A Little Spare Change

Transcription

1 A Little Spare Change Monitoring land-cover change by satellite by Introduction Problem Can city utility services use remote satellite data, processed with geographic information systems (GIS), to help track urban development? In particular, which areas of the city of Raleigh exhibit evidence of possible change in land cover during a period of high growth? The city of Raleigh, North Carolina, is interested in meeting and maintaining federal standards of water quality in the local waterways. As in many other municipalities, programs to reduce water pollution are funded through a utilities fee levied on city landowners based on the area of impervious surfaces on properties within the city. The city Public Works Department is responsible for maintaining accurate records of these surfaces, both for fair collection of these fees and for appropriate planning of projects to manage storm water runoff. This is a considerable task for the 45th-largest and eighth-fastest growing city in the US (City Data), but perhaps the use of high-resolution satellite imagery can make the process both faster and less expensive. Location Raleigh, North Carolina Time to complete the lab One and one-half hours Prerequisites This lab is intended for upper-class undergraduates who have a familiarity with the manipulation and presentation of raster data in ArcMap. Keywords: impervious surfaces; storm water; satellite imagery; change detection; change vector analysis; change magnitude; spectral space; components, distance formula; mask; raster calculator; Spatial Analyst; layer properties; reclassify; symbology

2 Data used in this lab IKONOS satellite imagery ( Geographic coordinate system: GCS_North_American_1983 Datum: D_WGS_1984 Projection: Universal Transverse Mercator Zone 17N Student activity Consider the following scenario: You have recently started a job with the Storm Water Management Division in the Public Works Department of the city of Raleigh, North Carolina. The division oversees construction projects and city maintenance that guards against flooding and helps maintain the quality of water that flows into local streams, rivers, and the local drinking supply. These projects (and your salary!) are funded through the collection of a fee on impervious surfaces i.e., buildings and paved surfaces such as parking lots, sidewalks and driveways, which all contribute to excess surface runoff by preventing rainfall from soaking into the ground (infiltrating). The city maintains a database of impervious surfaces and requires registration of new construction as part of the building permit process, but every few years it spends a great deal ($330,000) verifying the accuracy of the database for the entire city. Part of this process uses high-resolution satellite imagery to help identify impervious surfaces. Your first task on the new job is to explore whether such imagery could be used instead to discover priority areas where land-cover change has occurred, and perhaps establish a new procedure that limits the verification process to these areas and saves money that can be used for pollution reduction projects. In short, you wish to determine which areas of the city of Raleigh exhibit evidence of possible change in land cover during a period of high growth. Your data for this project consists of two sets of multiband high-resolution IKONOS satellite imagery (GeoEye foundation) for portions of the city: one image set for the year 2002 and another set for the year These two image sets are not perfectly aligned but do intersect over a significant portion of northeast Raleigh. In this lab, you will create a raster mask representing the area of available imagery data; extract satellite data for multiple bands; calculate multiple-band components of a change vector; create, analyze and visualize a change magnitude raster; create a map of likely change areas; and investigate the effectiveness of using a change magnitude raster. 2

3 PREPARE YOUR WORKSPACE You will be working with and generating a number of data files as part of this project, so it will be important to stay organized. Begin by creating a project folder on your computer or storage device (e.g., My Documents\SpatiaLABS). 1 Create a SpatiaLABS folder under the My Documents folder. 2 Create a Data folder under the SpatiaLABS folder. COLLECT THE DATA Now you will copy the lab data to your local workspace. 1 Copy the contents of the ALSCdata folder to your My Documents\SpatiaLABS\Data folder. View the data to confirm extraction before moving on to the analysis. 2 Start ArcCatalog. 3 Examine all the data in the Data folder. There should be 10 satellite data files. Eight contain values for each of four spectral bands for two years (blu02.tif, blu06.tif, grn02.tif, grn06.tif, nir02.tif, nir06.tif, red02.tif, red06.tif), and two are panchromatic images for 2002 and 2006 (pan02.tif, pan06.tif). PREPARE DATA 1 Start ArcMap. 2 Add each of the eight data bands as layers to the ArcMap data frame. 3 You may notice that the datasets appear featureless at first. In Layer Properties, modify the symbology of each layer so it is either Stretched or Classified in such a way that the features of the dataset are visible. ( Do you want to compute statistics? Click Yes if asked). It might also be useful to choose different color ramps for 2002 and 2006 data. 4 Display and observe each layer and compare the different bands and different dates. Question 1: How do the datasets differ for the different dates? 3

