Explore content

The City of Carson, California sought an efficient way to conduct storm water inspections at permitted businesses within City limits on a compressed schedule as part of their MS4 permit requirements. Partnering with GHD, the City was able to gain efficiencies in the following areas: Progress reporting QA/QC of field data during collection Linking field photos with inspection forms Digital data entry of previous hand-written field forms Scheduling follow-up inspections Tabular data upload to regulatory database The efficiencies translate into faster progress of over 1100 inspections, transparency on progress and data quality for all team members, and easier upload to compliance databases. Coupling these tools with GHD's comprehensive training on what to ask and look for in the field during inspections and what might indicate an illicit discharge or connection provided a comprehensive package and robust support team for the City of Carson. Purpose: We will be presenting the successful workflow using Esri's ArcGIS Online and mobile apps to show how the City accelerated a team of new storm water inspectors and new adopters of Esri mobile apps into an efficient and seamless program of storm water compliance. Custom scripting within pop-ups allowed inspectors to benefit from the map aspect of Collector and ease of data entry with Survey123. Dashboards and web apps in the Online environment allowed for concurrent review of field data in real time, and at-a-glance progress for program managers. Benefits of Presentation: This project is an excellent example of a ground-up program roll-out of first-time inspectors using Esri's ArcGIS Online and mobile tools. It demonstrates how quickly and easily storm water inspections and reporting can be prepared, packaged and delivered in a way to get a team started and inspections completed with maximum transparency. Status of Completion/Conclusions: At the time of submittal, the training, data preparation, and tool development described above is complete. Inspections have started and are scheduled to be completed and data uploaded to the regulatory database in the first quarter of 2022. We expect the project will be entirely complete, with time for reflection, before the June conference date. Originality: Esri offers many online tools that can be designed to fit the need of any project. We took this one step further to apply custom scripting within the mobile apps to leverage the advantages of both Collector and Survey123. The way we have combined the building blocks provided by Esri created a custom fit to this project and would fit any organization who is starting from square-one with their storm water inspections required through their NPDES permit, or who is looking for a more streamlined and repeatable approach to their inspection data management.

Municipalities often face difficult choices about how to allocate resources, where to build infrastructure, and with whom to partner to address complex stormwater management challenges. The City of Aiken in South Carolina is no exception as city leaders seek to restore predevelopment hydrology in the Sand River. The Sand River collects runoff from 1,109 acres of urban watershed in the City of Aiken and flows primarily through Hitchcock Woods, the largest privately-owned urban forest in the country. Hitchcock Woods and the Sand River are a haven for hikers and equestrians, and a driver of tourism in the City. What was once small enough for a person to step across, the Sand River has become a 70-ft deep canyon as a result of frequent and intense stormwater flows coming from downtown Aiken over the past 40 years [1,2]. In 2017, the newly-formed Sand River Stormwater Task Force recommended 20 capital improvement projects, sized to capture 58 ac-ft of stormwater, and costing approximately $22 million (2017 dollars) as a way to restore the river's predevelopment hydrology [3]. While this plan was based on the best available information at the time, recent developments in real-time control technology led the Task Force to reevaluate the recommended project list. In 2020, the Task Force recommended the use of forecast-based real-time control technology, also called continuous monitoring and adaptive control (CMAC), on the largest proposed stormwater project: a 24 ac-ft set of underground vaults (Figure 1). These vaults are positioned at the downstream end of Aiken's Sand River urban watershed, just upstream major erosion in Hitchcock Woods (Figure 2). CMAC was estimated to increase the effectiveness of the vaults by 2x to 3x by retaining water during and after storms, increasing infiltration, and only discharging in preparation for forecast storm events exceeding the capacity of the vaults [4]. To further quantify how the adaptively-controlled Sand River vaults perform compared to traditional passive underground storage, a modeling effort was undertaken for the period of calendar years 2001 through 2006, during which time Aiken received 295 inches of rainfall (average of 49 inches per year). The CMAC system was found to retain 71% of stormwater flow entering the site during wet weather periods, compared to only 26% wet weather capture achieved by passive storage. Modeling also showed the adaptive control vaults to exfiltrate 66% of all stormwater volume entering the system compared to 25% exfiltration by passive storage. This increase in performance will set a new baseline for stormwater flow in the Sand River, thereby mitigating the need for upstream stormwater projects and saving the City of Aiken millions of dollars in capital expenditures. This presentation will examine the decision to include CMAC on this project, discuss the design process, and present modeling results showing the performance improvement of the vaults with adaptive controls. Future work will include comparing the modeled performance to observed performance after the system is commissioned in 2022.

