Enhancing the T-shaped learning profile when teaching hydrology using data, modeling, and visualization activities

Abstract. Previous research has suggested that the use of more authentic learning activities can produce more robust and durable knowledge gains. This is consistent with calls within civil engineering education, specifically hydrology, that suggest that curricula should more often include professional perspective and data analysis skills to better develop the "T-shaped" knowledge profile of a professional hydrologist (i.e., professional breadth combined with technical depth). It was expected that the inclusion of a data-driven simulation lab exercise that was contextualized within a real-world situation and more consistent with the job duties of a professional in the field, would provide enhanced learning and appreciation of job duties beyond more conventional paper-and-pencil exercises in a lower-division undergraduate course. Results indicate that while students learned in both conditions, learning was enhanced for the data-driven simulation group in nearly every content area. This pattern of results suggests that the use of data-driven modeling and visualization activities can have a significant positive impact on instruction. This increase in learning likely facilitates the development of student perspective and conceptual mastery, enabling students to make better choices about their studies, while also better preparing them for work as a professional in the field.


Introduction 1
While there is a rising interest in and demand for civil engineering and hydrology education, 2 some have suggested a widening gap between how students are instructed in hydrology, and 3 the subsequent professional skill set required for a career as a hydrological engineer (Wagener 4 et al., 2007). Recent research has shown a potential for great variability within the 5 hydrological curriculum (Wagener et al., 2012). This variability includes differences in not 6 only what conceptual material should be taught (Gleeson, Allen & Ferguson, 2012), but also 7 how this material should be delivered pedagogically (Wagener, 2007). It has been suggested 8 that an emerging requirement for new hydrological engineers is the ability to not only develop 9 a well-defined knowledgebase of basic hydrological concepts, but also synthesize this 10 conceptual learning with more authentic 'real-world' knowledge gained from the 11 interpretation and application of this knowledge (Merwade and Ruddell, 2012). 12 Unfortunately, field and modeling activities are often lacking in the hydrological curriculum, 13 at least at the undergraduate and lower division level (ASCE, 1990;MacDonald, 1993;Nash 14 et al., 1990;Ruddell & Wagener, 2013;Wagener et al., 2007Wagener et al., , 2012. This is especially 15 concerning as unlike laboratory sciences such as physics and chemistry, hydrology is 16 fundamentally a place-based science. It can therefore be argued that hydrologists must 17 engage in field and modeling activities in order to fully develop the critical ability to link 18 hydrological concepts to applications in a specific place/instance (Eagleson et al., 1991). 19 This call to integrate experiential learning with traditional classroom instruction is not new , 20 and has been advocated in other fields of engineering (Duderstadt, 2007;Lattuca, Terenzini, 21 Volkwein & Peterson, 2006; National Academy of Science, 2007;Shulman, 2005), and has 22 also been suggested more generally within the educational literature (e.g., Bransford, Brown 23 & Cocking, 1999;Brown, Collins & Duguid, 1989). These suggestions are rooted in the 24 simple tenet that when learners engage more deeply in the formation and development of 25 relevant knowledge, the depth and quality of their understanding subsequently increases. This 26 constructive process is integral to numerous pedagogical philosophies such as problem-based 27 learning (PBL), guided discovery learning, and cognitive apprenticeship, to name a few 28 (Alfieri et al., 2011;Brown et al., 1989;Collins, 1991;DeJong & Van Joolingen, 1998;Duch 29 et al., 2001;Savery & Duffy, 1995;Wood, 2003). 30 While the various educational pedagogies mentioned above are different on several levels, 31 they share at least two important unifying characteristics. Fundamentally, (1) they require the 32 learner to be actively engaged in the learning activity in order to realize any learning benefit, 1 and (2) they are usually situated within an authentic or 'real' problem that the student must 2 work to solve or address. Importantly, these characteristics imply that the problem is difficult 3 enough that students must work towards a solution (i.e., they do not know the solution 4 initially), and that each student has explicit engagement with the pursuit of this solution, as 5 such activities are often implemented in group settings (Smith, Sheppard, Johnson & Johnson, 6 2005). It has been argued that such authentic engagement fosters a more deep conceptual 7 understanding of the material by 'anchoring' the more abstract learning material or concepts 8 to the more accessible authentic learning scenario (CTGV, 1992;Hake, 1998). Thus, the 9 contextualization of the material within an actual scenario increases not only retrieval cues 10 that the learner can use to more efficiently access factual knowledge, but also likely increases 11 the durability of the knowledgebase, thereby creating a more flexible state of information that 12 could be applied appropriately in multiple instances (Hansen, 2008;Smith & Van Doren, 13 2004). 