MODERNIZING THE MODEL FOR STREAMBANK, WATERSHED, STORMWATER MANAGEMENT, AND FLOODPLAIN PLANNING AND RESTORATION

Transcription

1 MODERNIZING THE MODEL FOR STREAMBANK, WATERSHED, STORMWATER MANAGEMENT, AND FLOODPLAIN PLANNING AND RESTORATION April M. Barkasi, PE, CEO, CEDARVILLE Engineering Group LLC, Pottstown, PA PURPOSE The purpose of the paper is address an emerging technology of Reality Models and to expand on the applicability of such models for environmental applications. Our case study of an urbanized area demonstrates the use, accuracy, and efficiency of Reality Models and outlines the benefits applicable throughout the Civil, Environmental, and Construction Industries. BACKGROUND As a Municipal Consultant serving a small city in Southeast Pennsylvania, the professionals of CEDARVILLE Engineering Group, llc (CEDARVILLE) are faced with many challenges. These challenges range in scope from planning initiatives and public safety to multi-million-dollar infrastructure projects. The use of Reality Modeling is impacting the decisions on capital improvements and economic development. Coatesville, PA is not a unique community in that economic trials over time have adversely impacted the infrastructure of the City. The exportation of manufacturing and industrial jobs have left a once thriving small city with its prosperity based on the production of domestic steel struggling to maintain and cope with a degrading infrastructure. The scope of the socio-economic problems is compounded by abandoned industrial complexes leaving behind Brownfield sites and various environmental concerns. The City thrived off the industrious workings of the steel mill, remembered best as Lukens Steel and now famous for providing steel for the World Trade Center buildings. Though there is still some industrial use of portions of the old mill by Arcelor Mittal, the northern 26+ acres fondly referred to as the Flats which lays along the Brandywine Creek has been effectively abandoned. The City through the Coatesville Redevelopment Authority claimed the Brownfield site as the largest single track of land in the growing inventory of blighted properties owned by the. The industrial structures were long ago demolished, leaving behind an expanse of abandoned foundations, soil contamination, railroad intrusions, and floodplain vulnerability. The Flats is the single largest opportunity for economic redevelopment of this town and options are being explored to how to best repurpose this land to make it attractive to developers. Subject to Act Two permitting in Pennsylvania, the remediation of the contamination by the steel operation requires capping the contaminated soils with asphalt or overfill the entire site with two feet of earth. As the last horizon of untapped opportunity in poorest community in the wealthiest county in Pennsylvania, Coatesville has many avenues for funding through Federal, State, and County grant programs. As the City Engineer, we have emphasized the priority of infrastructure rehabilitation as the platform for growth and economic improvement. Using the Flats as a case study, we intend to show how Reality Models can improve the efficiency and accuracy of data collection for planning and design.

2 Figure 1. Aerial of "The Flats" with Mapped FEMA Floodplain EXISTING DATA Every great recover starts with an executable plan. Getting the most of the dollars available to the City means more money spent on bricks and mortar and less on planning and paperwork. To begin looking into various avenues to immediate improvement, we began to compile various disparate forms of data that would be relevant to the site. The readily available information included: Orthophotos of the area FEMA Data LiDAR (10m) Base survey from (2005, CMX project) this was not accurate spatially and had no terrain data Roadway survey from Roadway re-alignment of State Route 82 and Lincoln Highway (State Route 30) limited to adjacent streets GIS Data in form of shape files such as parcels, waterways, contours (contours derived from the LiDAR data) When looking at the data as a whole, it appears to be a reasonable amount of data to begin a planning process. The available data could provide a backdrop for a high level planning of options, though in no way provides us with complete information to reliably predict the true impact of decisions or planned options. To begin we need to assess each one of the data to understand the limitations and it usefulness:

