University of Illinois Urbana-Champaign
Illinois Digital Environment for Access to Learning and Scholarship RepositoryNot a member yet
123813 research outputs found
Sort by
Discovering, autogenerating, and evaluating distractors for Parsons problems in CS1
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, David Smith IV, accepted the attached license on 2025-04-20 at 23:12.The student, David Smith IV, submitted this Dissertation for approval on 2025-04-20 at 23:22.This Dissertation was approved for publication on 2025-04-23 at 14:43.DSpace SAF Submission Ingestion Package generated from Vireo submission #21851 on 2025-10-19 at 18:18:27This dissertation presents work on evaluating the use of distractor blocks in Parsons problems, a problem type aimed at supporting novice programmers in learning to write code. Since their introduction by Parsons & Hayden [1], distractors have become common place in studies investigating Parsons problems. However, prior to the work presented in this dissertation, there had been a limited number of studies investigating their utility. In filling this gap, my work has focused on investigating methods of developing distractors and the impact of distractors in both summative and formative contexts. I first introduce a method of creating distractor templates from analysis of code writing errors and using those templates to support the autogeneration of distractors [2], a process that has been integrated into CodeSpec, a programming practice and eTextbook platform [3]. I present a series of studies comparing Parsons problems on exams that include distractors to those that do not. These studies have also included a comparison between different methods of including distractors, specifically comparing the use of distractors that are jumbled among the other options to distractors that are visually-grouped with their correct alternative [4, 5]. Additionally, I present a study comparing the impact on learning gains when practicing with Parsons problems that include distractors versus those without distractors. Learning gains are measured through pre- and post-assessment performance. In support of this work, I have also added the ability to associate feedback with distractors in the PrairieLearn assessment platform. The findings presented suggest that in summative contexts the presence of a single distractor group does not meaningfully increase question difficulty or item quality while causing students to spend significantly more time on the problems, a concern in a time constrained context such as an exam. However, as more distractor groups are added difficulty continues to increase with minimal impact—either positively or negatively—to item quality. This suggests that when distractors are added in sufficient quantities they can be useful for reducing the ceiling effect on score we observed in Parsons problems used in this study that do not include distractors. In formative contexts, the findings I present suggest that the presence of distractors causes students to engage more deeply with the code presented in blocks they are grouped with. The results of a randomized controlled trial indicate that students practicing with questions including distractors had measurable learning gains as measured through the post-test
Explicating the role of turning points and relationship satisfaction throughout mentoring relationships during medical residency: A communication perspective
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Lindsay Horan, accepted the attached license on 2025-04-21 at 15:12.The student, Lindsay Horan, submitted this Dissertation for approval on 2025-04-21 at 15:36.This Dissertation was approved for publication on 2025-04-21 at 16:24.DSpace SAF Submission Ingestion Package generated from Vireo submission #21869 on 2025-10-19 at 18:18:32Medical residents often engage in mentoring relationships during their residency to further their learning and professional development. Although mentoring has been cited as a way to mitigate some of the detrimental outcomes of medical residency (e.g., high stress, burnout), existing conceptualizations of mentoring leave knowledge gaps as to how residents experience relational shifts (i.e., turning points) with their mentors and how these turning points influence residents’ relationship satisfaction and well-being. To understand the nuance in mentoring relationship development, I conducted 42 semi-structured turning point retrospective interviews with Internal Medicine and Obstetrics/Gynecology residents in 26 residencies across the United States. First, turning point and trajectory analysis was conducted using residents' retrospective graphs collected during interviews to examine patterns within and across their mentoring relationships. The reported turning points revealed that residents were particularly attuned to their mentor’s (a) support in career development and (b) communication about relational norms (e.g., consistency of communication and quality of support). Additionally, residents’ graphs of their mentoring relationship trajectories revealed that residents primarily experienced a gradual increase in relationship satisfaction with their mentors, although some experienced immediate and maintained high satisfaction, significant disruptions, or turbulence. These turning point and