14 Active engagement in the learning process has also been suggested as a means to increase 15 interest in the topic to-be-learned (Paris & Turner, 1994; Scheifele, 1991), which might also 16 address issues of motivation within students. Traditional lecture-based instruction that forces 17 students to work towards normative educational goals in isolation is often cited as a major 18 complaint of engineering students, and has measurable negative effects on motivation levels 19 (Felder, Felder & Dietz, 1998). More authentic, problem-based activity has been shown to 20 produce an increase in student attitudes towards the content area in general (Watters & Ginns, 21 2000), offering an opportunity to offset such motivation issues. Importantly, this could not 22 only increase motivation within the lesson itself, but also potentially affect the likelihood to 23 continue with studies in a given domain. In other words, this motivation derived within a 24 specific context could have a direct effect on overall interest in the major or field, as learners 25 are better able to see how their own interests better align and apply to tangible problems. 26 However, efforts to adopt such authentic learning exercises within engineering education are 27 often hampered by unclear learning objectives and assessment, logistical constraints, and the 28 use of activities that do not necessarily optimize the learning experience (see Prince, 2004). 29 For example, it is unclear about what degree of 'authenticity' is required, and how does one 30 assess learning from 'field' activities relative to traditional instruction? For example, while 31 PBL has been implemented successfully with electrical engineering students (Yadav,Subedi,32 Lundeberg & Bunting, 2011), students who engaged with the PBL activity were compared to 1 students who only had a lecture component, without the opportunity to engage in an 2 equivalent control activity. As such, these studies cannot conclusively say that gains 3 normally attributed to the instructional manipulation are due to the activity alone, and could 4 reflect the influence of other factors (e.g., differences in time spent engaging with the 5 material). Further, what is an appropriate 'field activity' in an engineering discipline, and 6 how should these efforts be categorized and defined? As such, while this call for authentic 7 activity is often advocated and supported theoretically, unfortunately it is not often 8 consistently practiced, and thus leads to fragmented research on the issue (Prince, 2004). 9 There do also exist more specific pedagogical concerns regarding authentic learning within 10 the area of hydrological engineering education (Gleeson et al., 2012). For example, there is 11 little to no direct evidence that such activities are indeed effective at augmenting a 12 hydrologist's training, or even implemented with any kind of regularity for that matter 13 (Ruddell and Wagener, 2013). What little evidence that does exist supporting the 14 incorporation of student-centered activities into hydrology instruction is often anecdotal (e.g., 15 Thompson et al., 2012), without any kind of quantitative or measureable change in 16 performance outcomes. Pragmatic and logistical issues (e.g., faculty time and expertise, 17 student computer skills), and the use of curriculum materials that become rapidly outdated, 18 also stand as barriers to the adoption of a more discovery-based or student-centered approach 19 within hydrology (Merwade & Ruddell, 2012, Ruddell & Wagener, 2013). Finally, 20 hydrological instruction is also traditionally implemented using a teacher-centered approach 21 (e.g., lectures) that lacks the opportunity for applied experience (Wagener et al., 2007). Thus, 22 it appears critical to find new ways to achieve instructional goals that might incorporate this 23 real world experience, and are capable of side-stepping these methodological and logistical 24 issues. Fortunately, the emergence of rich and dynamic computer simulation techniques, 25 which allow students to interact with real data in ways that are consistent and appropriate with 26 the profession, might offer an alternative to such traditional instruction, and thus provide an 27 exciting opportunity for students to achieve this more authentic application of knowledge. 28

