Name of the labAuthor name
Paving the Planet
Impervious Surface as Proxy Measure of the Human Ecological Footprint
― by Paul C. Sutton
Introduction
Problem
This laboratory exercise will focus on three related questions:
1)“How could pavement be modeled, mapped, and measured using global satellite imagery?”
2) “How much of the earth’s surface have human beings paved over?”
3) “Is there a relationship between ‘pavement’ and the human ecological footprint?”
Questions associated with the idea of “sustainability” are of increasing importance in the 21st century. Ideas of ‘carrying capacity’ and the human ‘ecological footprint’ attempt to rationally quantify how sustainable our collective human behavior and activity is from a bio-physical or ecological perspective. This exercise develops a spatially explicit proxy measure of pavement (aka ‘anthropogenic impervious surface’) and compares nationally aggregated measures of impervious surface to state-of-the-art measures of the human ecological footprint. Many scientists and policy makers believe it will be increasingly important to develop and use several independent and spatially explicit measures of human impact on the Earth. Measurements of this nature will inform and focus policies thatguide humanity toward a more sustainable future.
Location
The geographic scope of this laboratory exercise is global.
Time to complete the lab
Completion of this laboratory exercise should take 2-3hours.
Prerequisites
An understanding of elementary GIS (preferably ArcMap), and basic statistics and mathematics.
Keywords
Topical: (Ecological Footprint; Carrying Capacity; Impervious Surface, DMSP OLS, nighttime satellite imagery, population density).
GISc: Concepts / tools / techniques used (raster calculator, cell based modeling, zonal statistics, Table Joining, linear regression) and types of data used (grids and polygons).
Data used in this lab
National boundaries of the world’s countries (ESRI)
Global Nighttime Satellite Image (NOAA and the United States Air Force)
Landscan: A Global Grid of Population Density(Oak Ridge National Laboratory)
Geographic coordinate system:GCS_WGS_1984
Datum: D_WGS_1984
Projection: Geographic Coordinate System
Data must have no copyright restrictions; otherwise, please provide data sources and permissions using the ESRI Press Copyright Permissions Form.
Student activity
Measuring and Mapping Pavement
It is recognized that human impact on the environment is a serious and growing concern. One of humanity’s most significant impacts on the environment is the conversion of various natural land surfaces to some kind of manmade impervious surface ( e.g. a road, a rooftop, a parking lot, etc.). An interesting question to ask is: “How much of the planet have we paved over?”. Satellite imagery used in a GIS framework is an excellent means for exploring this question. However, high resolution satellite imagery that we can find in Google Earth () would be overwhelming to analyze at a global scale. In fact it takes about one hour to classify ~200 points in a one square kilometer area (Answer Q1). One of the reasons we develop statistical models to map phenomena like impervious surface is because it is impractical to use high resolution satellite imagery for the whole world.
So, from the question: “How much of the Planet has humanity paved over?” we move on to another question: “How can we make measurements or estimates of impervious surface without getting bogged down in detailed classifications of millions of square kilometers of high resolution satellite imagery?”(Answer Q2). The approach we are going to take in this lab is to use two datasets that we believe serve as proxy measures of impervious surface or pavement: 1) A global nighttime satellite image composite and 2) a global raster dataset of population density (Figure 1). The theory is this: 1)High population densities will produce more ‘pavement’ than low population densities; and, 2) Well lit areas in a nighttime satellite image will be more lit in areas of high transportation activity, economic wealth, and urban development – all of which are proxy measures of pavement. Now, this is a nice theory. We do need to demonstrate the validity of this theory with some empirical observations that do indeed involve the tedious classification of thousands and thousands of points as ‘developed’ (e.g. impervious surface) or ‘not developed’ (e.g. natural surface) in high resolution images. Consider yourself lucky that you are not going to be asked to do any of this classification work. However, it is worthwhile for you to understand how one might go about performing such a task and what the resulting data would look like. The data is what we need to demonstrate that our theory is empirically valid and what we will need to develop a model that we can apply to our global datasets of nighttime imagery and population density.
Getting data to prove Concept and build a Statistical Model
Reasonably intelligent human beings are pretty good at identifying whether or not a specific point in a high resolution satellite image is a developed impervious surface or not. A team of scientists at the National Geophysical Data Center of NOAA in Boulder, CO has developed a web application to allow people to do just that ( ) (Figure 2). The ‘Impervious Surface Mapper’ is a web application that utilizes a Google Earth API that allows a user to zoom around ‘Google Earth’ style to any area of the world (Figure 2). The user can toggle the nighttime satellite image on and off to help identify urban areas. Once the user finds an area of the world to classify he or she selects a center point and begins to classify 225 points in a one kilometer square pixel according to the classification scheme portrayed in figure 2.The Impervious Surface Mapper web application can generate a table that can be imported into a software package such as Microsoft Excel or a statistical software package such as JMP (Table 1).
The first column in Table 1 is“% Impervious Surface” which is simply the proportion of the 225 points in a given 1 km2 pixel that were classified as ‘paved’. The columns labeled “Light Intensity” and “Population Density” are the spatially corresponding values from the nighttime satellite imagery and the Landscan dataset.