4 Question 2: Are there particular features of the landscape that are easier to identify with certain data bands? Question 3: Is there anything unexpected or strange about the data? (GeoEye) In order to make direct comparisons, all the datasets will need to be defined over the same spatial area. You will want to create a mask that represents the area where spatial data exists for both years, and then use that mask to extract the proper pixels from each dataset. You should have noticed that each dataset contains regions of blank data, which should also be excluded by the mask. Question 4: How are the blank regions represented in the dataset? Question 5: Why do you suppose there are blank regions in each dataset? MASK DATA There are a number of possible ways to mask data. In this case, the goal is to generate raster files for each year and in each band that can be matched pixel for pixel. Formulate a process of your own design or use the following steps. 4

5 Create a raster mask 1 On the main menu, click Customize» Extensions and, if necessary, activate the Spatial Analyst extension with the check box list under Customize» Extensions. 2 Open the ArcToolbox window and find the Raster Calculator under Spatial Analyst Tools» Map Algebra» Raster Calculator. The calculator allows you to manipulate one or more datasets by applying a mathematical function to the data in each pixel, matching those with the same spatial location in different datasets. For example, [data1.tif] + [data2.tif] would create a third file where each pixel would be the sum of the values from the matching pixels in the first two files. 3 Use the calculator to generate a new raster that is some mathematical combination of the 2002 and 2006 bands. Note that the calculator will only generate data where it exists for both sets i.e., for the intersection of the sets used in the calculation, which is what is desired for the mask. 4 Give the new intersection raster an appropriate name, and then modify the symbology of this raster so that the blank regions are still evident. If you can no longer identify these regions, you will need to rethink the choice you made in combining the two bands and return to step 3. 5 Look again in ArcToolbox for the Reclassify tool (Spatial Analyst Tools» Reclass» Reclassify). Select your newly generated intersection raster as the Input raster and leave value as the Reclass field. Edit the reclassification table to reassign a NoData value to the blank pixel values (assuming you know what they are), and a value of 1 to all other pixel values. You may want to use the Classify functions to reduce the number of classes. Name the output raster Mask. 6 As an alternative to step 5, use the Raster Calculator to normalize your intersection raster and divide the raster by itself. Compare this new calculation to your step 5 mask. Display one of your mask layers. Question 6: Why would a raster value of 1 for the mask be important? Question 7: Why does the alternate version in step 6 eliminate the blank values? 5

6 (GeoEye) Set a masked environment 1 Select Environments from the Geoprocessing menu to open the Environment Settings dialog box. On this dialog box, locate Raster Analysis from the 17 settings categories and click to reveal the associated settings. 2 Use the selection menu to set the cell size to Same as layer Mask. This should fill in the associated text field with a value of 4. 3 Select Mask (your layer name) as the Mask field. Click OK. Note that these environment settings will now apply to all your future calculations. 4 You may want to save your map document at this point, if you haven t done so already. 6

7 PERFORM CHANGE VECTOR ANALYSIS Each raster pixel in the masked area of the datasets is now represented by eight values: four measured in 2002 and four from Each of these four values can be considered a component of a four-dimensional vector (yes, you are working in 4D) representing the specific location at the specific time. You can imagine or draw these vectors as points in what is called spectral space, with axes that represent the spectral band components (see the examples below in two and three dimensions). nir 2002 Spectral Band Values green A change vector can be generated for each pixel by subtracting the 2002 vector from the 2006 vector, which is simply the subtraction of each component. nir band values values red band A simple measure of change would then be the magnitude of each change vector. A more advanced analysis would consider the direction of the change vector as well. 7

8 PREPARE THE ANALYSIS Repeat the following steps for each of the four data bands (red, green, blue, nir) to create the components of a Change Vector for each raster pixel. 1 Identify the 2002 and 2006 raster files for the chosen band. 2 Start the Raster Calculator tool. 3 Generate a new raster by subtracting the 2002 band from the 2006 band. 4 Give each new raster an appropriate identifying name. Note that because of the Environment Settings, these new rasters will be defined only over the masked area where both had data. 5 The length of a vector can be calculated with a multidimensional version of the distance formula: ( red) 2 + ( grn) 2 + ( blu) 2 + ( nir) 2 where red is the change in the red band, or the red component of the change vector (red06 - red02 ). Now calculate a Change Vector Magnitude Raster. 6 Start the Raster Calculator tool. 7 Generate the raster using the four change vector bands and the distance formula. Both Square( ) and SquareRoot( ) functions are available from the function menu; these functions are applied to whatever is put inside the parentheses e.g., SquareRoot( data.tif ). Be advised that Raster Calculator syntax uses the caret symbol (^) for the Boolean XOR function and not for exponentiation. VISUALIZE 1 Display the Change Vector Magnitude Raster. Modify the symbology so that the areas of high change are highlighted with a hot color and the areas of minimal change are associated with a cool color. 2 Choose and display images for the same band, but two different dates (e.g., red02.tif and red06.tif) so they can be compared with the change raster. 8