trajectory findings suggest that mentoring relationships are uniquely shaped by the communication patterns between mentee and mentor. Second, answering Dorrance-Hall and colleagues’ (2023) call for future turning point research to capture participants’ accounts of turning points, I conducted accounts coding and thematic analysis on the interview transcripts portion of the data. Findings suggest that medical resident mentoring relationships are a complex process that requires situated communication strategies to be both satisfying and supportive. At a broad level, the communication strategies that bolstered residents’ relationship satisfaction included (a) establishing and maintaining communication norms, (b) providing high-quality emotional and informational support targeting residents’ learning and identity, and (c) creating opportunities for residents’ passive (i.e., observational) and active learning that align with their career goals. These patterns closely aligned with known mentoring behaviors, and this study extends knowledge of these behaviors by exploring how communication qualities influence residents’ relationship satisfaction and well-being. Finally, this study offers practical interventions, including communication strategies and training criteria designed to optimize mentoring benefits for residency programs, resident mentors, and resident mentees. Together, this study contributes to both theoretical and practical advancements at the intersection of communication, mentoring, and relational turning points
Towards the downstream effects of insufficient content moderation on the internet
No full textSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Margie Ruffin, accepted the attached license on 2025-04-22 at 18:33.The student, Margie Ruffin, submitted this Dissertation for approval on 2025-04-22 at 18:44.This Dissertation was approved for publication on 2025-04-23 at 09:41.DSpace SAF Submission Ingestion Package generated from Vireo submission #21904 on 2025-10-19 at 18:18:40Internet platforms rely heavily on content moderation to help drive positive engagement. Historically, this task has been done by humans, but in the recent decade, ML/AI models have been developed to reduce the burden by filtering out abusive content. While a step in the right direction, some moderation practices, manual and automated (e.g., context labeling, deepfake detection, and self-moderation), can still inadvertently cause harm to individual Internet users. To prevent the perpetuation of these harms, we must understand how they manifest regarding individuals’ perceptions, sentiments, and behaviors. This dissertation investigates how the ineffective content moderation practices of internet platforms affect how individuals respond to online abuse. Specifically, this dissertation explores how explanations (content-labeling) impact their viewers’ perceptions and sentiments (Chapter 3), how unmonitored deepfake-enabled profiles can sway an individual’s perception of and engagement with fake news (Chapter 4), and how low-moderation of platforms results in self-moderation to combat copyright violation harms (Chapter 5). In summary, these works show the downstream effects of deceitful and unauthorized content that slips through the moderation cracks, and assessing them is essential for developing better interventions
The unpaved road: examining the rural student experience and overcoming college access barriers
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Corinne Smith, accepted the attached license on 2025-04-23 at 20:30.The student, Corinne Smith, submitted this Dissertation for approval on 2025-04-23 at 20:44.This Dissertation was approved for publication on 2025-04-24 at 13:17.DSpace SAF Submission Ingestion Package generated from Vireo submission #21935 on 2025-10-19 at 18:18:57Rural students graduate high school at rates on par with their urban and suburban counterparts, yet they are the least likely group to enroll in college (Gibbs, 2000; Lavalley, 2018; NCES, 2011; NCES, 2015; Smith et al., 1995). Moreover, rural students face distinctive challenges in accessing and completing a college education (Goldman, 2019; Guiffrida, 2008; Maltzan, 2006; Paterson, 2020). In many cases, these outcomes can be attributed to the overlap between rurality and students’ college generational or socioeconomic status (Byun et al., 2012; Donovan, 2018; Marré, 2014; Tieken, 2020). This dissertation explored the effects of rurality, along with first-generation and low- income status, on college-going tendencies. The study sought to understand the barriers facing rural college applicants as well as the ways high-achieving students have overcome them. A descriptive qualitative approach was utilized to investigate the experiences of first-generation and low-income rural students attending one highly selective university. This study revealed the implications of social capital on the college search, application, and matriculation processes. Further, it identified key knowledge gaps, which impacted participants’ transition to college. The findings suggest immense personal initiative and self-determination helped high-achieving rural students overcome college access barriers. Implications for theory and research include adding a unique perspective on Social Capital Theory’s concepts of sense of belonging and place attachment (Schaefer-McDaniel, 2004) as well as demonstrating how “community” is based on common knowledge and values rather than physical location (Plagens, 2011). Within policy and practice, applying a contextual review process when evaluating rural student applications and providing formal campus support systems are identified as critical to supporting the admission and transition of rural students
