ItemThe socio-cultural terroir of Irish craft brewing(University College Cork, 2024) Day, Shawn; Crowley, John; O'Connor, Ray; Murphy, OrlaThis thesis provides a rich and unique exploration of the world of craft brewing in Ireland. One of the key concepts underpinning the research is that of socio-cultural terroir, which captures the all-important nexus between craft, practice, and place. Cultural geography provides a way of seeing and understanding the craft of brewing in all its richness, diversity and complexity. Foregrounding the brewers’ own experiences reveals how the craft is learned, affirmed and sustained. Applying emerging digital humanities methodologies such as textual analysis, information and knowledge visualisation to more conventional cultural geographic approaches allows for an exploration of how the journeys and values of Irish craft brewers emerge from, shape and (re)create meaning, identity and place in a rapidly growing and evolving community. Consisting of two parts, this thesis first seeks to bring a cultural geographic lens to bear on the craft brewing trade while carefully detailing the historical tradition from which it emerged, and secondly, it demonstrates how digital humanities practice can be employed to expand, augment, amplify, and enhance that exploration. The design, development, and deployment of an exploratory interactive platform to disseminate the findings facilitates an open sharing of the data, inviting further exploration, interpretation, and engagement with the research by a wider network of interested parties, including most importantly, the brewers themselves who have been a central focus of this research. ItemGeospatial data obfuscation methods applied to agricultural data(University College Cork, 2023) Nowbakht, Parvaneh; Holloway, Paul; Cawkwell, Fiona; O’Sullivan, Lilian; Wall, David; TeagascGeoprivacy protection is a controversial subject within the field of agri-environmental research. Agricultural transformation and digital farming are widely used in the agriculture industry. This enhances sustainable agriculture and subsequently sustainable development and food security. At the same time, such technological advances can mitigate climate change by increasing agricultural productivity while protecting natural resources and reducing the environmental impact of agricultural activities. Digital farming generates a high volume of data including spatial data. This data can be used in data-driven modelling to improve agriculture systems and design and develop more sustainable agricultural policies and services to enhance sustainable agriculture. Sharing and making data accessible to a wide range of researchers and stakeholders is essential to gain maximum use of data. Data privacy and particularly geoprivacy protection is an essential factor when sharing agricultural data and needs urgent investigation. To date, point-based obfuscation methods are the predominant approaches used to protect object confidentiality with spatial pattern and statistical accuracy preservation. In this research, functionality of point-based obfuscation methods for polygon nature objects was investigated. The terms “non-unique obfuscation” was introduced to the scientific literature for the first time. A high percentage of false-identification and non-unique obfuscation was recognized as the main drawback of point-based obfuscation when applied on polygon centroids as point spatial data. Agricultural spatial data is often best represented as static polygon data with an association of attributes/properties of a field parcel or farm that includes coordinates, shape, size, topology, the relationship with the surrounding environment, and the impact of external factors on region characteristics that can be used to breach privacy. Agricultural spatial data with unique characteristics that distinguish them from other spatial data, therefore, require different geoprivacy protection techniques. Therefore, for the first time, to achieve a high level of geoprivacy protection, several polygon-based obfuscation methods including PN*Rand, PDonut-k, PDensity-k, PAHilb, PDonut_AHilb and PESOM methods were developed with consideration of these properties and avoidance of the occurrence of false-identification and non-unique obfuscation. The comparison of the performance of obfuscation methods, both point-based and polygon-based, in different aspects using various evaluation metrics indicated that the density-based obfuscation methods provided a better trade-off between level of confidentiality and accuracy. The results demonstrated that PESOM method maintained a high level of geoprivacy protection and absolute environmental and climatic clustering preservation with no false identification and non-unique obfuscation risk. Following on from this. a case study was conducted to showcase the useability of obfuscated data to solve real world problems and highlight the importance of the choice of obfuscation method in outcome and suitability of obfuscated data for a certain purpose. Several evaluation metrics were developed to examine and assess the performance of obfuscation methods in terms of determining the level of privacy protection and spatial pattern preservation, and statistical accuracy. The results confirm the importance of choosing the right obfuscation method based on the influence of internal and external features on the results of the data-driven model. Therefore, the results of this research should be of wide interest to those working in GIScience, agri-environmental research, and computer science, and be of relevance to researchers and data managers. ItemApplications of big data and machine learning in global energy system modelling(University