AI and earth observation-based mehodology for early warning of algal blooms in inland water bodies

Thomopoulou, Vasiliki; Θωμοπούλου Βασιλική

dc.contributor.author	Thomopoulou, Vasiliki	en
dc.contributor.author	Θωμοπούλου Βασιλική	el
dc.date.accessioned	2025-04-15T07:29:23Z
dc.date.available	2025-04-15T07:29:23Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/61749
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.29445
dc.rights	Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα	*
dc.rights.uri	http://creativecommons.org/licenses/by-nc-nd/3.0/gr/	*
dc.subject	Remote Sensing	en
dc.subject	Satellite Imagery	en
dc.subject	Water Resources	en
dc.subject	Eutrophication	en
dc.subject	Lake Yliki	en
dc.title	AI and earth observation-based mehodology for early warning of algal blooms in inland water bodies	en
heal.type	masterThesis
heal.classification	Water resources management	en
heal.classification	Remote Sensing	en
heal.language	el
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2024-10-04
heal.abstract	Water quality plays a critical role in various aspects of life, influencing human health, agriculture, and ecosystem stability. However, human activities affect, and many of them degrade, water quality, such as the excessive use of fertilizers in crop production. Fertilizers, consisting mainly of nutrients, are a major cause of eutrophication in water bodies. Specifically, excessive amounts of nitrogen and phosphorus - key nutrients in fertilizers - are washed into inland water systems, leading to phenomena such as harmful algal blooms. The link between water quality and agricultural practices, particularly those linked to eutrophication, poses the need for effective monitoring and early detection of bloom events in water bodies. A remedy to this challenge is the development of an early warning system for nutrient runoff in inland water bodies. However, such developments are hampered substantially, on the one hand, from the complexity and large-scale nature of water systems, and involved processes, and on the other hand, from the unavailability of ground metering stations, and hence large datasets. In this context, Earth Observation (EO) has a key role to play since it enables the large-scale, coherent, as well as near-real-time, monitoring of environmental and meteorological variables. This study introduces a methodological framework for the early warning of nutrient runoff in inland water bodies, exploiting a variety of ground-based and satellite-based observations. The framework is composed of the following three key components: 1. Crop-type classification component to estimate nutrient loads in the upstream watershed, utilizing Satellite Image Time Series in a supervised learning context. In this component a "Normalized Nutrient Load" index is proposed to quantify the severity of these loads. 2. The rainfall-runoff continuous simulation component, trained and fed by EO, generates daily inflows in the downstream water body. 3. The final component integrates the first two factors (the nutrient load severity and the discharge estimations) with meteorological EO forecasts. The outcome of the system is the bloom-occurrence probability in the lake for the next day.". The study area of the present study is the catchment area of the Boeotian Kifissos River, which has extensive agricultural activity, mainly in the area of Copaida. The case of Lake Yliki is of particular interest, not only because it is located downstream of a very fertilizer (and nutrient) polluted basin, but also because it is a major source of water supply for the Athens area, and is a key element of the external water supply system of EYDAP. The catchment is highly complex and is characterized by karstic formations, which imply significant water losses to the lower layers of the soil, and the existence of boreholes, which make hydrological simulation a particularly demanding process. At the same time, there are many crops in the area, which are characterized by a high need for fertilization throughout the year. The methodology integrates Artificial Intelligence (AI) techniques with Earth Observation (EO) datasets to create a holistic early warning system for nutrient runoff. By leveraging AI's analytical capabilities and the high spatio-temporal resolution of EO data—capturing various measured environmental variables—this approach provides comprehensive nutrient runoff estimations. Specifically, nutrient load severity estimation across basins utilizes data from the Sentinel-2 multispectral satellite mission to develop a computer vision algorithm for crop type classification. Rainfall-runoff estimation is achieved through a hydrological simulation that relies on EO data for soil properties, evapotranspiration rates, and a 24-hour precipitation forecast as the primary input driver. Finally, bloom occurrence probability is determined by combining meteorological EO forecasts with historical satellite data on algal blooms, as well as the estimated nutrient load and discharge levels. Each component shows promising results, demonstrating that EO data can be an effective tool in environmental monitoring and can be integrated to create a comprehensive early warning system for nutrient runoff. However, there is room for improvement, particularly in rainfall-runoff modeling, as the EO-based precipitation forecasts tend to underestimate actual precipitation when compared to in-situ measurements.	en
heal.advisorName	Makropoulos, Christos	en
heal.committeeMemberName	Kossieris, Panagiotis	en
heal.committeeMemberName	Chondros, Michalis K.	en
heal.committeeMemberName	Makropoulos, Christos	en
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Πολιτικών Μηχανικών	el
heal.academicPublisherID	ntua
heal.fullTextAvailability	false