National Technical University of Athens (NTUA)
Παρακολούθηση και προσομοίωση υδρολογικών διεργασιών στο έδαφος με τη χρήση υδροπληροφορικής
Abstract
dc:descriptionSoil, and specifically soil moisture, which is described by the soil volumetric water content per unit volume (θ) and soil water potential (Ψ), constitutes the critical interface between all the main flows in the hydrologic cycle. The coupling and feedback of soil moisture with key parameters of the hydrologic balance, such as air temperature and precipitation, confirm its regulatory role in the terrestrial hydrologic cycle. The heterogeneity of the unsaturated zone, shaped by the spatial distribution of soil texture and structure, the percentage of organic matter in the soil, and the overlying vegetation, affects the extensive θ and the intensive Ψ variable of soil moisture, modifying the soil hydrologic processes. The introduction of the doctoral dissertation extensively discusses soil sampling methods, the determination of sampling points and volume, the technologies for determining hydraulic properties of the unsaturated zone, which are inextricably linked to the energy state of water in the soil and the soil structure. Although various robust laboratory and field methods, both invasive and non-invasive, have been developed for θ measurement, the development of methods for determining the for determining the soil hydraulic properties, particurarly the measurement of the matric potential (h) does not cover the entire range of its values. Additionally, point measurements of θ, even with profile commercial sensors providing the ability to measure at multiple depths (6 – 9) up to 1 m, cannot capture the full range of processes at every possible point within the unsaturated zone. The recovery of spatially distributed data of θ from vertical waveguides using waveform reconstruction algorithms provides the ability to determine it at infinitesimal points (based on sampling frequency) along the transmission lines. The simulation of processes in the unsaturated zone is commonly conducted with physical-based models, on one-dimensional scale as well as on a broader spatial scale, discretizing the characteristics of the unsaturated zone into homogeneous parts for the application of one-dimensional codes. Data-driven models can be used for regression tasks on both point and spatial scales. The literature review conducted within the framework of the this doctoral dissertation has been summarized in a detailed encyclopedic information database (wiki) (http://e-pyrros.gr/wiki-2/), aiming to initiate a research discussion for the determination of universally recognized technologies for measuring certain parameters of the hydrologic cycle. To approach the topic of the doctoral dissertation, the monitoring and simulation of soil hydrologic processes were conducted at multiple spatial scales, starting from the level of the experimental field to the level of the hydrologic basin. The basin where the monitoring and simulation methodology was applied is the Arta’s plain, shaped by the deposits of Louros and Aracthos rivers. The geological, lithological, pedological and crop characteristics were determined through extensive literature research, while the selection of the experimental field was based on the ever-increasing spread in the research area (kiwifruit orchards) and the characteristics of the soil properties in relation with the broader study area. In a limited area of the experimental field, commercial point sensors for measuring soil volumetric water content per unit area were installed at two depths (10 cm, 50 cm), as well as custom-made vertical waveguides with insulated transmission lines for the spatial recovery of θ along the transmission lines. One of them was also installed in part of the saturated zone (1.67 m below the soil surface) to investigate the conditions in the interaction of the zones. The fluctuations of the groundwater table are determined by a vented pressure sensor, installed in a piezometer next to the area of installation of the commercially and laboratory-manufactured soil volumetric water content sensors. On the surface of the experimental field, sampling was conducted with a hand auger in the upper and lower soil horizons (0 – 30 cm, 30 – 60 cm, respectively) to investigate the grain size distribution. Additionally, mineralogical analysis was carried out on selected soil samples to determine the mineral in the lower soil horizon. The results show uniform grain size distribution both in depth and spatial extent, while the mineralogical analysis revealed the presence of active and moderately active clay minerals. The identification of calcite and quartz mineral delineates the location of the experimental filed at the interface of Louros and Aracthos basins, indicating that in hydrologic basins with uniform elevation, where the watershed is approximately delineated, mineralogical analysis can lead to a detailed determination of the watershed. For the determination of the saturated hydraulic conductivity Ksat in the soil surface zone, single- and double-ring infiltration tests were carried out and cumulative infiltration curves were fitted by parametric models. In addition, undisturbed soil samples were taken, in the area of installation of commercial and custom-made sensors, to a depth of one meter, to characterize the hydraulic properties of the unsaturated zone in deeper layers. The laboratory methods (falling head permameter, centrifuge