IntroductionVirtual globes are relatively recent phenomena in the computer world, but they are quickly becoming ubiquitous and popular. Their usefulness for geographic education is unquestionable, but the educators must be informed consumers of this technology and be aware of both the capabilities and limitations of virtual globes. Knowing what makes virtual globes different form Geographic Information Systems (GIS) and online mapping applications is also important. Learning how to use virtual globes and analyze spatial data that they depict may be an overwhelming task at first. There is a lot of information available on Internet on digital globes (blogs, user communities, etc.), but it is much dispersed. Our article is an attempt to consolidate and summarize the most important up‐to‐date information on virtual globes, with an emphasis on their educational applications. In this Teaching and Learning Guide we provide an annotated list of the best resources that exist on digital globes and focus questions for classroom discussions on this subject. We also provide four practical exercises with step‐by‐step instructions and color illustrations. These exercises do not use any GIS software and can be used in various geography and social studies courses.Key Readings Riedl A, 1999, Virtual globes: A new era for globes?, ICA 1999 19th International Cartographic Conference This paper provides an early idea of a ‘virtual globe’ that can use multimedia capabilities like sound, animation and interactivity that can enrich the way of visualizing natural phenomena and processes. Aurambout J P, Pettit C and Lewis H, 2008, Virtual globes: the next GIS? In: Landscape analysis and visualisation. Heidelberg: Springer Berlin, pp. 509–532 This paper compares the strengths and weaknesses of five major virtual globes and evaluates their capabilities to present the GIS data is the context of agricultural sciences and natural resources management. It shows the present limitations of virtual globes in visualizing the GIS data and suggests several improvements in the virtual globes. Brown M C, 2006, Hacking Google maps and Google earth. New York: Wiley This book provides a step‐by‐step tutorial for creating applications to make maps that reveal statistical data, plot and calculate distances for routes, highlight archaeological information etc. Butler D, 2006, Virtual globes: the web‐wide world. Nature (439): 776–778 With the help of various examples, the article shows that online tools like Google Earth and other easy‐to‐use virtual globes, are changing the way we interact and communicate with spatial data. Goodchild M, 2005, What does Google earth mean for the spatial sciences? Biennial Conference of the Spatial Sciences Institute. The author discusses technical aspects of showing GIS data and satellite images on the virtual globes. He named virtual globes as ‘Global GIS’ and said that the VG has the potential to extend spatial science to a much larger community of social scientists, users and students. Tuttle B T, Anderson S, and Huff R, 2008, Virtual globes: an overview of their history, uses, and future challenges. Geography Compass 2 (5): 1478–1505. This paper provides a commentary on the rise of virtual globes, reviews the current literature available and identifies the areas that require more research.Focus QuestionsThese focus questions/topics can be used in the context of any introductory mapping or GIS class to initiate the discussion of the importance and the limitations of virtual globes. How have virtual globes changed the way we visualize spatial data? What are the limitations of virtual globes? Can virtual globes replace GIS? What are the implications of inaccuracies in the data in virtual globes? Virtual globes and the digital divide Examples of Practical ExercisesBelow we provide four student activities that illustrate the use of three virtual globes (Google Earth, Virtual Earth, and Skyline Globe) in education. We believe that the rich visual environment of virtual globes enables students to frame interesting questions, and to visualize and analyze various spatial layers, produced and overlaid without any GIS software. The first two activities are designed for a novice user of virtual globes. Activity one helps in delineating the extent of urban sprawl in Long Island, NY using Skyline Globe. The second activity allows to visualize the country wise changes in internet use through time using Google Earth. The last two activities and intended for more advanced users as they require familiarity with Microsoft Excel (activity 3) and concepts of image classification (activity 4). Activity 3 helps in mapping the toxic release inventory (TRI) of the Worcester County on Massachusetts and finding out the economic status of the population living near the TRI sites the using MS Excel and Google Earth. Activity 4 shows how the students can use virtual globes for their various remotely sensed image classification processes.The Links to downloading instructions for the four globes are listed in the Appendix of the original paper. Activity 1: Estimating the extent of urban sprawl In this activity we will