Data modeling driven geoscience cybereducation 29
Standardized data and modeling driven geoscience cybereducation (DMDGC) modules, 30 developed and published by a dedicated community of educators, do potentially provide 31 access to such dynamic and realistic learning experiences, while also avoiding some of the 32 logistical barriers mentioned above (Habib et al., 2012; Merwade and Ruddell, 2012). These 1 modules utilize contextually specific, rich, and dynamic computer simulations that allow 2 students to interact with current field data in a fashion equivalent to professional hydrologists. 3 As students do not have to physically travel to a work site to collect data, nor do they require 4 specialized tools to work with the data, these simulation activities can be easily integrated into 5 normal laboratory sections via coursework. Most importantly, as the data is real, and also 6 contextualized within a specific example, it presents an opportunity to apply hydrological 7 concepts within a formally structured and valid situation, again consistent with professionals 8 in the field. 9 An open question, however, is whether such activities do in fact realize the potential 10 educational benefits that one might anticipate from authentic activities? Similarly, it is not 11 known for which content areas/aspects of the curriculum are such benefits localized or 12 strongest, if any? For example, do such activities help students better appreciate what it 13 means to be a professional hydrological engineer? Or is this benefit localized to better 14 understanding hydrological domain content alone? In other words, it must be evaluated 15 whether DMDGC modules do in fact serve as an adequate opportunity to gain such authentic 16 activity, while also permitting the learning and achievement of traditional class goals for 17 knowledge attainment? It has therefore become important to pinpoint the learning benefits 18 created by DMDGC activities, so that these activities can be optimized for content, structure, 19 and integration with the traditional lecture format. 20 This study directly examines the efficacy of such data-driven simulations for hydrology 21 education at the earliest point in a potential future hydrologist's university training: in a 22 mandatory lower-division undergraduate earth science context that is part of general 23 curriculum studies. At this level the student enters the classroom with very little (if any) prior 24 knowledge about hydrologic theory, hydrology models/methods, or the broad applications and 25 societal issues involved with hydrological engineering. Evaluating the effects of such an 26 intervention at this very early point provides an opportunity to examine the full effect of 27 DMDGC implementation, avoiding issues of self-selection bias and prior contextual 28 knowledge about the hydrology profession that might exist in upper division or graduate 29 students in the field. In other words, in this student population we can observe the effect of 30 DMDGC activities on a breadth of knowledge related to the field and its application, beyond 31 just core theoretical concepts and applied computer modeling skills. 32 Students in the DMDGC condition were given a data-driven hydrology activity that focused 1 on urbanization and flooding, while a control group was given a paper-pencil based laboratory 2 activity of equivalent general learning outcomes and effort, but lacked the specific applied 3 context and data-driven components of the DMDGC. The inclusion of this paper-based 4 activity is a critical methodological feature, as it permits a more appropriate evaluation of the 5 simulation activity against an activity that is likely equally effortful and time consuming from 6 the student perspective. Thus, any subsequent learning gains are less likely to be attributed to 7 other confounding factors, and must instead be more localized to the nature of the 8 manipulation itself. It was hypothesized that students who were presented with the DMDGC 9 learning activities would demonstrate a better understanding of theoretical and applied 10 hydrology concepts related to flooding, as their interaction with the material would be 11 contextualized and likewise permit a dynamic exploration of the data not otherwise possible 12 without such simulation. Further, it was also hoped that students in the simulation condition 13 would develop a better appreciation for the roles of hydrological engineers and hydrology 14 organizations in managing and preventing flooding problems, as they themselves are 15 engaging in a contextualized problem within a realistic community scenario that required the 16 intervention of hydrologists. 17 2 Methods 18

Participants and experimental design 19
One-hundred seven students (N=107) enrolled in an Introductory Earth Science course (and 20 corresponding laboratory sections) at a community college in the south-western United States 21 were solicited for participation. Participants were evaluated both before and after a sub-unit 22 within the course that focused on applying the Rational Method and a Synthetic Unit 23 Hydrograph to estimate hydrographs and flooding for urban areas experiencing land use and 24 climate change. Eighty-eight of these students successfully participated in both the pre and 25 post assessments, an overall completion rate of 82%. These 88 participants were distributed 26 among 2 different instructional conditions based on enrolled lab section: DMDGC modeling 27 (n=52; 79% participation rate), and paper-based activities (n=36; 88% participation rate). All 28 students shared the same single lecture instructor, and were thus given identical lecture 29 content over a period of approximately two weeks of class.

