Mapping Urban Land Use Land Cover Using Quick
Mapping Urban Land Use / Land Cover Using Quick. Bird NDVI Imagery for Runoff Curve Number Determination Pravara Thanapura 1, S. Burckhard 2, M. O’ Neill 1, D. Galster 3, and E. Warmath 4 Engineering Resource Center 1, Civil and Environmental Engineering 2, and Mathematics and Statistics 3, South Dakota State University, Brookings, South Dakota Department of Transportation, State of Nevada 4 Introduction The National Resource Conservation Service (NRCS), formerly the Soil Conservation Service (SCS), has developed a series of hydrologic models for water resource planning and design. The SCS Curve Number (SCS-CN) method is the best-known component of a series of SCS hydrologic models. A major catalyst for the development and implementation of the runoff CN methodology was the passage of the Watershed Protection and Flood Prevention Act (Public Law 83 -566) in August 1954. The CN method is well established in hydrologic engineering and environmental impact analyses. The method and its use are described in the SCS National Engineering Handbook Section 4: Hydrology (NEH-4). Today, the method has been widely used in numerous applications by practicing engineers and hydrologists nationally and internationally (Mishra and Singh, 2003. ( A critical component of the SCS-CN method is a runoff index known as the SCS Runoff Curve Number (CN) described in TR-55 (NRCS, 1986)[Figure 1. [ The CN is basically an index determined from the table in Figure 1. The CN is a function of three factors: a hydrologic soil group, the cover complex, and antecedent moisture conditions. Conventional ground-based methods for determining CN values are time-consuming and labor-intensive procedures. A basic problem exists in quantifying the detailed spatial extent and distribution of various land cover classes. Utilization of satellite remote sensing and geographic information system (GIS) technologies can provide spatially and temporally distributed input parameters for CN determination. The purpose of this applied research is to integrate remote sensing and GIS to produce scientific knowledge by designing technological methodologies for determining a composite of CN and deriving CN in urban watersheds. The theoretical foundation of this research is that mapping a high resolution NDVI image generated using the ISODATA algorithm is an efficient and effective information extraction approach for the determination of CN. To show the use of Equation 1, the CN values for industrial districts with 72% imperviousness are shown below. These values are the same as those shown in the TR-55 table in Figure 1. A 1 soil: 39 (0. 28) + 98(0. 72) = 81. 48 (~ 81) B 2 soil: 61 (0. 28) + 98(0. 72) = 87. 64 (~88) C 3 soil: 74 (0. 28) + 98(0. 72) = 91. 28 (~91) D 4 soil: 80 (0. 28) + 98(0. 72) = 92. 96 (~93) Note (Mc. Cuen, 1982): 1 A: The soil characteristics are deep sand, deep loess, and aggregated silts. 2 B: The soil characteristics are shallow loess and sandy loam. 3 C: The soil characteristics are clay loams, shallow sandy loam, soils low in organic content, and soils usually high in clay. 4 D: The soil characteristics are swell significantly when wet, heavy plastic clays, and certain saline soils. Figure 2. Sequence for digital processing and analysis. Study Area The study area for this project is centered on 43 O 31’ 18” north latitude and 96 O 44’ 42’’ west longitude in the southwestern part of the City of Sioux Falls. The study site is covered with various land use /land cover types and Benefits The results of this research could lead to an improved scheme for determining the runoff index used in two popular urban watershed runoff assessment methods - the SCS curve number method and the rational method. The application of this approach could benefit municipal engineers who are responsible for drainage analyses, the design of minor types of hydraulic structures, and maintenance and improvement projects in urban areas. Estimation of Curve Number for Urban Land Uses: A Review According to Mc. Cuen (2005), each CN described in Figure 1 is based on a specific percent of imperviousness. For example, the CNs for industrial districts are based on an imperviousness of 72%. For urban land uses with percentages of imperviousness different from those shown in the TR-55 Table, the CNs can be computed using a composite CN approach, with a CN of 98 used for the impervious areas and the CN for open space (good condition) used for the pervious portion of the area. Thus, CNs of 39, 61, , 74 and 80 are used for hydrologic soil groups A, B, C, and D, respectively. The following equation can be used to compute a composite CN (CN c) CNc = CNp(1 – ƒ) + ƒ(98)(1) In the preceding equation, ƒ is the fraction (not percentage) of imperviousness, and CNp is the curve number for the pervious portion (39, 61, 74, or 80). Classification Approach A Normalized Difference Vegetation Index (NDVI) using Quick. Bird (QB) Imagery was selected for this research project because of the land surface reflectance characteristics of the QB red and NIR bands. The relatively simple NDVI was utilized for the following reasons: > Reduce heterogeneous spectral-radiometric characteristics within land use / land cover surfaces portrayed in a QB image. > Improve the accuracy of mapping impervious surfaces and open spaces for different hydrologic conditions as used in the proposed SCS runoff curve number (CN) calculation. > Normalize potential atmospheric effects within the image. NDVI has been shown high correlation with green leaf biomass and the green leaf area index. Chlorophyll, the primary photosynthetic pigment in green plants, absorbs light primarily from the red and blue portions of the spectrum, while a higher proportion of infrared is reflected or scattered. As a result, vigorously growing healthy vegetation has low red-light reflectance and high NIR reflectance, and hence high NDVI values. Impervious surfaces (e. g. , asphalt and buildings) and bare land (e. g. , bare soil and rock) have similar reflectance in the red and the NIR, so these surfaces will have values near zero. Figure 4 shows a) the Quick. Bird multispectral image displayed using bands 4, 3, and 2 for red, green, and blue, respectively, b) the Quick. Bird NDVI image, and c) the natural color orthophoto used for reference and validation. In the NDVI image, the impervious surfaces are the dark areas with low NDVI DNs and the brighter areas are vegetated areas with high NDVI DNs. includes all or part of 26 hydrological urban sub basins (Figure 3). This area encompasses 2, 901. 