College Cork, 2022) Joshi, Siddharth; O'Gallachoir, Brian; Holloway, Paul; Glynn, James; Science Foundation IrelandGlobal efforts to limit atmospheric warming well below 2 degree celcius above pre-industrial levels form the backbone of our response to mitigate the detrimental effects of climate change. The energy sector contributes circa 75% of global GHG emissions, amongst which the Electricity and Heat sectors each contribute ~40%, and the Transport sector contributes ~20% to the total global energy-related GHG emissions. The recent IPCC AR6 report finds that in nearly all possible emission scenarios considered, the world is heading towards a 1.5 degree celcius global temperature rise by the early 2030s. Pursuant to this, Energy Systems Models (ESMs) and Integrated Assessment Models (IAMs) are essential tools that provide energy system pathways to limit global warming below the temperature threshold. Thus, improving the accuracies of ESMs and IAMs will lead to measurable improvement in energy policy formulation and evaluation,thereby increasing the likelihood of meeting the commitments under the Paris Climate Agreement. This thesis develops and applies novel frameworks and methods that use a big data and machine learning driven strategy to improve the technology potential assessment of global decentralised solar PV technology and projection of transport energy service demand. The frameworks and methods developed in this thesis are presented in a format of methodological design principles followed by a case study using them. Specifically, on the supply side, the thesis investigates the global high-resolution spatiotemporal technical potential of rooftop solar PV for 2015 and further growth in the technical potentials from 2020-2050. For this assessment case study, the developed framework utilises a suite of GIS derived geospatial metrics in conjunction with a custom machine learning framework to calculate the global rooftop area at a high spatial resolution. Further using an IAM, the role of decentralised solar PV in global future energy transitions is explored. On the demand side, the thesis introduces a new machine learning model called ‘TrebuNet’ that is capable of high accuracy in estimating future energy service demand in the transport sector. The thesis thus provides the first development of machine learning and GIS based methods to improve the accuracy of global ESMs and IAMs. Particular attention is also paid towards the reproduction and transparency of the methods and the frameworks developed in this thesis for cross- disciplinary research. The thesis contributes to the important task of climate change mitigation by providing a bridge between mature IAM and ESM modelling and emerging machine learning-big data-driven tools. In doing so, this thesis demonstrates how the emerging methods in conjunction with large geospatial open source data, can aid in improving the technology representation of variable renewable energy technology in energy systems. The thesis also lays the foundation for providing solutions to energy system related tasks that are currently limited by high computational costs and data. The datasets and analysis generated by this thesis are presently assisting in unlocking the global role of decentralised renewable energy technologies in future energy systems and are also encouraging shifts in national decarbonisation pathways. ItemOcean-surface heterogeneity mapping: exploiting hypertemporal datasets in support of seascape ecology research(University College Cork, 2022-04) Scarrott, Rory; Cawkwell, Fiona; Jessopp, Mark John; Cronin, Michelle; Cusack, Caroline; O’Rourke, Eleanor; Horizon 2020Seascape ecology provides us with a framework to explore the distribution of marine life throughout the world’s oceans. Studies often require both biological and environmental data from a variety of sources, which are increasingly complemented by data acquired using remote sensing and satellite-based sensors. Advances in remote sensing technology have equipped researchers with the capability to visualise environmental conditions, with great precision over large spatial scales. As records of environmental conditions, satellite-derived image data have proven useful in seascape ecology. Indeed, single-date and multi-date satellite imagery are already widely used in support of oceanographic and fisheries research and monitoring. However, with increases in the temporal frequency of imagery, a greater diversity of ocean surface features can be studied, modelled, and understood in terms of their development over time. This research examines the interface between oceanography, marine ecology, and remote sensing. It focuses on the use of a subset of temporally-rich satellite-derived imagery known as hypertemporal data, and its potential utility for seascape ecology studies. These high temporal resolution datasets are characterised by being: univariate in nature (e.g. Sea Surface Temperature); comprised of frequent, equally-spaced discrete time slices; precisely co-registered; and radiometrically consistent between images (i.e. they are measured using the same sensors, or are derived from inter-validated sensor systems). An in-depth review analyses nearly 25 years of available literature on the use of satellite-derived hypertemporal datasets. In general, they have been more widely used in terrestrial environments where in-situ validation costs are lower, and boundaries and features have greater permanency. By contrast, hypertemporal datasets have been under-utilised in ocean sciences. The review examines and describes the range of methodologies that have been adapted specifically for hypertemporal applications, in both marine and terrestrial contexts. It