Curricular materials 2
Both the DMDGC modules (Ruddell and Schiesser, 2012a;b) and the comparable paper 3 laboratory (Lab 9 in Schiesser, 2008) were designed to be implemented in parallel with 4 traditional lectures. In this unit, all students were given identical lectures (based on material 5 covered in Schiesser, 2008) that covered the fundamentals of flood frequency, urbanization 6 and land use change, flood risk, climate change effects on rainfall, and the roles and 7 responsibilities of agencies that provide flood prediction and management services in the 8 USA. In other words, the lecture component of the current design was identical for both 9 laboratory groups, and the only instructional difference was whether the students received a 10 DMDGC or paper laboratory module. 11 The experimental DMDGC module is written for Microsoft Excel™, a widely utilized and 12 highly accessible spreadsheet application. It is a simple stormwater hydrograph modeling 13 module that applies the widely utilized Rational Method and a Synthetic Unit Hydrograph to 14 estimate hydrographs and flooding for urban watersheds. The model is based on assumptions 15 optimized for a floodway in Maricopa County, Arizona; an urbanized desert area in the 16 southwestern United States. The module has the ability to accept both observed rainfall and 17 streamflow data so that a student may calibrate the parameters of the flood model to match 18 any observed event. Importantly, the module is also broadly applicable to urbanizing 19 watersheds anywhere in the world and can be adapted to other locations by simply adjusting a 20 few model parameters and obtaining observed streamflow data for a flood event. As such, this 21 DMDGC activity could be applied to nearly any urban area, an option that could be used to 22 tailor context and content respective to each student population and their corresponding 23 physical location. 24 The DMDGC module produces a visualization of modeled and observed hydrograph results 25 ( Figure 1). As is visible in Figure 1, the module emphasizes the determination of whether or 26 not a given channel will flood during a 100-year design storm event as land use is 27 progressively urbanized, and as the design storm changes due to climate change. These 28 multiple interacting characteristics served as the foundation for the rubric described in the 29 next section (and in Figure 2). The DMDGC module takes roughly two hours of preparation 1 control for comparison with the DMDGC module. The paper module requires students to 2 perform hand calculations and determine whether a channel will flood before and after 3 urbanization occurs in a watershed. Like the DMDGC module, a student considers the effect 4 of issues such as rainfall infiltration, watershed area, and rainfall intensity, and channel 5 capacity in determining a flood. Unlike the DMDGC module, the paper activity explicitly 6 addresses issues of flood frequency using recurrence interval calculations using a brief table  7 of historical peak flow events instead of a student's investigation of observed streamflow 8 data. Also, no visualization or interaction is possible with the paper method. The paper lab's 9 streamflow data is "stock" data that is hypothetical and not drawn from real-world or place-10 based sources. The paper module does not include customized data for the student's local 11 watershed, nor an observed rainfall event in the local watershed, and is not able to provide 12 visual feedback via the flow hydrograph when the student adjusts watershed parameters or the 13 rainfall intensity. Finally, this paper-based activity is not contextualized within the local 14 environment (e.g., Maricopa county). The estimated time to complete this paper activity is 15 also approximately 3 hours. Thus, this exercise requires students to complete calculations of 16 similar complexity and type as the DMDGC module, albeit in a paper and pencil form and 17 minus the place-based contextualization and interactive visualization components. As such, 18 effort and time with the material (across both lecture and laboratory components) are 19 comparable across these instructional conditions, and not likely explanations for any 20 subsequent effects. 21