9 3 Use the Classify button on the Symbology page to view the histogram of the change raster. Determine the break value between the bottom 90% and top 10% of the values in the change raster. Record this value for step 4. 4 Use the Raster Calculator a final time to generate a Top 10% change raster, selecting all the values greater than the 90% break value. 5 Modify the color and transparency of this Top 10% change raster so that it can be displayed as an overlay of the original data. Display the change raster with the multiband images for 2002 and Export these two maps as image files for your instructor. ANALYZE 1 Identify areas on the change raster that appear to have significant change. 2 Add the panchromatic images, pan02.tif and pan06.tif, to the ArcMap data frame. Zoom in on the change areas and compare them to the 2002 and 2006 panchromatic images. Change Magnitude Raster 2002 and 2006 panchromatic images (GeoEye) 9

10 Question 8: Can you identify regions that have experienced change between 2002 and 2006? Create close-up images of two different areas that demonstrate change and submit them, appropriately labeled, to your instructor. Describe the change that you see what was there in 2002, and how was it different in 2006? Question 9: Are there any false positives in your detection method i.e., are there areas that were identified by the method that perhaps don t represent actual change? Describe what you see. Question 10: Do you observe any other patterns in how the change raster relates to what you see in the two multiband images? Question 11: What additional or different processing would you suggest to improve the method? References Ferres, R. Land Change Detection Using High Resolution Imagery for Raleigh, NC. MS thesis, North Carolina Central University, Durham, NC, Advameg Inc. n.d. Raleigh, North Carolina. In City Data. Accessed March 23, City of Raleigh. n.d. Public Works Dept. In The Official City of Raleigh Portal. Accessed March 23, GeoEye. n.d. Satellite Imagery Products. In GeoEye. Accessed March 23, Johnson, R. D., and E. S. Kasischke Change Vector Analysis: A Technique for the Multispectral Monitoring of Land Cover and Condition. International Journal of Remote Sensing 19 (3): Lillesand, T. M., R. W. Keifer, and J. W. Chipman Remote Sensing and Image Interpretation, 5th ed. NY: J Wiley & Sons. Lu, D. et al Change Detection Techniques. International Journal of Remote Sensing 25 (12): Singh, A Change Detection in the Tropical Forest Environment of Northeastern India Using Landsat. In Remote Sensing and Tropical Land Management, edited by M. J. Eden and J. T. Parry, London: J Wiley & Sons. 10

11 Submit your work Submit to your instructor answers to questions 1 through 11 and your summary assessment of land-cover changes occurring in the city of Raleigh, North Carolina, between 2002 and Credits Sources of supplied data ALSCdata\blu02, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\grn02, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\nir02, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\pan02, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\red02, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\blu06, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\grn06, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\nir06, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\pan06, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye ALSCdata\red06, IKONOS satellite imagery, GeoEye Satellite Imagery courtesy of The GeoEye Instructor resources This module is written as an introduction to multispectral change analysis and can be used as a launching point for introducing more advanced techniques such as change vector analysis and principal components analysis. It is intended for upper-class undergraduates who have a familiarity with the manipulation and presentation of raster data in ArcMap. 11

12 This lab is followed well by the lab Change in the Right Direction: Monitoring land-cover change by satellite, which is designed to introduce tools and concepts that may be used in more advanced analyses and to prompt thought about how to approach such problems. It is intended for upperclass undergraduates who have a familiarity with the manipulation and presentation of raster data in ArcMap, including use of the Raster Calculator tool. Analysis and visualization tools This lab is designed to be completed with ArcGIS 10 with the Spatial Analyst extension, although the lab can be completed with ArcGIS 9. Lesson notes and comments PREPARE DATA The band values in the original.tif imagery range from 0 to 65,535, and because of relative sparseness on the high end will appear nearly monochromatic. The example image uses a Stretched visualization of the Standard Deviations, with a 40% transparency on the 2002 file. Question 1: How do the datasets differ for the different dates? Answer: Students should note that the datasets are defined over different spatial ranges. Question 2: Are there particular features of the landscape that are easier to identify with certain data bands? Answer: Answers will vary. Water bodies, forested areas, and dark asphalt are good examples. Question 3: Is there anything unexpected or strange about the data? Answer: This is a prompt to ensure that students notice the blank triangular regions on the sides of both datasets. Question 4: How are the blank regions represented in the dataset? Answer: The Identify tool can be used to demonstrate that the blank regions all have a pixel value of zero. Question 5: Why do you suppose there are blank regions in each dataset? Answer: Satellite sweeps do not have a perfect orientation with latitude and longitude, but data is most easily handled in perfectly rectangular blocks, which contain zeros outside the satellite sweep range. 12