Understanding the role of deep root water uptake in Amazon hydroclimate using modeling and observational approaches
No full textSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Carolina Bieri, accepted the attached license on 2025-04-25 at 15:24.The student, Carolina Bieri, submitted this Dissertation for approval on 2025-04-25 at 15:34.This Dissertation was approved for publication on 2025-04-28 at 10:32.DSpace SAF Submission Ingestion Package generated from Vireo submission #21960 on 2025-10-19 at 18:19:01Plant roots are critical pathways of moisture from the subsurface to the atmosphere. Deep moisture uptake by roots can act as a seasonal buffer mechanism in regions with a well-defined dry season such as the southern Amazon. Here, mature forests maintain transpiration (a critical source of atmospheric moisture in this part of the world) during the dry season. Most existing state-of-the-art earth system models do not have the necessary features to simulate subsurface-to-atmosphere moisture variations during drydowns. These features include groundwater dynamics, a sufficiently deep soil column, dynamic root water uptake (RWU), and a fine model spatial resolution (< 5 km). To address this, we present DynaRoot, a dynamic RWU scheme implemented within the Noah-MultiParameterization (Noah-MP) land surface model, a widely used tool for studying kilometer-scale regional land surface processes. Our modifications to Noah-MP include implementation of DynaRoot, eight additional resolved soil layers reaching a depth of 20 m, and soil properties that vary with depth. DynaRoot is computationally efficient and ideal for regional- or continental-scale climate simulations. We perform four 20-year uncoupled Noah-MP experiments for a region in the southern Amazon basin. Each experiment incrementally adds physical complexity. The experiments include default Noah-MP with free drainage (FD); a case with an activated groundwater scheme that resolves water table variations (GW); a case with eight added soil layers and soil properties that vary with depth (SOIL); and a case with DynaRoot activated (ROOT). Our results show the following: -DynaRoot allows mature forests in upland regions to avoid water stress during dry periods by taking up moisture from the deep vadose zone (where antecedent precipitation is still draining downward). -Valley vegetation can take up moisture from groundwater (while remaining constrained by the water table). -Temporally, we capture a seasonal shift in RWU from shallower soil layers in the wet season to deeper layers in the dry season, particularly over regions with dominant evergreen broadleaf (forest) vegetation. -Compared to the control (FD) case, there is a domain-average increase in transpiration of about 29% during dry months in the ROOT experiment. -Critically, the ROOT experiment performs best in simulating the temporal evolution of dry-season transpiration and evapotranspiration (ET), using a satellite-based ET product as an observational reference. Validation of modeled root depth is challenging due to a lack of in-situ observations. To address this, we describe a method for estimating root depth during drydown periods using observational products. These products include remotely sensed vegetation optical depth (VOD), precipitation, and enhanced vegetation index (EVI), as well as ET from a synthesized eddy covariance product. We compare rooting depth estimates from this method (ZVOD) with modeled rooting depth from the Noah-MP ROOT case (ZNMP) for 20-, 40-, and 60-day drydown events. We analyze results for the entire model domain as well as for two smaller subdomains: one with mostly forested grid points, and the other with mostly non-forested (savanna) grid points. We find that there is general agreement between the two methods for both subdomains. We also find the following: -Forest VOD and ET remain mostly constant throughout drydown events, while savanna VOD and ET decline. However, this decline is steeper in Noah-MP. -Modeled ET values are higher than shown by observational products. Median forest (savanna) ET is 2.5-3.0 (1.25-2.3) mm/day according to the observations-based method; according to Noah-MP it is 2.9-3.45 (1.0-3.0) mm/day. -Median ZVOD and ZNMP during drydown events differ slightly between the observations-based method and Noah-MP. Median ZVOD (ZNMP) is 10-15 m (10 m) for the forest subdomain and 1-3 m (5 m) for the savanna subdomain. -For the full domain, ZNMP is more influenced by water table variations, while ZVOD is more influenced by vegetation type. In addition to validation of the Noah-MP results, the observations-based method provides a measure of rooting depth that can be used when in-situ observations are not available. Future work should focus on further improvements to Noah-MP and application of the observations-based method to a larger domain. Moreover, it is important to explore effects of the DynaRoot uptake scheme on atmospheric variables in a coupled modeling framework