Learning Assessments 22
The pre/post assessment instrument (Appendix A) features eight questions spanning a range 23 of topics. Two of these 8 questions contained sub-questions, thus resulting in a total of eleven 24 questions overall. To provide a more coherent evaluation of performance in the learning of 25 hydrological concepts and the role of hydrologists, a rubric was developed resulting in nine 26 overall learning outcomes representing important hydrology concepts related to flooding. The 27 nine outcome areas presented in Table 2 represent areas of conceptual mastery regarding 28 climate, land cover, flood management, and hydrology. The first 6 areas specifically 29 emphasize mastery of the physical concepts determining flooding (e.g., rainfall intensity and 30 duration, hydrographs, infiltration, and stormwater management practices), and thus are 31 indicative of a good conceptual understanding of the material itself. However, the last 3 32 outcomes were designed to assess the understanding of the roles and responsibilities of flood-1 related professional agencies (e.g., agency roles & responsibilities, value of geoscience 2 knowledge), or in other words, the potential job duties of a professional working in the field. 3 To make this distinction more transparent, examples of conceptual mastery relative to the 4 learning outcomes are also presented in Table 2. Together, these nine outcomes cover the Two hydrology educators independently coded the level of conceptual mastery indicated by 16 student responses on the pretest assessment instrument, blind to condition, and indicated a 17 high degree of inter-rater reliability across all nine learning outcomes (all ICCs>.91, p<.01), . 18 The post-assessments (which were again identical to the pre-assessments) were then coded by 19 a single coder. Table 1 gives examples of conceptual mastery for each of the nine outcomes. 20

Results 21
To examine the effect of the DMDGC modules on the change in student knowledge in each of 22 the nine outcomes, a simple 2-way ANCOVA was conducted between laboratory groups on 23 the post-test scores for each outcome. Pretest scores for each measure were used as a 24 covariate in every respective analysis to control for any differences in initial knowledge 25 levels, and all results were evaluated for significance at the level of p < 0.05. Levene's tests 26 for all analyses indicated a non-significant result (p > .05), which affirms that variance was 27 equivalent between comparison groups. Descriptive statistics for each measure by group, and

Physical concepts of flooding 1
As is visible in Table 3, the use of a DMDGC module significantly improved performance in 2 all 6 areas (#1-6) save (#4) Effect of Decadal LULC Change on Flooding. Participants who 3 were given the opportunity to learn with the DMDGC modules were better able to not only 4 understand the effects of urbanization and other physical causes of flooding, but also 5 demonstrated better knowledge of maximum discharge rates and impacts of flood 6 management. The lack of result for outcome (#4) Effect of Decadal LULC Change on 7 Flooding was not entirely unexpected, as although this content topic was originally intended 8 to be emphasized in the lecture and lab settings, unfortunately it was not able to be covered in 9 depth due to time constraints. Thus, it is not surprising that this outcome showed little 10 divergence between groups as students were not explicitly instructed in this topic. As such, 11 this likely reflects a shortcoming in the overall content covered, rather than demonstrating a 12 lack of theoretical effect. 13

Professional role of hydrologists 14
Consistent with the content results above, users of DMDGC modules also appear to have 15 gained a better appreciation for the professional role of hydrologists and the field. Across all 16 3 sub-areas (#s 6-9), there was a significantly higher demonstration of expertise for the 17 simulation group, above those simply using the paper-and-pencil activities. This suggests that 18 not only does engaging in such authentic activity produce a measureable benefit in learning 19 content, but this benefit also results in a better understanding of the professional duties within 20 the field. This result is especially encouraging as it also could potentially indicate that such 21 activities allow students to become better prepared for eventual careers in hydrological 22 engineering, and thus provide a bridge between the content area and the application of 23 knowledge. 24 In summary, when one considers the overall pattern of results it appears that the benefit for 25 such dynamic simulation and visualization was not only limited to content knowledge areas 26 such as rainfall intensity and flooding, but was also realized in regards to better understanding 27 the professional and social impacts of hydrology. This suggests that not only did learners 28 better understand the material itself, but also better understood the role of hydrologists in a 29 more general sense. Further, the medium to large effect sizes (Miles & Shevlin, 2001) 30 realized by this manipulation further suggest that the inclusion of the DMDGC module produced a practical and worthwhile change in performance, above and beyond reaching 1 simple statistical reliability. 2