use interactive tools of virtual globes (on‐screen digitizing and area calculations) to estimate the extent of urban sprawl in Long Island, New York. Here virtual globe is used to perform tasks that were previously only available in GIS software. Required Software Skyline Globe Process Long Island, is it the largest island in the continental United States. We will use virtual globe to calculate the percentage of urban land from the total area of the island. Open Skyline Globe and search/zoom to Long Island, NY. Click on the Layers tab at the top of the screen and check the ‘Urban sprawls (US)’ check box in the left hand side panel. You’ll see the urban sprawl in Long Island in red color (Figure 1a). To find the area covered by urban sprawl, click on the Measure tab and select ‘Area measurement’ in the left hand side panel (Figure 1b). Digitize various urban sprawl areas on Long Island and write down the area covered by each of them. Add up the areas of the sprawl together. Find the total area of Long Island from literature/Internet. Calculate the percentage of area covered by urban sprawl in the Long Island. Estimating the extent of urban sprawl. (a) Urban sprawl in Long Island, NY. (b) Digitizing various urban sprawl areas on Long Island, NY. Application courtesy: http://www.skylineglobe.com image Questions to answer Is Long Island the most urbanized island in the United States? What factors contribute to high urbanization on this island? Use additional Internet resources to answer these questions. Activity 2: Visualizing global Internet use through time Using this activity we shall find the rate of increase in internet use around the world by creating a time‐series animation in Google Earth. We shall create a country wise dynamic world map of internet users per 100 people between the years 1990 and 2005 and look for trends in number of users over the years. The internet use data come from United Nations Statistics Division (http://data.un.org), which provides country level data on education, environment, energy, food, population, health etc. Any of these data series can be visualized in Google Earth using the following steps. Required Software Google Earth Process To create the thematic kml of internet users, go to http://thematicmapping.org/engine/Select ‘Internet users per 100 population, in the Indicator tab (Figure 2a). Select 1990 as the year. Select Prism in the Technique tab. Select yellow as the no value color. Select start color as red and the end color as green. Select equal intervals in the classifications tab. Select Time Slider in the Time tab. Select show color legend in the display tab. Click on ‘Download’. A kml file will be generated and then you will be asked to save it. Open saved kml file in Google Earth. It will open with a legend and a time slider (Figure 2b). Notice that in 1990 not many countries in the world had Internet. Which are the countries that had internet at that time? Where are they located? Slide the time slider to see the change/increase in Internet users per 100 in population around the world. You can also play the animation by clicking the ‘Go’ button on the right‐hand side of the time slider. To change animation speed click on the ‘clock’ button on the left‐hand side of the time slider. Visualizing global internet use through time. (a) Creating thematic kml of internet users. (b) Graphical representation of number of internet users on Google Earth. Application courtesy: http://www.thematicmapping.org image Questions to answer Notice that some of the smaller island nations have high penetration of Internet. What might be the reason for that? Name five countries with the high number of internet users per 100 people. Activity 3: Toxic release inventory mapping The aim of this activity is to graphically show the location and the amount of toxic releases from sites registered with the Toxic Release Inventory (TRI) that contribute to cancer hazards in Worcester County, MA, USA. We shall also use the census 2000 data to look at the income distribution of the people living near the TRI sites. This activity can be used in any environmental science or medical geography course. The students will learn to create and overlay data layers in Google Earth without using any GIS software. Map analysis will allow students to identify areas where high volumes of releases contributing to cancer risk and marginalized populations overlap in space. Please refer to Figure 3.Toxic release inventory mapping. (a) Copying TRI tabular data from the website. (b) Concatenating TRI data in Excel. (c) Adding column names in Excel. (d) Copying TRI data to batchgeocode. Application courtesy: http://www.batchgeocode.com (e) Validating addresses in batchgeocode. Application courtesy: http://www.batchgeocode.com (f) Downloading the geocoded TRI sites as a kml file. Application courtesy: http://www.batchgeocode.com (g) Checking the attributes of a TRI location in Google Earth. Application courtesy: Google Earth. (h) Copying geocoded address from batchgeocode to Excel. Application courtesy: http://www.batchgeocode.com (i) Formatting the columns in Excel. (j) Using GE