Fundamental limits and algorithms for sequence reconstruction, alignment, and analysis
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Kayvon Mazooji, accepted the attached license on 2025-04-26 at 19:32.The student, Kayvon Mazooji, submitted this Dissertation for approval on 2025-04-28 at 02:57.This Dissertation was approved for publication on 2025-05-01 at 11:22.DSpace SAF Submission Ingestion Package generated from Vireo submission #22011 on 2025-10-19 at 18:19:21This dissertation studies several problems concerning the reconstruction, alignment, and analysis of sequences. The results obtained are all motivated by or applicable to problems involving biological sequences, e.g. DNA, RNA, and proteins. These applications include privacy-preserving DNA sequencing, the reconstruction of proteins from tandem mass spectrometry data, DNA data storage, phylogeny inference, protein structure and function inference, and the clustering of cells into cell-types based on gene expression. The specific problems studied concern trace reconstruction, multiple sequence alignment, density-based clustering, flow graph decomposition, and information-theoretic privacy. The contributions in this dissertation are theoretical and algorithmic in nature, but applications to real datasets are included in some cases
A∞-algebras on directed graphs
Get PDFSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Omar Musleh, accepted the attached license on 2025-04-28 at 03:26.The student, Omar Musleh, submitted this Dissertation for approval on 2025-04-28 at 19:46.This Dissertation was approved for publication on 2025-04-30 at 12:29.DSpace SAF Submission Ingestion Package generated from Vireo submission #22019 on 2025-10-19 at 18:19:24Consider a directed graph Γ and a field k. We construct an associative algebra A(Γ) over k from Γ, whose elements are linear combinations of directed edges, where the multiplication is defined to be zero. The algebra on A(Γ) can be equipped with a graded structure by assigning integer degrees to the edges of Γ. We are particularly interested in the Hochschild cohomology CH∗(A(Γ)) of this graded algebra. Our primary focus is the A∞-algebras on A(Γ), which emerge as solutions to the Maurer-Cartan equation in CH∗(A(Γ)). Under this structure, the product of edges corresponds to their concatenation, though degree constraints on the edges significantly limit the number of such non-trivial structures. The second chapter develops the A∞-algebraic structures on directed graphs. This begins with a discussion of Hochschild cohomology on ungraded algebras, which then extends to graded ones. A∞-algebras are subsequently defined as degree 2 Hochschild cochains that satisfy the Maurer-Cartan equation. Through deformed Hochschild cohomology, an A∞-algebraic structure is then constructed for directed graphs. Later sections introduce tools and notations essential for simplifying work on these structures. The third chapter explores Directed Acyclic Graphs (DAGs), defined as graphs lacking loops. The absence of loops ensures that the Hochschild cochains associated with these graphs are finite-dimensional, which ensures that the MC equation becomes trivial beyond a certain degree. This chapter investigates such graphs, exploring degree assignments that maximize the number of products possible on them, referred to as MCAGs. Additionally, it introduces methods for visualizing and classifying the A∞-algebraic structures, documenting computational results derived using Wolfram Mathematica for these graphs with 5 and 6 vertices. The fourth chapter suggests various possible extensions to this research, revolving around discovering patterns and defining graph operations. The appendix provides the Mathematica file containing the code used to obtain computational results
Ultrasound-based functionalized biopsy markers for lesion tracking through neoadjuvant systemic therapy
No full textSubmission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2025-10-19 without embargo termsThe student, Jenna Cario, accepted the attached license on 2025-04-28 at 11:17.The student, Jenna Cario, submitted this Dissertation for approval on 2025-04-28 at 11:30.This Dissertation was approved for publication on 2025-04-28 at 15:38.DSpace SAF Submission Ingestion Package generated from Vireo submission #22029 on 2025-10-19 at 18:19:25As part of the standard care for breast cancer patients undergoing neoadjuvant systemic therapies (NST), a small biopsy marker, or clip, is inserted into or near the area being targeted for treatment. This strategy allows treated areas to be located more consistently after NST, despite any resulting morphological changes, than if no markers were used. However, challenges remain with the use of ultrasound to image these markers. Medical imaging applications, such as those used in a clinic or during surgery, demand consistency and reliability in a variety of environments and adverse conditions. Versatility and robustness of design are tantamount, which applies to both the markers themselves and the signals they return. Ultrasound is a highly attractive option in environments where real-time visualization of the body habitus is needed, such as during clip placement guidance and preoperative localization. It