| Exploration 1 |
| Explore the impact of rising sea levels as a consequence of
climate change. |
| One of the consequences of climate change will be rising sea
levels due to the melting of land-based glaciers. If all the
glaciers in the world were to completely melt, this could raise
sea levels by 80 meters (Williams 1999*). Let's explore the consequences
of sea level rise for the low-lying country of Bangladesh using
TerraViva! Global Data Analyst. The Elevation and Depth map provides
topographic information for this exploration and the Gazetteer
helps us quickly locate Bangladesh on the map of the world. Using
Create Mask, we'll mask areas under five meters in elevation. |
|
| Next, we'll apply the mask to the Population Density map and
use the Spatial Query tool to identify the number of people living
in the flood-risk area. The flood-risk area extends into eastern
India, but for this example we'll query Bangladesh only. The
query will employ data from the mask and data from the Population
Density map. Approximately 7.5 million people are living in Bangladesh
at elevations below five meters. |
|
| Finally, we'll identify the ecosystem type that will be inundated
by sea-level rise. We'll apply the flood risk mask to the Global
Ecosystems map. However, this time we'll invert the mask to more
effectively show the at-risk area. Using Spatial Query again,
we see that the ecosystem at greatest risk in Bangladesh is identified
as "Rice paddy and field."
Though flood-risk analysis is a complex endeavor that involves
more than just elevation studies, this example provides a starting
point for discussion. |
|
* Williams, Richard S., and Jane Ferrigno. (1999) “Estimated
Present-Day Area* and Volume* of Glaciers and Maximum Sea Level
Rise Potential” From: Satellite Image Atlas of Glaciers
of the World, Chapter A: Introduction, U.S. Geological Survey
Professional Paper 1386-A. Available at http://pubs.usgs.gov/fs/fs133-99/gl_vol.html.
TerraViva! Explorations |
| |
|
| Exploration 2 |
| Explore infant mortality rates worldwide, and development
variables that may influence infant mortality rates. |
| Infant mortality rates vary considerably around the world,
from a low of under 5 deaths per 1,000 live births to a high
of close to 150. To get a sense of the regions of the world with
the highest infant mortality rates, we'll use TerraViva! Global
Data Viewer (or TerraViva! Global Data Analyst) to produce a
choropleth map showing infant mortality rates (IMRs). We'll employ
the World Resources Institute dataset, People and Ecosystems,
to map the levels of infant mortality. Clearly sub-Saharan Africa,
Central Asia, and part of Central America have the highest levels.
The compare feature generates a ranked list of countries with
their infant mortality rates. This data table and the dynamic
map show that 9 out of the top10 highest IMR countries are in
Africa. |
|
| Now let's hypothesize which development variables would be
potential contributors to high infant mortality rates: GDP per
capita, calorie supply per capita, female literacy rate. But
for this example we'll choose "Access to an improved water
source" and test our hypothesis by exploring this variable
in relation to infant mortality on a scatter plot. Remember,
although we may find a high correlation between the independent
and the dependent variable (IMR), this does not necessarily imply
causality. The scatter plot shows "Access to an improved
water source" on the X axis, and "Infant Mortality
Rate" on the Y axis. Once we've created a regression line
we find that the R2 is 0.528. We could choose to animate the
scatter plot as a function of time. |
|
| Now let's identify outliers by clicking on the dots that lie
well away from the regression line and let's hypothesize why
they do not conform closely to the expected relationship. The
dot in the upper right quadrant corresponds to Burundi. Clicking
on the dot reveals variable details specific to Burundi. To continue
our investigation of Burundi we'll open the Gazetteer for information
about Burundi that may explain why it is an outlier.
Using plotting tools, data sets, and the Gazetteer we could
examine any development variables that may be contributors
to infant mortality rates. |
|
| |
|
| Exploration 3
|
| Identify variables with a high correlation to infant
mortality rates using multivariate regression analysis. |
| Your research may frequently lead you to a global or regional
issue that has many possible contributing factors. The ability
to quickly calculate the influence of multiple variables on the
issue under investigation and to eliminate non-contributing variables
can play a key role in the development of predictive models.
The Multivariate Stepwise Linear Regression (MVR) tool in TerraViva!
Global Data Analyst enables you to quickly: identify variables
that are statistically significant; produce a mathematical formula
that represents the relationship between the variable of interest
and the variables of influence; and plot the relationship. |
|
| Let's use multivariate regression to determine which of the
following factors most closely correlates with infant mortality
rates: GDP per capital; electricity consumption; household income;
unemployment rate; telephones in use; and military expenditures.