graph. Application courtesy: GE Graph: htttp://www.sgrillo.net/googleearth/gegraph.htm (k) Displaying TRI graphical data on Google Earth. Application courtesy: Google Earth. (l) Using GE Census. Application courtesy: http://www.gcensus.com (m) Displaying median household income in Worcester City for 1999 per census tract with TRI graphs. Application courtesy: Google Earth. Data courtesy: http://www.scoreboard.org imageimageimageimageimageimageimageimageimageimageimageimageimage Required Software Google Earth, Microsoft Excel, GE Graph, and Batch Geocoder. Process The location of the TRI sites in Worcester County can be accessed from http://tinyurl.com/trisites. Copy the tabular data from the website and paste that into a new text document and save (Figure 3a). Open the text document in MS Excel using Tab and Comma delimited options. Concatenate the last two columns to column G using this formulae:G2 = concatenate (E2,F2). Check the results of the operation and make changes if necessary (Figure 3b). Add a row on the top and name the columns as number, name, address, city, value1, value2, amount, state (in this order) (Figure 3c). Populate the last column with Massachusetts as the state name (Figure 3c). Copy the data from Excel and paste it on the following website: http://www.batchgeocode.com/. This website provides free geocoding facility (Figure 3d). Click on ‘Validate Source’ to verify that your data is read correctly and then click on ‘Run Geocoder’ (Figure 3e). Once geocoding is done, notice that two new columns – latitude and longitude – are added to the data. The geocoding operation will show location of the TRI address on a map as point locations. Click on ‘Download to Google Earth (KML) file’ and save the kml at an appropriate location in your computer (Figure 3f). This kml file can be open in all four virtual globes and it will show the location of the TRI sites. Clicking on the individual points will show the amount of toxic release associated with them (Figure 3g). Download and install GE graph from http://www.sgrillo.net/googleearth/gegraph.htm. This tool creates graphs that can be used for displaying attributes of spatial data in Goggle Earth. We will use it to create a bar graph showing the amount of toxic release from the sites. Copy all the data from step 9 (Batch geocoder results window) and paste it to a new text document and save it as ‘TRI_Graph’. Open the txt file ‘TRI_Graph’ in Excel as tab delimited text (Figure 3h). Format the data such that only four columns remain on the sheet (in the prescribed order): name, latitude, longitude, amount (Figure 3i). Copy the data from Excel, do not copy the first row with column names. Start GE‐Graph, click on ‘paste grid from clipboard’ to insert the data from Excel (Figure 3j). Select size as ‘constant’ (Figure 3j). Give size as ‘50’ meters (Figure 3j). Click on ‘Run’, the result will be a kml file that can be opened in Google Earth. This kml will extrude the TRI points as 3D columns on the basis of the ‘Value’ field. So, the site with the highest amount of toxic releases will have the highest column. Once the data is displayed graphically, look at spatial patterns and answer the following question: ‘What parts of the city are exposed to the highest amounts of toxins?’ (Figure 3k). Now we will visualize census data to know the income level of the people living near/around the TRI sites. Go to the following website: http://gecensus.stanford.edu/gcensus/ajax/gcensus.php using Firefox (Note: This site does not work well with Internet Explorer). Select US Census 2000 Summary File 3 > State > Massachusetts > County > Worcester County > Housing Subjects Summarized to Census Tract Level > Unified Over All Races > Household Income in 1999 > Median Household Income In 1999 (Dollars) By Tenure > Total (Figure 3l). Click on ‘Click here to download your map’ and give an output file name for the kml file. Open the file in Google Earth to show the median household income in Worcester City for the year 1999 per census tract. The data is divided into five classes (Figure 3m). Questions to Answer Look at the map in Google Earth and answer the following question: ‘What income group corresponds to the highest level of toxic exposures versus areas with no exposure (far away from TRI sites)?’ Activity 4: Image classification and virtual globes The aim of this activity is to show how to incorporate data from virtual globes into the following image classification processes: (i) identification of clusters from unsupervised classification, (ii) training sites delineation for supervised classification, and (iii) classification accuracy assessment. This approach has several advantages: it allows for more accurate delineation of the boundaries of training sites thanks to the high resolution imagery available in virtual globes; and it can potentially save enormous amount of time and effort associated with field work. We use data for Central Massachusetts in this activity, but these steps can be used for any area of the world for which there is a detailed image coverage within virtual globes (see section ‘Data available in virtual globes’ above). Required