CICEET Progress Report for the period 3/01/08 Through 8/31/08

Project Title: Vegetation, Impervious Surfaces, Soils, and Topographic Analysis Tools: Geospatial Technology to Promote Coastal Water Quality
Principal Investigator(s): Thomas R. Allen
Project Start Date: September 1, 2007
Report compiled by: Thomas R. Allen

Figures

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

B. Tasks to meet objective
1. Impervious surface cover mapping
The project reviewed newly available secondary sources of land cover, impervious cover, and potential primary sources. The available land cover included past USGS Multi-resolution Land Characteristics Consortium (MRLC) land cover and impervious cover, NOAA Coastal Change Analysis Program (C-CAP), and historical land cover data. In addition, newly available Landsat and very high-resolution satellites (IKONOS and QuickBird) were also evaluated. Pilot projects by the NC Center for Geographic Information and Analysis had reviewed these data for the same purpose of impervious mapping, also in comparison to the efficiency of classifying low-cost NASA MODIS ASTER satellite data. This project focused on the applicability of the high-resolution (30m x 30m) versus very high-resolution data (~1m x 1m to 4m x 4m) multispectral imagery. New data would be acquired and classified specifically to meet the needs of the study areas in Camden and Southern Shores, North Carolina. Initial results of this analysis would be presented at the American Society for Photogrammetry and Remote Sensing (ASPRS) annual meeting to solicit value and feedback.

2. Soil attributes database development
A variety of preexisting models have applied the USDA Natural Resources Conservation Service (NRCS) Soil Survey GIS data to agricultural and other non-point source pollution. We undertook an analysis of the data quality, mapping unit extent, and specificity of runoff curve numbers, soil horizon depths, and infiltration capacity of soils in the coastal plain. In addition, we sought to compare the variation of runoff versus infiltration into soils in our inland coastal and barrier island study areas.

3. Topographic runoff models and moisture indices
We desired a summary index of potential topographic runoff that would have a high predictive capacity to point out source areas of overland runoff, since much of the water quality problems in our sites originates from sheetwash and channelized suburban road networks lacking stormwater infrastructure. This task required comparison of existing availalable topographic runoff models and indices and the possibility to develop a new, hybrid model more predictive of water quality degradation (i.e., incorporating non-point as well as channelized point source pollution.)

4. Integrated soil and topographic models in ArcGIS ModelBuilder
The dissemination plan for this project builds upon the existing GIS capacity of local municipalities. These entities are provided with exemplary data in the State of North Carolina’s Floodplain Mapping Program. These data, augmented by available digital soil surveys and other models (a-c above), will nonetheless require technical query, data conversion, and analyses that may be beyond the technical or logistical capability of local personnel.

Recognizing the need to develop a tools with wide applicability (yet with robustness and customizable for expert users), we have initially adopted the ArcGIS ModelBuilder framework for product development. Each of the models of topographic runoff and soil infiltration would be developed in a separate model tool. A suite of such models would be efficiently packaged in a portable "ArcToolbox" and shared with partners and other users. This phase of the project requires exploiting the latest available software functionality, at times pushing against bugs and engaging with the vendor on technical limitations. New enhancements in the modeling software now allow submodels and functions, variable-based iteration, and the possibility of server-based processing. We also explored the possibility that with extensive high-quality data, we could provide a server-based geoprocessing tool. This would potentially overcome personnel capacity limitations of communities as well as exploit the latest efficient server technology and ability to work with the best statewide Lidar digital elevation models in the nation.

C. Progress on tasks
1. Impervious surface cover mapping
The product of the MRLC National Land Cover Database is currently in preparation for 2006. NOAA CCAP data have become available for 2001. No other applicable, current state or local land cover data exist.

Our project acquired IKONOS imagery for the Town of Southern Shores and proceeded to rectify and classify impervious cover.

Initial attempts also utilized Definiens object-based image analysis software. However, extensive vegetation canopy cover and ultimately required higher resolution to accurately define impervious covers at the 5m resolution of our resampled DEM. We combined the above supervised classification with manually digitized building footprints delineated from digital orthophotography provided by Dare County.

In addition, we completed an accuracy assessment of impervious cover, resulting in an overall accuracy of 91%.

Figure 1 depicts the true color composite of the study area in IKONOS imagery. Figure 2 Shows a subset of classified impervious cover. Buildings and roads were separately classified, with vegetation canopy obscuring some areas. To improve upon the accuracy, roads, driveways, and buildings that were not visible in the satellite data were manually digitized using Dare County orthophotography then burned-in by GIS overlay onto the impervious layer.