Discussion 3
While prior research in education has suggested that the use of applied examples could likely 4 benefit learning, this suggestion was explicitly tested here in the context of hydrological 5 education, using DMDGC modules. It was anticipated that the use of such dynamic and 6 flexible simulation tools, which enable learners to contextualize and visualize the impact of 7 minute changes in data over time, would lead to a marked increase in learning performance. 8 The results of this classroom study support exactly that. Learners who were permitted to 9 interact with such simulations not only were better able to understand the content itself in the 10 form of general knowledge, but these same learners were also better able to appreciate the 11 role of professionals within the field. This increase was significantly larger than that 12 experienced by the control group, which engaged with materials that required similar skills 13 but lacked the contextual and simulation components of the DMDGC module. It is our 14 contention that the increase in both areas (breadth and depth) was a direct result of the 15 experience with the DMDGC module. For example, in terms of conceptual learning, the 16 DMDGC modules allowed learners to better understand the interaction of conceptual units 17 and how to use tools like the hydrograph to anticipate flooding conditions. Similarly, this 18 direct experience also allowed learners to better appreciate the job duties of practicing 19 hydrologists, providing a tacit understanding of the role of agencies and geoscience education 20 in society, which in turn led to better recall. While certainly speculative, given that both 21 groups received identical discussion regarding agency duties in lecture (and thus in a 22 decontextualized, abstract sense), the fact that the DMDGC group was able to better 23 appreciate this kind of information seems to again suggest that the concrete experience helped 24 make this understanding of professional duties more accessible to these learners. 25 As such, this overall pattern of results suggests that learners were gaining a more complete 26 'T-shaped profile' of hydrological education (Ruddell & Wagener, 2013), balancing an 27 increase in not only their specialized conceptual, quantitative, and modeling skills within the 28 field, but also achieving a more broad understanding of the role of professionals in the field 29 relative to real-world scenarios. This is a very encouraging result, as it suggests a dual benefit 30 for such DMDGC training.
Further, another interesting point is that it is likely such multiple effects were observed 1 because the participants in this study were just beginning their education in the field of 2 hydrology, so issues of contextualized knowledge or self-selection were likely minimized in 3 this sample. In other words, because learners were lacking a well-defined representation of 4 not only the knowledge of the field, but also the role of working professionals in the field, this 5 training experience permitted them to gain greater insight into both the field and requisite 6 application. This fact is even more encouraging as it suggests that such interventions, 7 introduced early in the educational trajectory, can provide a more robust and complete 8 learning experience at all levels. It is possible that such increases in depth and breadth of 9 knowledge early on could translate into more success with the material, thus likely increasing 10 the likelihood of learners persisting in the pursuit of education in this domain. It appears that 11 working with authentic data increases the appreciation of a novice student for the importance 12 of the hydrology profession and for the physical problems this profession addresses. 13

Conclusions 14
For the fundamentally place-based geosciences such as hydrology, the integration of concepts 15 will inevitably require exposure to real-world contexts and data. The results of this study 16 demonstrate that computerized learning content can effectively bring the 'real world' into the 17 classroom and make it accessible, especially in the case of students at lower levels and across 18 the general curriculum. The findings of this paper also indicate that it is possible to deliver 19 this type of content in a localized place-based context, and to realize learning gains on both 20 physical and professional learning outcomes without introducing a great deal of complexity in 21 the way of computer modeling and programming. A simple spreadsheet, combined with 22 readily available online hydrological data, is sufficient in this case. In other words, these 23 computerized techniques afford instructors the opportunity to have their students engage in 24 realistic and authentic problem-based activities without the need to manage other logistical 25 constraints often encountered with field research (i.e., transportation, materials, etc.). It is our 26 hope that the positive findings of this study encourage investment in development of high-27 quality DMDGC learning materials, and the wider adoption of place-based DMDGC learning 28 materials across the civil engineering curriculum. Implementing such learning experiences 29 into the curriculum will ideally create more enriching experiences for student learners, and 30 hopefully also develop more well-rounded and skilled practicing hydrologists.
While the current study focused on lower-division students, in future work it would also be of 1 interest to expand this program longitudinally throughout the curriculum to identify how best 2 to deliver DMDGC content at all levels of the hydrology curriculum to maximize its 3 effectiveness. Efforts are currently ongoing to do exactly this, and also expand the 4 application of DMDGC content to hydrological concepts beyond flooding and urbanization. The USGS is the best answer, but NOAA is a good second choice.
(4) What U.S. Federal agency is the primary provider of rainfall and weather data?
NOAA is the best answer, but the USGS is a good second choice.     Method and a triangular Synthetic Unit Hydrograph are applied to model a rainstorm's 16