software Any image processing software, Google Earth or Virtual Earth, ‘Export to KML’ tool (free). Any commercial (e.g., ERDAS Imagine, Idrisi Andes, ENVI, Geomatica) or free image processing software can be used in this exercise for image classification. The following free packages are widely known and can be used in this exercise: Multispec –http://cobweb.ecn.purdue.edu/~biehl/MultiSpec/TNTlite –http://www.microimages.com/tntlite/MIPS –http://homepage.ntlworld.com/paul.mather/ComputerProcessing3/MIPS_Software_Updates.html Data Landsat ETM+ data downloaded for free from Global Land Cover Facility Process Go to Global Land Cover Facility's Earth Science Data Interface (http://glcfapp.umiacs.umd.edu:8080/esdi/index.jsp) and search for ETM+ imagery for path 012, row 031. From the resulting list of 37 scenes select scene with ID number 074‐814 (the acquisition date for this image is 31 July 2002) and click on the Download button at the top of the page. This scene consists of seven bands of multi‐spectral images and one panchromatic band, all in TIFF format. Download all files except for the panchromatic band (it has the largest file size) and unzip them. Use remote sensing software tools in import TIFF files into the format necessary for image processing. Use clipping tools to select a smaller area from the entire scene for each band using these coordinates for the bounding rectangle (in latitude/longitude in WGS84):minimum X: −71.90maximum X: −71.75minimum Y: 42.20maximum Y: 42.35This smaller area corresponds to Worcester, the second largest city in the state of Massachusetts. Run unsupervised classification using multi‐spectral bands and the default clustering algorithm offered by your software. Display the resulting image alongside with the false color composite created from bands 2, 3 and 4. Zoom in both images into a smaller area in the Northwestern part of the image, near these coordinates: X = −71.86, Y = 42.30. You will see several clusters near the lakes, corresponding to various shades of red in the false color composite image. As you can see, these are clusters of natural vegetation. Let's see how this area looks on a virtual globe (e.g., in Virtual Earth). Open Virtual Earth by going to http://maps.live.com and search ‘Worcester, MA’. Zoom to the area of our interest and try to identify land cover from the imagery by clicking on ‘Bird's Eye’ imagery. The imagery used in Virtual Earth was taken during the season when deciduous trees lost the leaves (‘leaf‐off period’), so patches of evergreen vegetation stand out. Reposition two windows side‐by‐side on your screen – the classified image from the image processing software and the Virtual Earth window. Compare visually clusters with what you see in Virtual Earth. If needed, further zoom into the image and if there are shadows in the image then you can use the ‘rotate’ tool in Virtual Earth to see the location from four different angles. Use editing tools of your image processing software to label these clusters as deciduous and evergreen forests. Continue zooming into various sections of the classified image and comparing them to the view in Virtual Earth. Virtual Earth's very high resolution Bird's Eye images proved to be a very handy, time‐saving tool for labeling clusters. The same approach can be used with other tasks of image processing and analysis. For example, it is possible to use ‘Virtual Earth’ as a guide in training sites development for supervised classification. With both windows displayed side‐by‐side, on‐screen digitizing can be done within image processing software using imagery from Virtual Earth. Similarly, Virtual Earth can be used as a surrogate for field work required for accuracy assessment after an image is classified. The general steps for this task are as follows: Generate a set of sampling points for accuracy assessment using your GIS or image processing software and stratified random sampling scheme. If necessary, convert these points into shapefile format and use free ‘Export to KML tool’ in ArcGIS to create a kml file of these sampling points. Open this kml file in Virtual Earth by clicking on Collection > Open your collections > Import and record land cover classes for each point in a table. This table will have two columns of data – the ID of the sampling point and the corresponding land cover class. Based on this table, create a ground truth image in your image processing software. Perform accuracy assessment using information from the ‘virtual field work’ and the classified image. There is one limitation to this approach: the classified imagery and the imagery available via virtual globe may be several years apart (virtual globe imagery being more recent in most cases). So, there may be areas that have changed between the two dates, such as a new residential development appeared where forest used to be. Therefore, a careful visual examination and comparison of the classified imagery and the virtual globe should be done first, and locations with disagreements in shapes, spatial patterns, and expected spatial reflectance pattern should be noted and not used in the subsequent analysis.