3 OrthoPhotos Orthophotos are relevant in that they give us a visual starting point from a planning aspect. Two sets were reviewed, one from PA Map program and the second from DVRPC Both sets of imagery provided for one foot (30cm) ground resolution. The data was useful in comparing changes from 2008 to 2015, however the resolution did not provide a high level of detail to use in an engineering design and was taken when the canopy was on the trees making an accurate determination of features such as the stream bank impossible. FEMA Data DFIRM The data provided reflects the flood zones along the section of the Brandywine creek adjacent to the Flats. From this data we see the area is directly located in a flood zone AE therefore any grading will have impacts on this flood zone. While the DFIRM provided stream elevations there is no indication of a special study being done along this section of the mapped Floodway. We can conclude from this that any derived elevations were based on available elevation data as described below under LiDAR. We also must understand the floodway is an approximation and in no way can predict impact on area up or downstream. LiDAR PA MAP Lidar Las files ( ) the Lidar data available through this program includes bare earth point data with a density of 10meters. Horizontal Accuracy was set to NSSDA-1998 standard 95% of which is 5 feet or better. Vertical Accuracy was tested and met NSSDA/FEMA and NDEP/ASPRS methods for vertical accuracy the tested areas were within.67 to.34 feet dependent on the testing method. While this data is well within the accepted tolerances requirements for National Standard for Spatial Data Accuracy (NSSDA), Federal Emergency Management Agency (FEMA), National Digital Elevation Program (NDEP), and American Society for Photogrammetry and Remote Sensing (ASPRS), the models only provide a minimum level of detail to accurately model the Brandywine creek which is in direct contact with the site. The level of detail ideally suited would require an accurate field survey. Base Survey (2005) This survey was originally used in the layout of a walking trail located between the Brandywine Creek and the Flats The usefulness of this survey is suspect as it is not a complete survey and lacks detail. Additionally, no digital information was provided in the form of a surface Digital Terrain Model (DTM). When checked against a more recent survey the spatial reference was off > 10m, for these reasons this survey is virtually useless. Roadway Survey In a separate project related to the realignment of the major intersection of N First Ave (SR-82) and Lincoln Hwy E (SR-30 business), we were able to acquire the digital files of the existing conditions of this intersection. This survey was completed in a recent time frame and was tied to ground control. This survey provides many control we can use moving forward as it is in close proximity to the Flats GIS Data- Chester County had taken a very active role compiling and generating data in the form of overlays of various layers including Parcels, waterways, contours etc. The data provided a platform for useful information in the form of a database. However, we found much of the vector data is derived from other forms of data mentioned, i.e. the contours are a product of the LiDAR so are less accurate as they provide an interpolation of the original source data.

4 Before considering the acquisition of additional (costly) data we needed the potential needs. The potential for redevelopment, based on marketing the site as a viable economic opportunity, would the incorporate the following: 1. A recent general topographic and boundary survey. Survey would need to encompass an area > 30 acres and provide coverage of surrounding areas. 2. Flood Study Using HEC-RAS of the existing section of the Brandywine adjacent to the Flats 3. Flood study using HEC-RAS of the section of the Brandywine impacted by any changes 4. Flood Study needs to incorporate large structures present on the site 5. A 3D Model to be used for planning purposes 6. Public access to information 7. Cost As resources are low what are the expected costs for the data 8. A means to provide survey at various stages ADDITIONAL DATA NEEDS When looking at the list of requirements, it was clear to accurately model the site, the floodway, and the impacts, we would need to obtain a survey of a large area. There are several was we can accomplish this including typical survey with a Total Station, Laser Scanning, Aerial mapping, and a new emerging technology Photogrammetric Reality Modeling. Below is a chart of various means to accomplish this along with advantages and disadvantages of each. Advantages Limitations and Solution Advantages Limitations Survey Total Station Laser Scanner Aerial Survey (Stereo) Camera - Photogrammetric Reality model Reliable - Widespread use 3D High precision - instantaneous cover large area quickly 3D Texturized mesh, precision Highly portable Table 1. Comparison of Means and Methods of Site Data Collection level of detail is low, cost is high as it requires several days for the area n Requires skilled professional, calibration training, costly Inaccurate- limited to + or - a contour - Low detail value - cost Computing processing time Accuracy information of Low- will not provide precision needed on the scale proposed High-3D point clouds Low- Poor precision Very High, georeferenced mesh or point clouds deliverable, 3D model