Since 2001, the Metropolitan North Georgia Water Planning District (District) has focused on water supply, wastewater, and watershed planning through an integrated Water Resources Management Plan (Plan) for the 15 counties and 95 cities that make up the Metro Atlanta Region. As part of the watershed planning efforts, the District has coordinated strategies to protect watershed conditions and manage stormwater in conjunction with existing regulatory requirements. However, despite these efforts, urban stormwater runoff remains a leading cause of nonpoint source pollution and flooding in the District, leaving watershed managers with ongoing challenges related to water quality, streambank erosion, and nuisance flooding. Historically, managers have focused on strategies to address water quality rather than water quantity, even though they are fundamentally linked. For these reasons, the District has developed a novel water quantity-based indicator, called the Stormwater Forecast (Forecast), as part of the integrated 2022 Water Resources Management Plan. The Forecast is a planning-level estimate of the total potential runoff management volume from development, calculated at the basin scale using site-scale post-construction stormwater performance standards. The calculation estimates runoff volume in cubic feet for three post-construction stormwater performance standards used in the District: Water Quality Volume (1.2-inch storm event), Channel Protection Volume (1-year, 24-hour storm event) and Overbank Flood Protection Volume (25-year, 24-hour storm event). The Forecast provides a planning-level estimate of runoff volume from all historical and existing development through 2019 (using the most recent USGS land cover data) and calculates future runoff volume through 2040.With this approach, the Forecast represents total potential runoff management volume from development with the potential to be managed by Stormwater Control Measures (e.g., stormwater ponds or bioretention basins) if current post-construction stormwater management standards were fully in place. The goals of the Forecast are to provide additional information for improved decision making at both the site-scale and basin-scale, to support a better understanding of overall needs, and to help generate new solutions and policies that address ongoing stormwater management challenges. Future use of the Forecast is intended to complement and support the implementation of existing water quality and flood reduction goals in the District. This presentation will focus on the basic framework of the Stormwater Forecast as well as an overview of the technical approach and future implementation.

Introduction: An open channel is a concrete waterway that is widely used in many regions. Like other concrete structures, open channels suffer from many deficiencies, such as scour, cracking, spalling, and tilting. Inspections need to be performed on a regular basis aimed at identifying and rating deficiencies. Ratings represent the severities of a deficiency, which are often associated with descriptions and suggested actions. Table 1 shows an example of spall rating classification. Manually classifying deficiency ratings is often difficult, since it requires both expertise and extensive physical inspection. Therefore, automating deficiency rating work has been an industry challenge. Machine learning (ML) is a type of artificial intelligence (AI) that can learn from various data, identify patterns, and make decisions. In this study, we are interested in evaluating the power of ML on providing accurate rating classifications of channel deficiencies without human intervention. The developed ML model is described in Methodology and followed by preliminary Model Results and Conclusion. Methodology: Since the goal is to rate defects or deficiencies, a supervised ML model is selected. Specially, we train a convolutional neural network (CNN) on 80% of the total available data, then validate & test model results on the remaining 20%. For each deficiency, the total available data consists of thousands of pre-labeled deficiency photos. Table 2 lists the total available photos for each test deficiency. Figure 1 provides a proposed modelling flowchart. Convolutional neural network (CNN) is one of the most popular neural networks in computer vision. Compared with traditional neural networks, CNN consists of not only fully connected layers, but also convolutional layers and pooling layers. This structure allows CNN to utilize spatial pixel interaction information as well as reduce model complexity, which makes CNN especially suitable for image classification. The convolutional layer is used for parsing the input photo. Three brightness intensity values (red, green, and blue, or 'RGB' channels) are assigned to each pixel on a color photo. Figure 2 gives an illustrative example using a tree photo. After parsing a photo into its RGB channels, the convolutional layer employs several 'filters' that contain trainable weights to apply convolution operations on these channels. These filters move over the input layer until all pixels have been covered. The dimensions of input layer will be reduced after the convolution. A pooling layer typically follows a convolutional layer, which further decreases the input layer dimension. The pooling layer also summarizes the regional pixels information so that the model becomes robust against objective position variations. Figure 3 demonstrates a CNN example that is used to predict if a photo is a tree, streetlamp, or a stop sign. In this study, we further adopt a 'transfer learning' technique: A pre-trained CNN, ResNet-50, is used and followed by a trainable fully connected layer to yield deficiency rating predictions. Our model is developed based on open-source software libraries: TensorFlow and Keras frameworks. Model Results: Results are briefly discussed in this abstract. Figure 4 show both the loss and accuracy curves against 70 epochs for cracking as an example, with class weight considered in training. As the training proceeds, in general, the validation loss is decreasing and the validation accuracy is increasing before reaching a relative stable condition, which indicates the convergence of the model. Figure 5 plots the confusion matrix of cracking based on the cracking test data. There are three ratings for cracking (R1, R2, and R3). Each row represents the model prediction while each column is the true label. For example, there are 333 photos with the true R2 labels that are also predicted as R2. Based on the confusion matrix, the model accuracy, precision, and recall are calculated. In this study, we select accuracy and macro-averaged F1 score as two evaluation metrics. Accuracy is defined as the number of correct predictions divided by the total number of predictions, and it assumes that all the classes are equally important. Macro-averaged F1 score is a harmonic mean (tradeoff) between precision and recall, which is useful when the false negatives and false positives are essential. For some deficiencies, rating classes can be imbalanced, where one rating class only has several photos while another rating class has thousands of photos. In this case, F1 score is a preferred over accuracy since it accounts for the model ability to predict the minority class. One way to increase F1 score is to incorporate class weight into loss function, where class weight is the inverse data ratio among different rating classes. Table 3 shows the class weight information for each deficiency. Based on the class weights, Table 4 shows the test results for all four deficiencies, where 60%-70% accuracy can be generally achieved. The accuracy is generally lowered by 2% in exchange for 0.02 increase of F1 score, when comparing the no-class-weight model to the class-weight model. Due to the limited data (Table 2), tilting has the least accuracy performance. Also, because of the relatively balanced classes (Table 3), F1 score does not increase for tilting when transiting from the no-class-weight to the class-weight model. Conclusion: This abstract presents the method used to develop a machine learning model based on CNN, with the objective of assigning open-channel deficiency ratings. Four deficiencies were selected (cracking, spalling, tilting, and vegetation), and the preliminary results indicate a general 60%-70% accuracy and 0.4-0.5 F1 score. In general, the achieved accuracy performance is positively related to the number of training data. Adding class weights into the model can increase F1 score, if the deficiency rating classes are imbalanced. More discussions will be provided in the following paper.