42 acres (1174. 2 hectares) and primarily consists of soil Objective The objective of this study was to demonstrate and evaluate Normalized Difference Vegetation Index (NDVI) data derived from Quick. Bird (QB) satellite imagery to map land use / land cover surface characteristics such as impervious areas and open spaces for runoff curve number (CN) determination. Runoff Index Spatial Model ↓ Digital Data ↓ Pre-Processing Data Merging and Integration Quick. Bird NDVI Imagery & GIS Layers ↓ Decision and Classification ↓ Image Classification Unsupervised – ISODATA Algorithm ↓ Classification Output ↓ Accuracy Assessment ↓ Reports and GIS Data ↓ GIS Spatial Modeling The Composite of Runoff Curve Number Calculation ↓ Reject / Accept Hypothesis group B (~ 86. 34%). Figure 1. TR-55 runoff curve number table for urban areas (United States Department of Agriculture, 1986) (a) QB multi spectral image (4 -3 -2) (b) QB NDVI image Methods This research paper presents a new approach of utilizing high spatial resolution satellite data for mapping impervious areas and open spaces and developing GIS spatial modeling for determining runoff curve number (CN) as utilized in the SCS CN method in urban watersheds. To achieve this goal, a composite runoff index spatial model (Model 1) and a sequence for digital processing and analysis were proposed and implemented for mapping impervious areas and open spaces using Quick. Bird (QB) NDVI imagery. A GIS spatial model was also developed for determining the composite CN and the CN values (Figure 2. ( In order to assess the utility of QB NDVI imagery and to validate the composite CN calculation, various procedures and comparisons of the generated CN were presented and reviewed by practicing professionals, including the City Drainage Engineer and the City GIS manager, for the City of Sioux Falls, South Dakota in August 9, 2005. Model 1. The Composite Runoff Index Spatial Model 2005 ©]Pravara Thanapura. Used with permission]: RIc = [the ƒ of area covered (j/i) x RI(j)] Where: RIc or Runoff Indexcomposite is the sum of the component runoff index within : area (i) delineated by the description of area. an The ƒ or the fraction of area covered (j/i) is the component (j) of the area(i) divided by the total area (i). Note that the component of the area is delineated by surface characteristics of land use / land cover, hydrologic soil group, and/or slope. RI(j) or Runoff Index(j) is a runoff index of the component (j) of the area (i) determined by the surface characteristics and/or its hydrologic soil groups. (a) The QB multi spectral image (4 -3 -2) (b) The QB NDVI image Figure 3. The 2004 images of the study area at the same scale. Digital Data & Preprocessing The 2004 remotely sensed data and vector GIS data layers were provided by the City of Sioux Falls. Data used in this project include the following: > Quick. Bird image (blue, green, red, and near-infrared bands) collected on April 26, 2004, with 7. 96 ft. (2. 39 m) resolution. > Orthophoto mosaics acquired on April 23, 2004 and May 20, 2002, with 2 ft. (0. 6 m) and 0. 5 ft. (0. 15 m) resolution, respectively. > GIS data layers such as parcels, hydrology, and streets. > The 2004 NRCS SSURGO 1: 24, 000 data set. Data was processed and merged to combine the remotely sensed data with the GIS layers for the study area. Software used in this project included ERDAS Imagine 8. 7, Arc. Map 9, Arc. View 3. 3, and Microsoft Excel 2002. The April 23, 2004 orthophoto was used as a reference to integrate all data sets into the same map projection and units (Universal Transverse Mercator map projection, World Geodetic System 1984 datum, and horizontal units in feet). The QB subset scene was generated and registered to the 2004 orthophoto using 31 ground control points, resulting in an RMS of 0. 6940 pixel. The red channel (band 3: 630 -690 nm) and the near infrared channel (band 4: 760 -900 nm) were processed to create unsigned 8 -bit QB NDVI image (NDVI= Band 4 – Band 3 / Band 4 + Band 3) for the study area. (c) Orthophoto (1 -2 -3) Figure 4. Quick. Bird multispectral (a) and NDVI (b) images showing impervious areas and vegetative open spaces and the higher resolution orthophoto used for reference and validation. Acknowledgements The author would like to thank the following employees of the City of Sioux Falls, South Dakota, for their contribution to the success of this research project: Steve Van Aartsen, GIS supervisor, and his staff, for all the data and information contributed for this project; Jeff Dunn, City Drainage Engineer, for his time and the literature provided on hydrology in Sioux Falls; and Sam Trebilcock, Transportation Planner, for his assistance and interest. The author would also like to thank the co-authors for their support, input, effort, and specialized expertise in making this project a success. Additional thanks go to Mr. Kevin Dalsted, Director, Engineering Resource Center, South Dakota State University, Brookings, South Dakota, for the use of the hardware and software that made this research possible.
- Slides: 1