2. Soil attributes database development
We collected geophysical field data to qualitatively assess the use of NRCS digital soils data as well as ascertain the functional effect of road rights-of-way (ROWs) on surface runoff. In addition, we explored the possibility of mapping subsurface peat, a locally important factor in determining water infiltration, percolation, and throughflow drainage in soils. With collaborating Dr. Michael O’Driscoll (ECU Department of Geological Sciences), a series of ground resistivity transects and point samples and soil water infiltration measurements were made. The results of this analysis confirmed that local areas of subsurface peat appear to impact lower parts of the landscape, primarily occupying inter-ridge swales in the geomorphology of the region (a prograding beach spit with linear dune ridge-and-swale topography (Figure 3) In addition, the infiltrometer revealed that the ROWs exhibited far lower infiltration rates as compared to nearby undisturbed soils (Figure 4) This finding leads the GIS analysis to be conservative with respect to impervious cover of roadways. In fact, the disturbed roadbed adjoining the actual tarmac is only semi-pervious. As a result of surface runoff from asphalt to ROW, surfactants of many types, erosion of coarse sediment, and denuded vegetation rooting may effectively lead these areas to be considered impervious.

3. Topographic runoff models and moisture indices
Progress on modeling topographic runoff includes results from employing an ensemble of available models and implementing a new ArcGIS model on a previously published, ecological-oriented "Integrated Moisture Index" (Iverson et al. 1997.) A suite of other tools were also obtained and evaluated, including an ArcGIS model for curve-number calculation from ESRI (Darren Baird, pers.comm) and a runoff model developed for post-wildfire watershed analysis (Richard Easterbrook, pers. comm.)

We completed an ArcGIS implementation of IMI and continue refining it to evaluate its portability. In our coastal plain landscape, it is important to note that the quality of any runoff model is highly dependent on the accuracy and resolution of the digital elevation model used. In our Camden County study site, ditches are widely prevalent yet not represented in the Lidar DEMs. We are currently developing a "stream burning" algorithm to automate the adjustment of DEMs (and associated runoff models) by reducing pixel elevation values for ditches. This stream-burning model will be an ancillary byproduct of the project useful for a wide array of coastal GIS applications. We are comparing the accuracy of two alternative approaches; 1) mapping ditches and tagging their depths by type with subsequent burning into the DEM or 2) mapping ditches and buffering from raw Lidar points followed by re-interpolation of DEMs, to ensure ditches and not artifacts of them, are represented in the final surface model. This requires a field trip to also ensure that we have adequately mapped the ditches and have reliable depth points from the Lidar surveys.

4. Integrated soil and topographic models in ArcGIS ModelBuilder
This objective is preliminary in this phase of the project. Rather, it is the initial scoping and plan for the final modeling and tool product. The above models are in various stages of validation to assess their accuracy, portability, and ease of use. We are also engaged in informal and public presentations of the project to assess the end-user community’s perspectives. The project is still considering alternative formats for the tool, from an Arcobjects-based ArcGIS extension, a suite of ArcToolbox Models, or a online GIS server-based geoprocessing service. The project has revised the budget to also include a consultant to support programming. In addition, the PI has acquired ArcGIS database and webservers from another project that could serve as platforms for the online GIS server dissemination.

We also contrast our approach with an existing inferential method for predicting nutrient loading by human development in watersheds. The first of these is the NOAA Coastal Services Center Impervious Surface Analysis Tool (ISAT). The tool was implemented for the Town of Southern Shores using Census 2006 estimated block population and 2001 CCAP land cover. While the automated low, medium, high range-graded scale did not highlight impervious problems, our custom legend better distinguished the percent impervious cover. It should also be noted that we did not use the default National Hydrologic Database (NHD) basins boundaries, but instead supplied the algorithm with our higher resolution subwatersheds. Using this tool, the major runoff problem areas were generally distinguished from non-problem basins. However, the algorithm, as designed, points to the aggregate basin and not the actual sub-basin parcels or lots. In addition, the use of the Census underestimates impervious cover, since many of the town’s beachfront and soundfront homes are second or vacation residences, and not counted as populated by the Census.

Further analysis also revealed that the topographic modeling could identify likely overland runoff channeliziation (Figure 8) Overlays with road networks revealed that the road and rights of way in some instances actually would channelize the runoff.