is a clinical mainstay and among the most common imaging modalities. It does not use ionizing radiation as X-ray or computed tomography (CT) scans do, and systems are portable and low-cost to acquire and maintain. Patients can be imaged at bedside, and there are no material compatibility concerns as exist with magnetic substances and magnetic resonance imaging (MRI) machines. Ultrasound probes consist of small piezoelectric elements that store electrical voltage as tension in their crystal lattice structures, and release this tension as high-frequency mechanical vibrations. Conversely, mechanical vibrations incident on the elements induce an electrical voltage. In this way, an acoustic pulse can be sent out from these elements, bounce off of structures in range of the pulse, and be received as echoes whose intensity directly correlates to a voltage magnitude. One of the most common configurations for clinical ultrasound probes is an array of uniform piezoelectric elements arranged in a single line, where the dimensions of the elements themselves will influence image width and lateral resolution. Ultrasound images tend to be leveraged in a qualitative manner, and radiofrequency (RF) data on the returned acoustic echoes in the imaged area must be beamformed to reconstruct an image that can be evaluated by a human observer. A beamformed ultrasound image is a compression of underlying numerical data, which can contain visually imperceptible patterns. In quantitative ultrasound (QUS), RF data is processed with numerical and statistical learning methods in order to ascertain information about the imaged environment that is not visible in an ultrasound image. QUS data can reveal structures associated with disease or malignancy even in the absence of visual indicators. We are currently evaluating QUS as a technique to identify early the response of breast cancer to NST. To use QUS, calibration is needed with a reference material in order to compensate for effects inherent to a given imaging system and settings. Doing so isolates the aspects of the numerical data that are inherent to the imaged medium from those inherent to the imaging system, and provides a method of standardizing measurements across multiple setups, systems, and institutions. One point of focus in this thesis is the evaluation of a spherical titanium bead that can act both as an in situ calibration target and a biopsy clip or marker. Using a reference which is adjacent to and at the same depth as an area of interest means the calibration and target experience very similar overlying conditions. This further isolates properties of interest, such as scattering, from other tissue effects such as attenuation. In a simple imaging environment with a spherical calibration target, the interplay of subtle positional shifts against physical probe limitations such as array element spacing can complicate the efficient recovery of a consistent signal, which is a requirement for accurate QUS measurements. Even without considering the restrictions that must be imposed in order to obtain accurate QUS estimates, ultrasound probe orientation and position relative to a small target such as a biopsy marker play an outsized role in successful visualization. Commercial off-the-shelf marker designs come in many non-spherical shapes designed to fit within a biopsy needle barrel for ease of deployment and patient comfort. These asymmetries create even greater variability in conspicuity relative to imaging orientation, which can complicate successful localization. Exploring how positioning affects these commercially-available clips is another feature point in this thesis, and significant differences in conspicuity with respect to viewing angle arise. This leads to a consideration of how electronics engineering may be used in order to improve ultrasonic clip conspicuity over what is achievable with passive marker design. By making the markers into active electronic devices, which transmit a specialized acoustic signal back to the imaging probe, we can decouple the clip response from the magnitude of the pulse used to image and potentially also from the orientation of the probe relative to the clip. So long as the imaging pulse is strong enough to activate the marker, the clip can send out a signal that encodes information such as an identification number or sensor readout. The signal will have its own inherent visibility in an ultrasound image, increasing the likelihood of successful localization. Any test platform designed to prototype such a device should capture as many of the design requirements for the final hardware, communication channel, and received signal processing as possible. Power, timing, and size considerations, as well as processing capability of the imaging system and the desire to obtain real-time results, introduce engineering tradeoffs in all parts of the system. These design challenges are explored primarily through experiment where possible, and steps to propagate the design through later stages into clinical translation are provided. This thesis outlines multiple strategies for increasing the functionality of radiological clips by leveraging already-common imaging equipment. It ultimately presents an electronic system, a signal design, and the corresponding processing that can meet or be readily adapted to enable communication with an ultrasound imaging system from an implanted device. With the design principles proposed herein, a standalone electronic clip design can progress toward a scale prototype