method and salt solution method), which were used, were optimized for undisturbed soil samples, preserving the soil structure, by developing innovative laboratory manufactured equipment. To extract the hydraulic properties of the unsaturated zone, in addition to Ksat the data of soil water characteristic curve were fitted through unimodal and bimodal parametric models. The spatial mapping of infiltration tests results indicates heterogeneous zones within the field, despite the uniform grain size distribution, indicating the influence of soil structure on the results. The investigation of soil hydraulic properties in depth revealed discrepancies with field infiltration tests in the upper soil layers, as field and laboratory methods tend to overestimate and underestimate infiltration rates respectively. Corrections based on the spread of wetting front to the single ring measurements balance the results with the laboratory measurements. The decreasing values of Ksat in the deeper layers of undisturbed soil samples are coupled with increased water retention. The observations of point commercial sensors and custom-made waveguides were managed through the development of classes in Python programming language. The data from the commercial sensors were used for the calibration of custom-made waveguide measurements, which are connected with time and frequency domain systems. The calibration results indicated insignificant discrepancies between the observed values of θ from the installed waveguides at the calibration depths, which coincide with the installation depths of the commercial sensors. An intriguing observation concerns the capillary rise at the time in which the groundwater level recovered from pumping. The coincidence of measurements from both TDR (Time Domain Reflectometry) and FDR (Frequency Domain Reflectometry) systems for the calculation of spatially distributed θ data indicates the potential application of low-cost systems for the recovery of θ data at infinitesimal points along the waveguide transmission lines. The simulation of the flow in the unsaturated zone at the point scale using the HYDRUS-1D code demonstrates the importance of experimentally determined hydraulic parameters in combination with data assimilation methods to achieve greater accuracy in the simulation. At the spatial scale of the experimental field and in the process of simulating flow in the subsurface zone (both unsaturated and saturated zone), the HYDRUS-1D and UZF codes coupled with MODFLOW-2005 were used, in which the flow in the unsaturated zone is described by discrete equations. The equation describing the flow in the UZF code ignores capillary flow in the unsaturated zone, which is a dominant hydrologic process in the experimental field, according to the measurements of insulated waveguides, and generally in soil systems where the groundwater table fluctuates in the proximity of the soil surface. The solution of the discrete equations leads to different recharge volumes in the hydrogeological model, with the HYDRUS-1D code producing 40% more recharge than the UZF code, concluding that the simulation of flow in the unsaturated zone with UZF code cannot capture the full range of soil hydrologic processes, especially in soil systems where the groundwater table lies in the proximity of the soil surface. The employment of machine learning algorithms to predict θ at discrete depths is common, however geographic extrapolation of algorithms to uncalibrated locations provides the ability to predict at points without observations. In the methodological framework developed for applying the three machine learning algorithms (Support Vector Machines, Random Forests and Voting Regressor), the prediction of θ was conducted with high accuracy in both temporal and geographical extrapolation context. The pre-trained models in the feature space of each agro-meteorological station in the plain, apart from temporal extrapolation with data from the same station, were also evaluated in geographic extrapolation with the feature space of other agro-meteorological stations in which were not trained. The spatiotemporal prediction of θ at high spatial resolution (250 x 250 m) over the extent of the plain was carried out by applying a methodological framework for decomposition of target attributes, while feeding a deep neural network with the hydraulic parameters (filed capacity, Ksat, θs, and wilting point) of the unsaturated zone retrieved from the three-dimensional soil hydraulic parameter database for Europe. The accuracy of the prediction at the evaluation station is adequate, indicating that the methodological framework can be applied to the spatiotemporal prediction of θ at the available hydraulic parameter locations of the plain.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Chrysanthopoulos, Efthymios
- Χρυσανθόπουλος, Ευθύμιος
Subjects
dc:subject × 22- Υδρολογία ακόρεστης ζώνης
- Αριθμητική προσομοίωση
- Μηχανική μάθηση
- Αισθητήρες
- Υδροπληροφορική
- Soil hydrology
- Numerical modeling
- Machine learning
- Sensors
- Hydroinformatics
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Τεχνητή νοημοσύνη
- Περιβαλλοντική Μηχανική
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη και Τεχνολογία υδάτων
- Computer and Information Sciences
- Natural Sciences
- Artificial Intelligence
- Environmental Engineering
- Engineering and Technology
- Water Science and Technology
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59452