Further, when we analyze the proportion of impervious cover in a basin between problem runoff sites and non-problem sites (Figure 9), there is a qualitative difference between the two. This led to an empirical statistical evaluation, as to whether percent impervious cover and spatial metrics could be used predictively. The topic was presented at the Association of American Geographers annual meeting, but details are for publication. The IMI model has shown potential to summarize the conditions, as it did for ecological application in the initial research by Iverson et al. (1997.) In this case, we would use the IMI index as a guide to the site potential for inundation versus runoff (instead of vegetation moisture availability, a factor in "site index"). Such a simple index, if shown robust and portable, could be more widely interpreted and used to prescribe best management practices for stormwater.

D. Difficulties in accomplishing tasks
Significant difficulties were overcome during the period. We found parsing and modeling of soil attributes data for integrated modeling cumbersome. The result was a solution in a model form that would be much more efficient for end-users. In addition, we suffered the loss of a collaborator to job change in one partner community, requiring new contact and logistical arrangements. We did receive strong support from the community in a public workshop and continue to have the support of NCNERR for our field control site in Kitty Hawk Woods reserve.

We encountered certain bugs in the ModelBuilder software that have been overcome with work-around solutions and the new release of ArcGIS version 9.3 New capabilities in the geoprocessing tools also will improve our modeling approach. The PI and assistant have developed their capabilities in this area by taking ESRI Python programming training.

In our terrain modelilng, we discovered relatively poor rendering of ditches in the Lidar bare earth surface models of the coastal plain. This required us to critically evaluate the role of these DEMs, with the result that we would initiate the "stream burning" technique and evaluate the cost and benefit of this to the model. The need for this, guided by our collaborator’s desire for high resolution, requires continued analysis on this and possible incorporation of the technique within our modeling toolkit.

II. Data generated
Geospatial data products to the current phase of work include several layers. Among the new data generated, we have focused on producing:

• Impervious surface cover for the Town of Southern Shores
• Building footprints and rights-of-way
• SSURGO soils data
• Distribution of soil resistivity along transects
• Locations of ground water monitoring wells in Southern Shores and Kitty Hawk Woods (NCNERR)
• Integrated Moisture Index
• Major and subbasin watersheds delineated by terrain modeling (TauDEM and ArcHydro)
• NOAA Impervious Surface Analysis Tool (ISAT) tool results, including runs with high-resolution Lidar-derived watersheds
• Ditches and stream-burn DEM for Sawyer’s Creek watershed, Camden County

In addition to new data, more importantly, we have assembled a set of models for assessing their utility as well as applying and redeveloping them (e.g., IMI) in the latest software as ArcGIS geoprocessing models:

• Integrated Moisture Index (IMI) model for ArcGIS ArcToolbox
• ArcCN-Runoff (modifications to work with NOAA C-CAP land use/land cover)
• Miscellaneous tools for soil attribute conversions and queries, high-resolutionsub-basin NOAA ISAT, etc.

III. Project objectives for next reporting period

1. Technical objectives
a. Complete the modeling of IMI and stream-burning coastal plain DEMs and reassess the effectiveness in Camden County.

b. Augment the ArcCN-Runoff model with a lookup table for NOAA CCAP land use/land cover, allowing portability for coastal plain analyses

c. Plan for the local, high-resolution analyses, model, and query capability (i.e., parcel level contribution metrics)

d. Assemble derivative GIS data into a database for possible inclusion in a webserver application.

2. Non-technical objectives
a. Obtain end-user community feedback
i. NC ArcUser Group Meeting presentation and discussion, Wilmington (Sep. 17-19)
ii. On-site meetings with partners in Dare and Camden Counties
iii. Presentation to broader scientific community (e.g., Coastal Zone ’09) and CICEET (Dec. 2008)

IV. Expenditures
Approximate expenditures during the semi-annual reporting period include:
Principal Investigator (1 mo. summer salary) $10,000.
Undergraduate student assistant (spring and summer) $ 4,000.
Travel (ASPRS conference presentation and local coordination) $ 1,200.

References
Iverson, L.R.; Dale, M.E.; Scott, C.T.; Prasad, A. 1997. A GIS-derived integrated moisture index to predict forest composition and productivity in Ohio forests. Landscape Ecology. 12:331-348.

Zhan, X. and M-L. Huang. 2004. ArcCN-Runoff: an ArcGIS tool for generating curve-number and runoff maps. Environmental Modelling and Software. 19(10):875-879.