5 REALITY MODELING Given the limitations primarily cost it was decided to generate an existing condition model using a new technology referred to as Reality Modeling. This new technology provides the user the ability to use virtually any digital camera and create a Reality Model. The models generated could be a texturized 3D mesh or a point cloud. There are numerous products available, so the criteria given below was used to determine which application to use. 1. Scalability Before making the investment we wanted to be sure the software we used could not just handle this project (34 acres) but any scale project we work with in the future 2. Engineering Precision, it is understood that the images directly impact the models precision, our project required 1-2cm accuracy or better. We estimated 750+ images or gigapixels (unit of measurement equal to one billion pixels) note: one megapixel equals one million pixels. 3. Accurate 3D mesh would be needed to work with design cross sections and profiles 4. Interoperability in a design (CAD Engineering) environment Based on our requirements we opted for Bentley Systems ContextCapture. Workflow The workflow for the project involved: 1. Compile existing Data 2. Capture Imagery for the Reality Model 3. Generate the Reality Model 4. Generate stream alignment 5. Generate cross-section of existing conditions along stream > Analyze using HEC-RAS 6. Generate cross-section of proposed conditions along stream > Analyze using HEC-RAS 7. Present City with findings, Conclusions, and proposals Capture Imagery Due to the size of the area to be modeled CEDARVILLE employed the use of an Unmanned Aerial Vehicle (UAV) to capture the images. Based on the desired precision the camera specifics and altitude would be determined. To make these determinations the follow formulas were used to estimate the expected precision of the output from ContextCapture. - Ls is the greater size of the sensor (mm) - D is the distance between the camera and the subject (m). - f is the focal length of the camera (mm) - L is the greater size of the photograph (Px) - R is the spatial ground resolution of the photos in (m/px) - P the precision of spatial positioning of the vertices of the 3D mesh. R=Ls Df L P=3 R

6 Ls, f and L are usually fixed for a given camera. Generally, the only way to improve accuracy is to decrease the flight height. This will also result in taking more photos to cover the same area. P is the relative precision of the 3D mesh. The absolute precision of your model will only make sense if it is georeferenced. In capturing the imagery other things to consider are overlap and the capturing of oblique imagery. Our photogrammetric software was capable of producing a high precision detailed geometric mesh because the computer vision algorithms optimized for working with oblique angles. Four oblique angles were captured in the original model (N, S, E, W). The imagery for the reconstruction of the model was captured in < 1 hr. of flight time. It consisted of 745 images taken at an altitude of ft. Figure 2- Oblique Imagery used as Part of the Dataset, captured at 180' Altitude with a Canon EOS M2 Camera Creating the Reality Model The Reality model was created using 745 images (all oblique). Four ground control points from a recent survey were used to give the model greater spatial accuracy. The final model Details are listed below. Area mapped- 32 acres or.13km2 745 Images taken from 4 oblique angles Gigapixels Resolution Area of interest- 1cm. Outside the area of interest 1-2cm Processing time approx. one day on one machine