also identifies the priority research that should be done to maximise data usage for the ocean arena. In particular, the review highlighted the importance of prioritising data-driven methodologies in the near-term, and developing methods to map spatio-temporal heterogeneity (i.e. spatial and temporal variations in physical, chemical, or biological conditions) within datasets. A data-driven method to map Spatio-Temporal Heterogeneity (STH) using hypertemporal datasets is then presented. The Ocean-surface Heterogeneity MApping (OHMA) algorithm uses unsupervised ISODATA clustering, an ensemble approach, and a data-driven optimisation process, to identify where boundaries between different ocean regions frequently occur. OHMA effectively produces a suite of complementary datasets – a single STH image dataset that highlights the frequency at which boundaries between regions occur, and an optimised spatio-temporal classification of the ocean surface (both spatial cluster distributions, and their associated temporal profiles). A Sea Surface Temperature (SST) dataset, comprising weekly images of the North Atlantic Ocean from 2011, was used to develop the method. Validation, using Irish Marine Institute temperature measurements acquired along vessel tracks while the vessel is “Underway”, demonstrated an association between higher STH values and areas where locally extreme temperature transitions were more likely to occur. Generalised Linear Modelling of STH versus a combination of possible ocean-surface heterogeneity characteristics (fronts and currents), and oceanographically-known drivers (bathymetry), showed that the drivers of STH vary regionally. This highlighted the utility of OHMA to generate higher-level products. These contain spatial and temporal information that cannot be produced using only in-situ or infrequent satellite data. Following development of the OHMA methodology, the research examined its performance at a range of spatial and temporal scales, and the considerations needed to integrate STH mapping into seascape ecology analyses. Two OHMA-applications on SST data used relatively well-understood waters around the West European Archipelago, and ecologically important thermohaline and tidal front systems, to explore the methodology’s potential. The first OHMA-application focused on the algorithm’s performance at higher spatial scales, and sub-annual timeframes. STH values, derived from a 48-day hypertemporal SST dataset from summer 2019, were compared to a difference grid generated from a network of Conductivity, Temperature and Depth (CTD) measurements, and hydrographic information derived from Underway and CTD cast data. Higher value STH features were also compared to a number of well-documented tidal fronts, and a thermohaline front. Statistically significant positive correlations were identified between STH values and a difference in surface water characteristics between CTD sites. While the approach conveyed detailed spatio-temporal information on tidal fronts, it performed poorly at framing the thermohaline front. Examining the hydrographic information revealed the complexity in interpreting these associations. The study highlighted the value of using STH datasets to objectively characterise the spatio-temporal behaviour of fronts, and the surface structure of stratified shelf seas. It also suggested an annual cycle of STH should be analysed, to clarify whether thermohaline front information could be gathered. The second application examined the use of multiple temporal scales within a preliminary seascape ecology study, comparing SST-derived STH to a number of ecosystem indicators concerning primary producer abundance (phytoplankton) and fishing fleet activity. In doing so, it also further examined temporal scale considerations when using hypertemporal data. STH datasets were produced from different temporal inputs (an annual 7-day SST dataset versus 30/31-day daily SST datasets over 2018). SST-STH, model-derived phytoplankton concentration, and apparent commercial fishing effort data (derived from a behavioural classification of Automatic Identification System data, recording vessel tracks) were compared for February, March, July, and August using a correlation-based approach. Positive associations were found between SST-STH and fishing effort. This study demonstrated there could be important, usable, biophysical linkages between trophic indicators, and the higher-level heterogeneity conveyed by a STH dataset. It also highlighted the challenges the OHMA approach faced when used on multi-temporal, rather than hypertemporal, data. Additionally, it clarified SST-STH maps could convey information on thermohaline fronts when OHMA was used on SST datasets at annual timeframes, or datasets framing times when continental shelf seas are not thermally stratified. The thesis culminates in a discussion of how hypertemporal data could be used to support seascape ecology applications, using the development of the OHMA approach to provide evidence and guidance. Discussions highlight the key advances the research has made. Namely, the research has clarified that there are benefits to exploiting the hypertemporal data resource within seascape ecology research. The high-resolution data records subtle changes that can be used to better characterise ocean surface habitats. The research also produced a methodology that delivers objectively-derived, high-level heterogeneity information from hypertemporal data, demonstrated its use within a seascape ecology study, and described it along with best practice guidance for use. The limitations of the methodology are also clearly outlined, and include a cautionary note concerning the use of OHMA on non-hypertemporal