Using World Factbook 2004 dataset we set the dependent variable – “Y” in
the images on the right – as “Infant Mortality rate:
total.” And we select our six independent variables – “X1” through “X6” in
the images – as described above. As we select each variable
we can apply a logarithmic transformation to the variable in
an attempt to produce a decent bell curve and can choose appropriate
normalization options. The “View Plot” pane gives
us a look at the correlation of each selected “x” variable
to infant mortality and displays the results as a histogram or
scatter plot. |
|
| After selecting all variables we initiate "forward" regression
analysis. Forward analysis cumulatively incorporates each independent
variable in order of strongest correlation. As we progress through
each step we can see the relationship expressed as "R Squared" and "Adjusted
R Squared." The results leave us with a good two-variable
model showing the strongest correlation using GDP and household
income: an "R Squared" equal to 0.867. A strong correlation
does not necessarily prove causality. Statistics is a discipline
that requires a skilled and artful interpreter and it would be
wise to use other supportive evidence in a rigorous examination
of infant mortality rates. Effective use of multivariate regression
does offer a good starting point for broader investigation into
causality. |
|
| |
|
| Exploration 4 |
| Import a map showing vegetation deviation in Thailand. |
Data-driven maps can give a closer look at “your” corner
of the world, or offer deeper insight into a particular global
theme. To take advantage of the many visualization and analysis
tools available in TerraViva! Global Data Analyst you may want
to import maps that were created from other sources. To use these
maps, the original data files are simply converted to a TerraViva!
readable file format which is then imported, allowing the map
to be viewed, labeled, masked, and mined. TerraViva! MapConverter,
a companion program included with TerraViva! Global Data Analyst,
enables you to convert map data files into a readable file format.
|
|
| For this exploration we'll convert and import map data for
a vegetation deviation map of Thailand. First, we'll open the
TerraViva! Map Converter program. The wizard prompts us to browse
for the desired map data file, called ‘ThaiVeg.flt’.
Subsequent prompts (see top image) require us to enter map-specific
parameters: map name; data type; map projection; map information
(pixel size, map height/width, map coordinates); data range;
scaling value; and, parameters for converting, such as pyramid
levels and color table selection. Once all parameters have been
entered the conversion takes about three minutes. The result
is a new file called ‘ThaiVeg.xtvm’. The conversion
process could have taken much longer, depending on the map size
and map projection. |
|
| Now that the map data file of Thailand has been converted into
an ‘*.xtvm’ file we import it into TerraViva! Global
Data Analyst and take a look. The Import Map File menu (middle
image) prompts us to browse for the new “*.xtvm’ file.
Once imported (bottom image) we see the new map as a rectangle
that includes Thailand and portions of the surrounding countries.
We can use the zoom tools for a closer look and use the map legend
for reference. As we move our cursor around on the map we see
thematic information displayed on the bottom of the map in the
gray status bar. As we pause our cursor over any point on the
map the percent of normal vegetation for that pixel will be displayed
in a small text box. The vegetation deviation map, used along
with other imported maps of Thailand such as precipitation, temperature
and ground wetness deviation, offers a broad picture of that
country’s vegetation conditions. |
|
| |
|
| Exploration 5 |
| Explore changes in GDP per capita growth worldwide
to identify areas that may be experiencing economic stress. |
| GDP per capita growth can serve as an effective window into
the economic “soul” of a nation over time. A histogram,
a plot style that employs groupings or “bins” of
numbers, will aid this examination. A histogram is simply a bar
graph in which the height of each bar is proportional to the
frequency or relative frequency represented. The discrete statistical
groupings are called “bins.” In analyzing values
associated with any variable it can be illuminating to determine
distribution across a finite, linear set of groups. From a histogram
you can easily observe where, categorically, the numbers fall – perhaps
a good first step toward understanding changes on a global scale. |
|
| First, we’ll build a histogram using the variable “GDP
per capita growth” from the WDI (World Development Indicators)
database. The histogram in the top image shows that the greatest
number of countries experienced GDP per capita growth rates between –0.244%
and 3.62%, with the remaining countries falling on either side.
To explore this variable further, you could experiment with different
numbers of bins, could click on a bin to see which countries
occupied the bin, or could map a selected bin to get a global
visualization. You might also animate the histogram by cycling
through available years of data. This histogram “movie” would
give a quick visual history of changes in GDP per capita growth
that have occurred over time. After experimenting with binning
and animation, what conclusions might be drawn about these changes? |
|
| Finally, we’ll generate a simple data table indicating
country rankings based on GDP per capita growth for the most
recent year. We could scroll through the data table to examine
country rankings. And, we could reverse the order of the list,
as shown in the bottom image, so that we could easily see which
countries are experiencing negative growth. Since it is reasonable
to assume that positive growth rates are desirable, any country
experiencing a negative growth rate for any length of time is
worth further study. Using information gained from the animated
histogram and the data table, you might then begin to examine
possible causes of the economic stress and to anticipate the
social and political instability that may result from this economic
degradation. |
|
