Hannah Julia O'Sullivan

Ms. Hannah Julia O'Sullivan

orcid

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
5 April 2024 in Remote Sensing
Verified
Journal Article
Open access
PDF file
CDCesar Ivan Alvites Diaz
Cesar Ivan Alvites Diaz
HOHannah Julia O'Sullivan
Hannah Julia O'Sullivan
Saverio FranciniSaverio Francini
Saverio Francini
MMMarco Marchetti
Marco Marchetti
Giovanni SantopuoliGiovanni Santopuoli
Giovanni Santopuoli
EBERIKA BAZZATO
ERIKA BAZZATO
Accurate structural information about forests, including canopy heights and diameters, is crucial for quantifying tree volume, biomass, and carbon stocks, enabling effective forest ecosystem management, particularly in response to changing environmental conditions. Since late 2018, NASA’s Global Ecosystem Dynamics Investigation (GEDI) mission has monitored global canopy structure using a satellite Light Detection and Ranging (LiDAR) instrument. While GEDI has collected billions of LiDAR shots across a near-global range (between 51.6°N and >51.6°S), their spatial distribution remains dispersed, posing challenges for achieving complete forest coverage. This study proposes and evaluates an approach that generates high-resolution canopy height maps by integrating GEDI data with Sentinel-1, Sentinel-2, and topographical ancillary data through three machine learning (ML) algorithms: random forests (RF), gradient tree boost (GB), and classification and regression trees (CART). To achieve this, the secondary aims included the following: (1) to assess the performance of three ML algorithms, RF, GB, and CART, in predicting canopy heights, (2) to evaluate the performance of our canopy height maps using reference canopy height from canopy height models (CHMs), and (3) to compare our canopy height maps with other two existing canopy height maps. RF and GB were the top-performing algorithms, achieving the best 13.32% and 16% root mean squared error for broadleaf and coniferous forests, respectively. Validation of the proposed approach revealed that the 100th and 98th percentile, followed by the average of the 75th, 90th, 95th, and 100th percentiles (AVG), were the most accurate GEDI metrics for predicting real canopy heights. Comparisons between predicted and reference CHMs demonstrated accurate predictions for coniferous stands (R-squared = 0.45, RMSE = 29.16%).

5 October 2021 in Annals of Botany
Verified
Journal Article
Open access
PDF file
HOHannah Julia O'Sullivan
Hannah Julia O'Sullivan
PRPasi Raumonen
Pasi Raumonen
PKPekka Kaitaniemi
Pekka Kaitaniemi
JPJari Perttunen
Jari Perttunen
RSRisto Sievänen
Risto Sievänen
Background: Woody plants (trees and shrubs) play an important role in terrestrial ecosystems, but their size and longevity make them difficult subjects for traditional experiments. In the last 20 years functional–structural plant models (FSPMs) have evolved: they consider the interplay between plant modular structure, the immediate environment and internal functioning. However, computational constraints and data deficiency have long been limiting factors in a broader application of FSPMs, particularly at the scale of forest communities. Recently, terrestrial laser scanning (TLS), has emerged as an invaluable tool for capturing the 3-D structure of forest communities, thus opening up exciting opportunities to explore and predict forest dynamics with FSPMs.Scope: The potential synergies between TLS-derived data and FSPMs have yet to be fully explored. Here, we summarize recent developments in FSPM and TLS research, with a specific focus on woody plants. We then evaluate the emerging opportunities for applying FSPMs in an ecological and evolutionary context, in light of TLS-derived data, with particular consideration of the challenges posed by scaling up from individual trees to whole forests. Finally, we propose guidelines for incorporating TLS data into the FSPM workflow to encourage overlap of practice amongst researchers.Conclusions: We conclude that TLS is a feasible tool to help shift FSPMs from an individual-level modelling technique to a community-level one. The ability to scan multiple trees, of multiple species, in a short amount of time, is paramount to gathering the detailed structural information required for parameterizing FSPMs for forest communities. Conventional techniques, such as repeated manual forest surveys, have their limitations in explaining the driving mechanisms behind observed patterns in 3-D forest structure and dynamics. Therefore, other techniques are valuable to explore how forests might respond to environmental change. A robust synthesis between TLS and FSPMs provides the opportunity to virtually explore the spatial and temporal dynamics of forest communities.