7 The Flats Reconstruction (Reality Model) can be viewed here: Working with the Reality Model (RM) The Resulting Reality model was generated as a 3mx format. The 3mx is a multi-resolution format and the file size is nearly 30% smaller than a point cloud. Overall the processing power and responsiveness of the data is improved with the reduction in file size. Another other advantages we discovered in the process concerning working with a texturized mesh over a point cloud is that the texturized mesh provided the user with a complete visual story. Conversely, a point cloud is always perceived by the user s eye as vertices which may be unrelated, where the texturized mesh is easily recognized by the user. This made working with the 3mx reality model much easier for the user. No special tools outside of civil geometry functions were needed to create the alignment and profile for the stream bed. Generating the Design option grading with the RM A design option had to be proposed using the RM as existing ground. Two large stockpiles of clean material were previous placed on the site. The RM allowed for easy volumetric calculations to know the amount of fill needed to complete the two foot overfill for the mitigation area of the Brownfields area. The RM also provides a complete surface model for design, grading and infrastructure planning. The elevated grading areas, two ft. above the site area, were easily added and grades to the RM surface. The process to create the RM requires minimal effort after the images are acquired. The software used in the process uses a combination of Photogrammetry and computer vision, the process analyzing images to produce numerical or symbolic information, to generate the RM. No additional processing except for dropping the RM to a base CAD element. A designer can interactively move the newly created surface to achieve the desired grading for drainage, permit compliance, and infrastructure planning. Figure 3-volume analysis on existing stockpiles

8 Creating and exporting to for HEC-RAS Prior to working with HEC-RAS, a stream alignment and profile had to be established. Once generated, existing, and proposed cross sections could then be extracted from the RM and exported to HEC-RAS. The RM provided the designer with excellent information to base the stream profile. The designer could easily prepare the stream alignment based on observing the deepest section of the channel. The RM also provided relatable context when establishing the stream profile. The stream bottom was easy to follow as the capture of the photos were taken when the stream was low and clear. This allowed for the stream bottom to be profiled and accurately interpolated from the RM. Following the generation of the steam alignment, cross-sections were extracted for the entire channel, and extended portion of the property along the alignment for Floodway analysis. Generating the crosssections for export to HEC-RAS is a straight forward process given most design applications have this capability. Especially noteworthy is the detail the Reality Model provides in the cross-sections. Structural details from the abandoned mill foundations and bridge structures are clearly depicted in the cross sections. Figure 4 Cross-sections taken from the RM clearly detail existing structures. This level of detail would not have been available in the DFIRM analysis. Accuracy As mentioned earlier, accuracy was a key consideration in using a Reality Model. Because a field survey of the two adjacent streets was available, we were able to use ground controls from the established survey. For our RM we used four control points. Two points located on SR 82 were spaced approximately 350 ft.

9 apart. On SR 30 the controls were spaced 750 ft. apart. Using the intermediate survey, we were able to check existing surveyed features to verify the RM accuracy. The horizontal survey accuracy was within the 1.25cm (this was also the pixel size). Vertical accuracy in the controlled area of the streets was measured at < 2cm (3/4 inch). Vertical accuracy on the controls was within.07 of an inch). For our purposes the RM exceeded expectations for overall accuracy. SUMMARY By using this new technology of Reality Modeling on a localized project, we were ultimately able to bring surety to the issue of how new grading would impact of the Floodway and resulting effective economic viability of redevelopment on the Flats. The visual impact of the reality model is self-evident, and will continue to provide the City value throughout planning and design. Utilizing the photography and the Reality Model, we produced into a time lapse video for both public presentations and private funding and development interests. We continue to find new uses for this model and can only expect the usefulness or Reality Modeling to increase as they become more accepted within the engineering community. Figure 5- HEC-RAS Floodway Illustration as Demonstrated Reality Model Author Bio APRIL M. BARKASI, P.E. Founder, President and CEO of CEDARVILLE Engineering Group, LLC, April is an accomplished Professional Engineer with over 18 years of experience in the industry, who specializes in Municipal Engineering, Public Works, Sustainable Site Design, and Environmental Studies. She is passionate about redevelopment through investment in infrastructure and finds funding to drive that mission. She founded CEDARVILLE Engineering Group, LLC in 2010 in determination to provide excellent service and environmentally-focused, high quality work for our clients. April is personally dedicated to the success of each project and exceeding the expectations of clients.