Today we have the benefit of having digital data available at our fingertips. However, in the stormwater profession, are we ready to maximize the use of that data? Each project presents opportunities to maximize the application of this data and how we can best collect stormwater data today to further the digital world we live in. When we collect stormwater data are we thinking about how that data can be used in models to address watershed solutions? Are we thinking about how our modeling results can be used to engage stakeholders? We have the technical resources to keep up with the rapidly changing digital world; the challenge is applying those technical resource efficiently to maximize the use of the data. Ultimately, it is using data to create more sustainable and resilient solutions that reduce costs and increases public benefit while managing the risk associated with a changing climate. Our toolbox includes stormwater models. Stormwater models are used every day to simulate design and historical rainfall events. We use them to demonstrate the impacts and the benefit those stormwater solutions given our assumed inputs. We use models to map floodplains, to evaluate enclosed systems, and to predict future flooding based on land cover changes or increases in rainfall intensities. The potential uses are endless, but do we get the full benefit of the resources put into developing these models over time? Are we building an adaptable and transforming digital platform or are we just developing a list of static projects? For stormwater solutions to advance and be more adaptive, resilient, and sustainable we need to have a plan and a vision of how stormwater data should be collected, how that data is maintained, what that data is going to be used for, and, who is using the data. We can't keep doing stormwater modeling in a black box every 15-20 years. Stormwater data and the subsequent models need to be living and growing digital systems that are refined and updated regularly with an eye towards smarter stormwater solutions that provide greater long term public benefit. This presentation will address these issues using the City of Lawrence, Kansas as a case study. The City of Lawrence 20-years ago went through a typical stormwater master planning process that resulted in a list of projects. Many of those projects have been completed but stormwater issues remain and, in some cases, are worse. Instead of going through another typical stormwater master planning process, the City is fully understanding their stormwater system through comprehensive field evaluation of their assets; developing the processes and procedures for using digital data in comprehensive watershed models; using the results of the watershed models to define improvement projects and engage stakeholders; and to create a holistic, dynamic stormwater improvement plan. The field data was collected using the latest technology, organized within a stormwater asset management system, and delivered as geodatabases, shapefiles, and digital photos and videos to the modeling team. The modeling team then used that data with recent LiDAR to develop stormwater models of the enclosed system, the 2D overland flow, and the 2D open channels and streams. The stormwater models include PC-SWMM with 2D and HEC-RAS 6.0 2D. The model files were spatially referenced and fully integrated with the City's GIS. The model inputs and outputs were easily mapped and overlaid with other City GIS data. This resulted in digital stormwater data and model simulation results that formed the foundation for the development of an adaptive and dynamic stormwater improvement process. This process will allow the City to use the models to make decisions, update stormwater improvement plans periodically, and provide watershed visualization materials to educate and engage stakeholders on stormwater issues. This presentation will focus on stormwater modeling in the digital world and how the latest field data collection technology, asset management systems, and stormwater improvement projects can be done more effectively. Lessons learned in using digital data for model inputs, 2D modeling, and linking enclosed storm sewer models with open channel models will also be presented. And finally, a process will be presented of how this can be accomplished on a watershed scale to drive innovation, improve the resiliency of stormwater solutions, and reduce a city's overall stormwater improvement cost.