. Results In our study, bee abundance was higher in areas with proportionally more agricultural area surrounding the fields. This effect was mostly driven by an increase in honeybees. Bee richness and beetle abundances were not affected by landscape characteristics, nor was flower damage or fruit set difference between bagged and open flowers. We did not observe a positive effect of bee density or richness, nor a negative effect of florivory, on fruit set difference. Discussion In our study area, pigeon pea flowers relatively late—well into the dry season. This could explain why we observe higher densities of bees in areas dominated by agriculture rather than in areas with more semi-natural habitat where resources for bees during this time of the year are scarce. Therefore, late flowering legumes may be an important food resource for bees during a period of scarcity in the seasonal tropics. The differences in patterns between our study and those conducted in temperate regions highlight the need for landscape-scale studies in areas outside the temperate region.more » « less
-
Climate change is likely to alter both flowering phenology and water availability for plants. Either of these changes alone can affect pollinator visitation and plant reproductive success. The relative impacts of phenology and water, and whether they interact in their impacts on plant reproductive success remain, however, largely unexplored. We manipulated flowering phenology and soil moisture in a factorial experiment with the subalpine perennial Mertensia ciliata (Boraginaceae). We examined responses of floral traits, floral abundance, pollinator visitation, and composition of visits by bumblebees vs. other pollinators. To determine the net effects on plant reproductive success, we also measured seed production and seed mass. Reduced water led to shorter, narrower flowers that produced less nectar. Late flowering plants produced fewer and shorter flowers. Both flowering phenology and water availability influenced pollination and reproductive success. Differences in flowering phenology had greater effects on pollinator visitation than did changes in water availability, but the reverse was true for seed production and mass, which were enhanced by greater water availability. The probability of receiving a flower visit declined over the season, coinciding with a decline in floral abundance in the arrays. Among plants receiving visits, both the visitation rate and percent of non-bumblebee visitors declined after the first week and remained low until the final week. We detected interactions of phenology and water on pollinator visitor composition, in which plants subject to drought were the only group to experience a late-season resurgence in visits by solitary bees and flies. Despite that interaction, net reproductive success measured as seed production responded additively to the two manipulations of water and phenology. Commonly observed declines in flower size and reward due to drought or shifts in phenology may not necessarily result in reduced plant reproductive success, which in M. ciliata responded more directly to water availability. The results highlight the need to go beyond studying single responses to climate changes, such as either phenology of a single species or how it experiences an abiotic factor, in order to understand how climate change may affect plant reproductive success.more » « less