Table 1: Typical Data generated by the Impervious Surface Mapper
Using a dataset like Table 1(that has been derived from the classification of tens of thousands of points using the Impervious Surface Mapper), we can obtain numbers we call ‘regression parameters’ that we use to predict the fraction of a grid cell that is ‘impervious surface’ from both the light intensity and the population density. This statistical approach is called multi-variate regression. Without going into detail on the statistics, the simple result is the following model:
Predicted % Impervious Surface = (0.0795)*Light Intensity + (0.00868)* Population Density (equation 1)
The numbers 0.0795 and .00868 are the regression parameters derived from a statistical analysis of data that look like the numbers shown in Table 1. What you will now do using the power of a Geographic Information System is apply these parameters to some global datasets to produce a global estimate of impervious surface which you can then analyze. To test your understanding of this of idea of applying parameters to a cell based model answer exercise question number three using equation 1 (Answer Q3).
Using ArcMap to create your model of Impervious Surface
Getting to know your data and the ArcMap display functions
Start ArcMap by either clicking on the icon on the desktop or by going to ‘Start’ > ‘All Programs’ > ‘ArcGIS’ > ‘ArcMap’. Open the mxd titled: Pave_The_Planet.mxd (mxd is the filename extension for an ArcMap map document). If you are successful in opening the file you should get a window that looks something like the image below (Figure 3). Now is a great time to explore your data. Play around with the display tools in ArcMap. Use the zoom tool to zoom in on parts of the world you know. Turn layers on and off to allow you to see how they are stacked (the layers in this case are: World_Nations.shp, popden_ls, and nitelites_dmsp_mw.tif). Notice the spatial correspondence between areas of high population density and high levels of light intensity as measured by the nighttime satellite image mosaic. Become familiar with turning layers on and off, changing the display of the layers (right click on the layer, choose properties, and select the symbology tab), and pan and zoom around (Answer Q4).
Use Map ALgebra to build a global model of Impervious Surface
Here you will take advantage of the computing power of a GIS to perform the calculations you did by hand in question three - millions of times (e.g. for every land pixel of the world).
Here you will use an application called “Single Output Map Algebra” in the ‘Spatial Analyst Tools’ of Arc Toolbox. Figure 4 shows you what the tool that you can build the model with looks like (Figure 4). Use the equation below as your ‘Map Algebra Expression’:
int((.0795 * nitelites_mw) + (.00868 * popden_ls))
For your ‘output raster’ create a grid in your data_mollweide folder called ‘isa_model’. The processing of this model may take a couple of minutes – be patient. When the process is complete a new layer titled ‘isa_model’ will appear in the table of contents of your data frame. Using the pan and zoom tools and the information tool explore your model of Impervious Surface. The model was built so that the numbers in the grid cells represent the percentage of that pixel that has been converted to impervious surface by human activity. Many of your cells have values that are over 100 which is physically impossible (Answer Q5).
USE ZONAL Statistics to calculate total pavement per country
You now have a ‘representation’ of impervious surface (aka ‘pavement’) at roughly 1 km2 spatial resolution for the entire surface of the world. Now use the ‘zonal statistics’ function of spatial analyst to create a table that tabulates the total area of pavement for each country of the world. This is a basic ‘overlay’ function of a geographic information system in which a vector coverage (World_Nations.shp) is overlaid on a raster coverage (ISA_model). To do this go to the ‘spatial analyst’ pull down menu and choose ‘Zonal Statistics’. You should get a window that looks like Figure 5. Fill out the window as Figure 5 is filled out. The ‘Zone Dataset’ is the vector coverage of country boundaries. The ‘Zone Field’ is the column of the attribute table that you want to be used for summing the impervious surface of. In other words, you are asking the GIS to add up all the impervious surface grid cell values on a per country basis (CNTRY_NAME). The ‘Value Raster’ is ISA_model, in other words you are telling the GIS to add up the impervious surface grid cell models as opposed to the population density values (popden_ls) or the light intensity values (nitelites_mw). Once you click ‘OK’ be patient, this may take a minute or two. A table similar to the one below should be displayed. The
columns have the following meanings: Count (# of pixels in that country), Area (areal extent of country in square meters), MIN (lowest ISA pixel value found in that country), MAX (highest ISA pixel value found in that country), Range (max ISA pixel – min ISA pixel in that country), MEAN (average ISA pixel value in that country), STD (standard deviation of ISA pixel values in that country), SUM (total sum of ISA pixel values in that country), Variety (# of different ISA pixel values in that country), Majority (most common ISA pixel value in that country), Minority (least common ISA pixel value in that country), Median (median ISA pixel value in that country). You are most interested in the SUM column. It is not a measure of area of pavement exactly but it is directly proportional to such a measure. You can export this table to a spreadsheet package such as Microsoft Excel by clicking on ‘Options’ at the lower right of the table and choosing ‘Export’. Do this and say yes when it asks to add this table to your ‘current map’. ArcMap will also allow you to manipulate and query this table. For example, right click on the SUM column header and choose ‘Sort Descending’. This should rank the countries from highest amount of impervious surface to lowest (Answer Q6).