datasets where the image number, or dataset variability, may be insufficient for the process to produce the full range of possible products. Concerning validation, the research highlighted the need for in-situ data to be collected which is spatio-temporally comparable to hypertemporal datasets, to improve quality assessment of, and confidence in, information products derived from hypertemporal datasets. Within this research in particular, the in-situ data should allow for estimates of spatio-temporal heterogeneity to be derived, which are comparable to those which can be derived from hypertemporal datasets. The discussion also emphasises a range of near-term (1-5 year) future research avenues, which include (i) exploiting available hypertemporal datasets from the ESA Climate Change Initiative, (ii) using OHMA-derived STH maps to stratify in-situ sampling of ocean-surface features, (iii) demonstrating the use of OHMA coupled with time-series analysis of ocean-surface parameters, and (iv) examining multi-trophic interactions between primary producers, predators, and STH features in waters around the West European Archipelago and the North Atlantic. Such near-term studies would enhance our understanding of the methodology, and use hypertemporal data to expand our understanding of ocean-surface structure at different spatial and temporal scales. In addition, they would use spatio-temporal heterogeneity mapping to diversify the range of hypertemporal methodologies being deployed on ocean-surface parameters, further enabling seascape ecology studies to exploit the hypertemporal data resource. ItemNational farm scale estimates of grass yield from satellite remote sensing(University College Cork, 2021-12-03) Marwaha, Richa; Cawkwell, Fiona; Green, Stuart; TeagascGlobally, grasslands are an important source of food for livestock and provide additional ecosystem services such as greenhouse gas (GHG) mitigation through carbon sequestration, habitats for biodiversity, and recreational amenities. Grass is the cheapest source of fodder providing Irish farmers with an economic benefit against international competitors. Hence, to maintain profitability, farmers have to maximize the proportion of grazed grass in cow’s diet or save it as silage. The overall objective of the current research project was to build a machine-learning model to estimate grass growth nationally using earth observation imagery from the Sentinel 2 satellite constellation and ancillary meteorological data, which are known to influence grass growth. Firstly, the impact of meteorological data and Growing Degree Days (GDD) was assessed for Teagasc Moorepark experimental farm (Fermoy, Co Cork, Ireland). GDD was modified to include Soil Moisture Deficit (SMD), which included the impact of summer drought conditions in 2018. Results demonstrated the importance of GDD for grass growth estimation using ordinary linear regression (OLS). The potential evapotranspiration (PE) 0.65 (r=0.65) and evaporation (r=0.65) were equally significant variables in 2017, while in 2018 the solar radiation had the highest correlation (r=0.43), followed by potential evapotranspiration and evaporation with r of 0.42. The standard and modified GDD were equally significant variables with r of 0.65 in 2017, but both had a reduced correlation in 2018 with modified GDD (0.38, p<0.01) performing slightly better than the standard GDD (0.26, p<0.01) calculation. These models only explained 53% (RMSE of 18.90 kg DM ha-1day-1) and 36% (RMSE of 27.02 kg DM ha-1day-1) of variability in grass growth for 2017 and 2018, respectively. Considering the importance of meteorological data, an empirical grass model called the Brereton model, previously used for Irish grass growing conditions were tested. Since this model lacks a spatial element, we compared the Brereton model with the previously used machine-learning model ANFIS and Random Forest (RF) with the combination of satellite data and meteorological data for eight Teagasc farms. Overall, the machine-learning algorithms (R2= 0.32 to 0.73 and RMSE=14.65 to 24.76 kg DM ha-1day-1 for the test data) performed better than the Brereton model (range of R2=0.03 to 0.33 and RMSE=41.68 to 82.29 kg DM ha-1day-1). The RF model (with all the variables except rainfall) had the highest accuracy for predicting grass growth rate, with (R2= 0.55, RMSE = 14.65 kg DM ha-1day-1, MSE= 214.79 kg DM ha-1day-1 versus ANFIS with R2 = 0.47, RMSE = 15.95 kg DM ha-1day-1, MSE= 254.40 kg DM ha-1day-1). When developing a national model, meteorological data were missing (except precipitation). A different approach was followed, whereby the grass growing season was subdivided (January-June Agmodel 1 and July–December Agmodel 2). Phenologically, the peak grass growth in Ireland typically occurs in May, with a slow decline in subsequent months. Spring is the most important season for grassland management, where growing conditions can impact the grass supply for the whole year. The national models were developed using Sentinel 2 band metrics, spectral indices (NDVI and NDRE), and rainfall for 179 farms. Data from 2017-2019 was divided into training and testing data (70:30 split), with 2020 data used for independent validation of the final trained model. Test accuracy was higher for Agmodel 1 (R2 = 0.74, RMSE= 15.52 kg DM ha-1day-1) versus Agmodel 2 (R2 = 0.58, RMSE= 13.74 kg DM ha-1day-1). This trained model was used on validation data from 2020, and the results were similar with better performance for Agmodel1 (R2 =0.70) versus Agmodel2 (R2=0.36). The improved spatial resolution of Sentinel 2 and the availability of red-edge bands showed improved results compared with previous work based on coarse resolution satellite imagery.