27 March 2022
Verified
Conference Paper
Open access
HOHannah Julia O'Sullivan
Hannah Julia O'Sullivan
Terrestrial laser scanning (TLS) is quickly becoming an indispensable tool for quantifying individual tree and whole forest structure. In recent years, developments have been made in several areas, including tree segmentation from whole forest scans, quantitative structural model (QSM) reconstruction and leaf-wood separation (Åkerblom et al., 2017, Burt et al., 2019, Wang et al., 2020). One research area that could benefit greatly from a widespread use of TLS is that of functional-structural plant models (FSPMs), which simulate both the structure and function of plants. In FSPMs, the structural component of these models is usually derived using theoretical methods that reproduce remarkably realistic tree forms (Perttunen and Sievänen, 2005). However, it is often challenging to compare and validate these structural models with real trees, particularly those occurring in ‘natural environments’.   TLS-derived structural models in FSPMs have the potential to uncover new insights into the role of tree architecture on physiological plant processes, particularly in how tree 3D shape influences resource capture. Here, we will assess the impact of light fluctuations in the crown on individual tree productivity, scaling up to the productivity of whole forest stands. We will outline a new generic method that bridges the gap between TLS data and FSPMs as well as introduce ways of utilising other remote sensing techniques to validate FSPM outputs. Firstly, we will produce individual tree QSMs from a whole forest point cloud. Secondly, we will simulate tree-by-tree productivity in the context of the original forest environment. Lastly, we will use drone and satellite imagery to validate the FSPM productivity outputs.    Explicit 3D tree structure is an often-overlooked component of vegetation modelling, despite the feedback process between form and productivity.  We aim to highlight the role of light microenvironments within the crown and understand how uneven resource capture might extrapolate to whole forest estimates of productivity. We hope that this new method will encourage overlap of practice between researchers in this growing field and lead to further use of virtual plants in studies of tree evolution and ecology. Citations: Åkerblom, M., Raumonen, P., Mäkipää, R. and Kaasalainen, M., 2017. Automatic tree species recognition with quantitative structure models. Remote Sensing of Environment, 191, pp.1-12  Perttunen, J. and Sievänen, R., 2005. Incorporating Lindenmayer systems for architectural development in a functional-structural tree model. Ecological modelling, 181(4), pp.479-491.  Burt, A., Disney, M. and Calders, K., 2019. Extracting individual trees from lidar point clouds using treeseg. Methods in Ecology and Evolution, 10(3), pp.438-445.  Perttunen, J. and Sievänen, R., 2005. Incorporating Lindenmayer systems for architectural development in a functional-structural tree model. Ecological modelling, 181(4), pp.479-491.  Wang, D., Momo Takoudjou, S. and Casella, E., 2020. LeWoS: A universal leaf‐wood classification method to facilitate the 3D modelling of large tropical trees using terrestrial LiDAR. Methods in Ecology and Evolution, 11(3), pp.376-389.

20 January 2025
Verified
Conference Paper
Open access
Samuli JunttilaSamuli Junttila
Samuli Junttila
MBMinna Blomqvist
Minna Blomqvist
VLVille Laukkanen
Ville Laukkanen
HOHannah Julia O'Sullivan
Hannah Julia O'Sullivan
APAntti Polvivaara
Antti Polvivaara
HPHeli Peltola
Heli Peltola
Simultaneous increases of severe drought and heat extremes with bark beetle outbreaks have recently increased tree mortality globally. The lack of accurate tree mortality data over large areas has limited the development and applications of tree mortality models. Available tree mortality data has mainly been collected using field observations with limited spatial extent. However, the use of high-resolution remote sensing data, such as aerial imagery with automated imagery analysis, may change this situation. In this study, we analysed the development of tree mortality (standing dead wood) and factors contributing to it during 2017-2023 over an area of 117 366 ha of boreal forested landscape in Southern Finland. For this purpose, we developed a convolutional neural network (CNN) based on U-Net architecture, allowing segmenting of standing dead wood automatically from aerial imagery in 2017, 2020 and 2023 with a spatial resolution of 0.5 m. We trained the model using 22300 segments of manually delineated dead trees from various geographic regions in Finland. The model showed high accuracy in detecting the dead trees with an F1 score of 0.93 based on an independent validation dataset. We also combined the information on detected dead trees with open forest resource information based on extensive field campaigns and airborne laser scanning, to estimate standing dead wood volume during 2017-2023. The total standing dead wood volume increased substantially in our boreal study area, by 543 % from 8660 m3 to 52659 m3 between 2017 and 2023. Similarly, the total forest area of standing dead wood increased by 456 % between 2017 and 2023. Both variables followed an exponential growth curve with a nearly perfect fit for the 2017, 2020 and 2023 time series, indicating that tree mortality increased rapidly. Tree mortality occurred mainly in Norway spruce (Picea abies) dominated forests on relatively fertile soils. The mean age of forest stands suffering from tree mortality decreased from 69.7 to 62.6 years between 2017 and 2023, indicating that tree mortality has transitioned towards younger forests. Our findings highlight the increasing risk of tree mortality in boreal forests and the need for large-scale monitoring to keep up to date on the fast-paced changes in boreal forest mortality. This is also required for timely risk management measures in forestry under changing climate, associated with simultaneous increases of severe drought and heat extremes with bark beetle outbreaks.

20 January 2025
Verified
Conference Paper
Open access
HOHannah Julia O'Sullivan
Hannah Julia O'Sullivan
Samuli JunttilaSamuli Junttila
Samuli Junttila
Drought-induced tree mortality in the Anthropocene has increased at an unprecedented rate and is expected to increase even more sharply in the next few decades due to anthropogenic climate change. Fennoscandian boreal forests are particularly at risk of drought-related mortality because they are less adapted to drought conditions compared with lower latitude forests. Currently, tree mortality research is inhibited by a lack of data on tree deaths, but we can still anticipate mortality events by monitoring forest health. In this study we determine indicators of forest ‘hydraulic health’ which we define as the ability of a forest to withstand and recover from drought events. Ultimately, this may help to predict and prevent drought-induced tree deaths under future climatic shifts. To describe the hydraulic health of Fennoscandian forests we examine vegetation water content at both short (sub-daily) and intermediate (weekly) timescales using remotely sensed vegetation optical depth (VOD) measurements. Whilst short-timescale VOD is more tightly linked to water stress, intermediate-timescale VOD can demonstrate forest resilience and recovery to drought events. Here, we use a multivariate time series model to integrate VOD measurements at different timescales, presenting a novel opportunity to disentangle the drivers and responses of forests to drought. The method introduced here is a first step towards quantifying abiotic hazards for forests. By focusing on smaller time steps of VOD data we can minimise the noise attributed to biomass changes over time and can better detect changes in vegetation water content. We hope that by quantifying hydraulic health indicators, they can serve as an early warning system for drought-induced mortality or help in predicting other hazards such as wildfire. Although we focus on Fennoscandian forests, in principle, these methods could be applied to other biomes.