Compare ‘ISA’ for each country to their Ecological Footprint
You are now going to join a table that contains Ecological Footprint information for each country of the world to the table you created with Zonal Stats. The Eco-Footprint data came from the Global Footprint Network (). The purpose of this exercise is to see if there is a relationship between the amount of land a country has paved over and that country’s ‘ecological footprint’ as calculated by the Global Footprint Network (GFN). Choose the ‘Source’ tab at the bottom of your ‘Table of Contents’ on the left side of your ArcMap window. The Table you created from your Zonal Statistics operation should be visible. Now add a table called Nation_Ecofootprint.xlsinto your current map. Right click on the table (in the table of contents) that you created using the Zonal Statistics function and choose ‘Join’. You should get a dialog box that looks like the figure to the right (Figure 6). Clicking ‘OK’ will merge the two tables using ‘CNTRY_NAME’ as a ‘key’ column such that the GFN’s ‘EcoFootprint’ number for each country will be ‘joined’ to the original table you made using Zonal Statistics. You now will have a table that has an ISA number (SUM) and an EcoFootprint (EcoFootprint) number associated with all of the140 or so countries that the GFN calculated an Ecological Footprint for. Try to export the table you have joined and read it into a package you are comfortable doing analysis with. If you like you can use ArcMap to make scatterplots. From the ‘options’ button you choose ‘create graph’ to explore relationships between variables. I whatever way you feel most comfortable try to explain the relationship between impervious surface area (ISA) of a country and the Ecological Footprint of that country as measured by GFN. For ease of analysis you can use the table titled Nation_ISA_EF_Pop_Ext.xls that is in the data_mollweide directory. This contains ISA, Population, EcoFootprint, ISA/person, EcoFootprint/person, ln(ISA), and ln(EcoFootprint) data for 137 countries. Using either your data or the data in the Nation_ISA_EF_Pop_Ext.xls file provided, explore and explain the relationship between impervious surface and ecological footprint (answer Q7).
Submit your work
Answer and/or Solve the Following Questions and Problems
Q1: If it takes 1 hour to classify 200 points in a 1 square kilometer area, how long would it take to classify all of the earth’s land surface with a similar level of analysis?
Q2: Other than what we use in this lab (nighttime lights and population density), what easy to measure phenomena do you think would make good spatially explicit measures of impervious surface or pavement?
Q3: Using the 3x3 grids representing light intensity and population density below and the model (equation1)of impervious surface to produce a corresponding 3x3 grid cell estimate of the impervious surface percentage of the corresponding cells (see figure below).
Q4: Describe the three ‘layers’ in the Pave_The_Planet.mxd. All three layers are in a Mollweide projection. The mollweide projection is one of the better global equal area projections. Why might it be important to use an equal area projection in this study? Approximately how large are the grid cells in the nitelites_mw and popden_ls datasets? What is the average cell value of the nitelites_mw and popden_ls datasets?
Q5: The layer you have created called ‘isa-model’ has some grid cell values over 100%. How do these errors manifest? How serious of a problem are they? What fraction of the pixels have values in excess of 100% ? (To figure this out right click on the isa_model layer and open its attribute table, this attribute table is called a VAT (Value Attribute Table). Read the help system on the VAT) Do the spatial patterns of impervious surface make sense as you display areas of the world that you know well? Zoom in on Los Angeles, California. What is the highest grid cell value in the Los Angeles area that you can identify? Fire up a web browser such as Mozilla or Internet Explorer and use the Impervious Surface Mapping tool to look at high resolution imagery of some area you know well. How reasonable an estimate of impervious surface percentage was your model of this area. Explain.
Q6: According to the model of ‘pavement’ (aka ‘Impervious Surface’) what are the top ten countries of the world in terms of area of pavement? Does this make sense to you? Of these countries, which country has the highest pavement per capita? Which country has the lowest pavement per capita? Using the average nitelites_mw number and average popden_ls number from Q5 and the regression parameters you used in Q3, determine the relative contributions of nighttime lights and population density to the estimate of impervious surface globally. The ‘zonal statistics’ function added up all the ‘percent impervious surface’ values in all the pixels of each country. What do the total numbers mean? How could you convert those numbers to a total area of impervious surface for each country?
Q7: What is the ‘Ecolgical Footprint’ (EF) of a nation? How does the Global Footprint Network calculate the Ecological Footprint of a country? Why do you think there might be a relationship between the Impervious Surface Area (ISA) of a country and that country’s ecological footprint? Describe or display the scatterplots of ISA and EF; ISA/capita and EF/capita; ln(ISA) and ln(EF). Do these relationships appear to be significant? Explain the database procedure or ‘trick’ of the table join. What is the ‘key’ column and how is it used? What countries of the world have the greatest ‘Ecological Footprints’? Why might a simple measure of ISA as used here be useful as a proxy measure of the EF developed by the Global Footprint Network?