Embodying the Americas through Antigone and Medea
No full textSubmission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2027-05-01The student, Mary Casey, accepted the attached license on 2025-04-09 at 15:12.The student, Mary Casey, submitted this Dissertation for approval on 2025-04-09 at 15:19.This Dissertation was approved for publication on 2025-04-10 at 13:20.DSpace SAF Submission Ingestion Package generated from Vireo submission #21731 on 2025-10-19 at 19:14:34This dissertation traces adaptations of the Greek tragedies Antigone and Medea across the American hemisphere, spanning two hundred years of cultural production stretching from the Southern Cone to Canada. By beginning with a history of the Antigone and Medea’s prolific circulation across Latin America, this research situates hemispheric injustices spanning over five hundred years in the context of a longstanding cultural tradition in order to understand the prevalence and pervasiveness of such injustices across time and space. By anchoring this hemispheric study in the Latin American trajectory of Antigone and Medea, the U.S. and Canadian adaptations become part of a broader trend that can be better understood through this contextualization. Following this diachronic study, this dissertation turns to three particular social injustices plaguing the contemporary neoliberal state across the entire American hemisphere: immigration and anti-immigrant sentiment, racial and gender violence and the state’s culpability and complicity in such violence, and women’s bodily autonomy and reproductive rights. Antigone and Medea tell different stories under each of these categories, but both have been called upon across time and space as empowered figures of resistance
Evaluating functional performance of evapotranspiration models based on causal discovery methods
No full textSubmission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2027-05-01The student, Jiaze Cao, accepted the attached license on 2025-04-15 at 13:30.The student, Jiaze Cao, submitted this Thesis for approval on 2025-04-15 at 13:40.This Thesis was approved for publication on 2025-04-17 at 14:48.DSpace SAF Submission Ingestion Package generated from Vireo submission #21766 on 2025-10-19 at 19:14:42Investigating causal relationships is critical in ecohydrological systems to understand how variables interact, anticipate future trends, and develop effective models. Traditional predictive performance measures, which compare observations to modeled outputs, cannot fully verify generalization ability and do not address representations of causal linkages between inputs, observations, and outputs. Functional performance focuses on relationships between the target and source variables, but similar to predictive performance metrics, there are multiple ways to quantify these relationships. We compare four causal discovery methods (Granger causality, Transfer Entropy, PCMCI, and Convergent Cross Mapping) to analyze the functional performance of models of evapotranspiration (ET) in an intensively managed agricultural landscape based on eddy covariance measurements over seven years. Evapotranspiration is a critical component of both the energy and water cycles, playing a key role in water resource management and irrigation planning. We first compare methods with linear, nonlinear, and chaotic test cases to evaluate how causal discovery varies with system dynamics, temporal aggregation, and noise. We find that the four methods identify accurate causal relationships to a large extent, but temporal aggregation, and particularly data resampling, can lead to diverging results. Subsequently, we apply the methods to analyze causal sources within three ET models: Priestly-Taylor (PT), Surface Flux Equilibrium (SFE), and Soil Water Balance (SWB) relative to tower observations. The causal discovery methods consistently identify model inputs as primary causal sources, such as net radiation, relative humidity, and temperature for PT and SFE, and soil moisture for SWB. In a functional performance assessment, we rely on the causal strength measures derived from Transfer Entropy and Granger Causality. The functional performance metric offers improved interpretability and flexibility, enabling model evaluation across varying temporal intervals. Among the evaluated models, the functional performance aligns closely with predictive performance, identifying SFE as the best-performing model. We further apply functional performance metrics to evaluate the OpenET dataset, which contains multiple satellite-derived ET products. Although OpenET models rely primarily on satellite-based inputs, they exhibit higher functional performance, indicating a more accurate representation of evapotranspiration processes. The analysis reveals that functional performance metrics offer unique insights into model behavior, aligning with predictive performance, especially when substantial performance differences exist. This study shows how functional performance complements traditional model evaluations, but also highlights sensitivities to different causal inference methods and time scales