Publicação
GeoAI approach to Vineyard Yield Estimation
| datacite.subject.fos | Ciências Naturais::Geografia Económica e Social | pt_PT |
| dc.contributor.advisor | Neto, Miguel de Castro Simões Ferreira | |
| dc.contributor.advisor | Gil, Artur José Freire | |
| dc.contributor.author | Barriguinha, André Figueiredo | |
| dc.date.accessioned | 2023-07-05T14:01:15Z | |
| dc.date.available | 2023-07-05T14:01:15Z | |
| dc.date.issued | 2023-06-26 | |
| dc.description | A thesis submitted in partial fulfillment of the requirements for the degree of Doctor in Information Management, specialization in Geographic Information Systems | pt_PT |
| dc.description.abstract | Knowing in advance vineyard yield is a key issue for growers, winemakers, policy makers, and regulators being fundamental to achieve the best balance between vegetative and reproductive growth, and to allow more informed decisions like thinning, irrigation and nutrient management, schedule harvest, optimize winemaking operations, program crop insurance, fraud detection and grape picking workforce demand. In a long-term scenario of perceived climate change, it is also essential for planning and regulatory purposes at the regional level. Estimating yield is complex and requires knowing driving factors related to climate, plant, and crop management that directly influence the number of clusters per vine, berries per cluster, and berry weight. These three yield components explain 60%, 30%, and 10% of the yield. The traditional methods are destructive, labor-demanding, and time-consuming, with low accuracy primarily due to operator errors and sparse sampling (compared to the inherent spatial variability in a production vineyard). Those are supported by manual sampling, where yield is estimated by sampling clusters weight and the number of clusters per vine, historical data, and extrapolation considering the number of vines in a plot. As the extensive research in the area clearly shows, improved applied methodologies are needed at different spatial scales. The methodological approaches for yield estimation based on indirect methods are primarily applicable at small scale and can provide better estimates than the traditional manual sampling. They mainly depend on computer vision and image processing algorithms, data-driven models based on vegetation indices and pollen data, and on relating climate, soil, vegetation, and crop management variables that can support dynamic crop simulation models. Despite surpassing the limitations assigned to traditional manual sampling methods with the same or better results on accuracy, they still lack a fundamental key aspect: the real application in commercial vineyards. Another gap is the lack of solutions for estimating yield at broader scales (e.g., regional level). The perception is that decisions are more likely to take place on a smaller scale, which in some cases is inaccurate. It might be the case in regulated areas and areas where support for small viticulturists is needed and made by institutions with proper resources and a large area of influence. This is corroborated by the fact that data-driven models based on Trellis Tension and Pollen traps are being used for yield estimation at regional scales in real environments in different regions of the world. The current dissertation consists of the first study to identify through a systematic literature review the research approaches for predicting yield in vineyards for wine production that can serve as an alternative to traditional estimation methods, to characterize the different new approaches identifying and comparing their applicability under field conditions, scalability concerning the objective, accuracy, advantages, and shortcomings. In the second study following the identified research gap, a yield estimation model based on Geospatial Artificial Intelligence (GeoAI) with remote sensing and climate data and a machine-learning approach was developed. Using a satellite-based time-series of Normalized Difference Vegetation Index (NDVI) calculated from Sentinel 2 images and climate data acquired by local automatic weather stations, a system for yield prediction based on a Long Short-Term Memory (LSTM) neural network was implemented. The results show that this approach makes it possible to estimate wine grape yield accurately in advance at different scales. | pt_PT |
| dc.identifier.tid | 101356340 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/154859 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | pt_PT |
| dc.subject | Vineyard | pt_PT |
| dc.subject | Yield | pt_PT |
| dc.subject | Estimation | pt_PT |
| dc.subject | Prediction | pt_PT |
| dc.subject | Forecasting | pt_PT |
| dc.subject | Systematic Literature Review | pt_PT |
| dc.subject | Remote Sensing | pt_PT |
| dc.subject | Vineyard | pt_PT |
| dc.subject | Yield | pt_PT |
| dc.subject | Estimation | pt_PT |
| dc.subject | Prediction | pt_PT |
| dc.subject | NDVI | pt_PT |
| dc.subject | Climate | pt_PT |
| dc.subject | Machine Learning | pt_PT |
| dc.title | GeoAI approach to Vineyard Yield Estimation | pt_PT |
| dc.type | doctoral thesis | |
| dspace.entity.type | Publication | |
| person.familyName | Barriguinha | |
| person.givenName | André | |
| person.identifier.ciencia-id | 791D-C1E0-1B0D | |
| person.identifier.orcid | 0000-0002-9533-3419 | |
| person.identifier.rid | E-2851-2013 | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | doctoralThesis | pt_PT |
| relation.isAuthorOfPublication | 4039d1d1-071c-43b6-9b71-ba6454c08187 | |
| relation.isAuthorOfPublication.latestForDiscovery | 4039d1d1-071c-43b6-9b71-ba6454c08187 | |
| thesis.degree.name | Doutoramento em Gestão da Informação, especialização em Sistemas de Informação Geográfica | pt_PT |