13 MASK DATA One feature of the raster calculator is that it only produces output for spatial areas where all included datasets are defined, so it effectively executes an intersection of datasets. The most direct process for creating the raster mask is to use the Boolean AND, which assigns a value of 1 where both sets are non-zero and 0 where either set is zero. Multiplication of the two sets will also preserve the blank areas. Other mathematical combinations (e.g., simple addition) will alter the blank areas and make it difficult or impossible to subsequently mask them out. Question 6: Why would a raster value of 1 for the mask be important? Answer: The value of 1 is the multiplicative identity, meaning that multiplication by the mask will preserve the values of the original raster. Question 7: Why does the alternate version in step 6 eliminate the blank values? Answer: Division by zero is an undefined numerical value, which the raster calculator does not include in the output set (assigns a NoData value). Dividing any dataset by itself normalizes all nonzero values to 1 while eliminating all zero values. 13

14 PERFORM CHANGE VECTOR ANALYSIS PREPARE THE ANALYSIS Students may note that it is not necessary to directly generate the component bands of the change vector (steps 1 5) to generate the change magnitude raster (steps 6 7). You may wish to point out the Power(, ) function as the appropriate syntax for exponentiation, again emphasizing that the caret symbol (^) is not proper syntax in this context. (It has common usage as an indication of exponentiation in other computing environments). 14

15 VISUALIZE Value duplication makes it impossible to single out exactly 10% of the dataset. Using the Quantile Classification Method yields a 90% threshold value of 508.5, but that represents only 357,646 pixels of 4,852,114, or 7.37%. Manual manipulation yields a threshold value of 468.9, representing 483,917 pixels, or 9.97%. The latter value was used as the threshold for the following images. Change Magnitude Raster with multiband imagery (GeoEye). 15

16 Top 10% Change Raster with multiband imagery (GeoEye). ANALYZE The example area is located in the upper-left corner of the masked area against the western boundary. Question 8: Can you identify regions that have experienced change between 2002 and 2006? Answer: The mall parking lot in the upper-right corner is the most obvious example. There are cleared development areas in the lower right and in the middle and new buildings scattered throughout the area. Question 9: Are there any false positives in your detection method i.e., are there areas that were identified by the method that perhaps don t represent actual change? Answer: Clouds are the biggest source of false positives, the most notable being at the bottom center of the masked area. This could lead to a discussion for strategies for identifying clouds (e.g., by irregular shape or using multiple time images). There are also a few examples of changed roof color, which will register as change, but students might note that this would not likely represent a change in imperviousness. 16

17 Question 10: Do you observe any other patterns in how the change raster relates to what you see in the two multiband images? Answer: Students might recognize that additions of dark surfaces are not highlighted as much as the addition of light surfaces. Compare, for example, the regions covered by cloud shadow, which are highlighted in the middle of the spectrum change. Students may also discover new buildings with dark roofs that are not prominently identified as change areas. Comparing two different surfaces that both have lower reflectivity across all bands will indeed result in a lowermagnitude change vector, and this might be particularly important given how much of impervious surface is often dark in color. This could prompt discussion about algorithmic improvements including the incorporation of the change vector direction or principal components analysis (PCA). Students may also notice that there are a large number of either individual pixels or small groups, and that these are often associated either with shadows or with misalignment or different angles in the imagery. This can be particularly useful in illustrating some of the issues associated with higher resolution imagery and could lead to a discussion of how to clean up the resulting visualization, including nearest-neighbor techniques. Question 11: What additional or different processing would you suggest to improve the method? Answer: You may want to encourage students to look at the individual delta bands representing the components of the change vector and to explore other functions available under the Spatial Analyst menu. Good students may be able to discover strategies already established in the literature such as image ratioing (dividing two temporal rasters instead of subtracting), deriving a new component index (e.g., the vegetation index = nir - red / nir + red, which is used for vegetation classification), nearest-neighbor filtering techniques, or if they have a solid grasp of spectral space, analysis that includes the change vector direction component or involves principal components. 17