Mapping
ArcGis is a fantastic tool for making maps to display collected data in an understandable way. These maps can be of geographic boundaries, biodiversity in an area, or any other trend that can be visualized through a map. ArcGis is an online program which takes sets of data as input and produces map images as output. All one needs to plot data into a map in ArcGis is latitude and longitude coordinates. Many sets of data which are available online come in the form of an excel spreadsheet or some form of tabular file. We call these either XLSX or CSV files. It is the data from these files that ArcGis will use as input. Once the file has columns for latitude and longitude, it can plot a map. However, most datasets will also have values associated with these coordinates which give the map its meaning. We call this vectorized data. A map of soil type in America might have values corresponding to sand in Southwestern desert regions, loamy clay in swampy areas of Florida, and hard rock in the Appalachians. A map would show these differences by associating the different types of soil with different colors on the map. When given clean, clear cut data, ArcGis can produce beautiful and accurate maps.
However, the map is only as good as the data that creates it. What is tricky is dealing with huge, messy datasets. These are “cleaned up” using a program that implements codes called R. R is a programming language that easily accesses files like XLSXs and CSVs and allows you to manipulate the rows and columns and to generate shapefiles from specific rows and columns. A shapefile is a form of data storage for vectorized points, shapes, polygons, etc. It saves maps in ArcGis in a similar fashion to XLSXs saving spreadsheets in Excel. Here is a sample code that I used to take only specified groundwater values from one column, their respective site IDs, and their coordinates and put these into a new CSV file.
Mapping
This new CSV file is ready to load into ArcGis. Once opened here, one simply needs to assign the variable which corresponds with the given latitudes and longitudes, and select a corresponding geographical coordinate axis to produce a map from this information. This code enabled me to produce the following map. From here, the map can be saved as a shapefile.
.png)
Application of Mapping
Satellites are perfect for providing accurate data across entire areas of land. However, only certain types of data are available at this scale. The factors listed above which contribute to a model for evapotranspiration are, fortunately, available. This gave us the ability to render accurately detailed maps of evapotranspiration for all of India. The code for these models provided point values for each “pixel” of India at a resolution of 1 km, and stored these values in a shapefile. The shapefiles for Bhattarai’s models produced the following ensemble mean map which is compared with ET maps of other resolutions below:
Application of Mapping
The pixelated patterns of these maps indicate that they are universally vectorized. However, the map from the previous mapping post showed a map with fewer point values. This is because trends in less easily accessible climate measurements are most accurately retrieved on the ground, on-site. This leaves far fewer measurements to accurately deduce trends in, for example, groundwater.
For the purposes of my research, I retrieved a dataset from India’s Central Water Commission (ICWC). Cleaning the groundwater data, sourced from the ICWC, included filtering out all data that was marked as not available and repeat data that gave more than one value per site. Completing the R code for this provided a CSV file that I could open right up in ArcGis. Indicating groundwater as the variable I wanted to map to show and assigning a geographical coordinate axis built me the map of groundwater values in India.
Although maps that are not universally vectorized are by default less accurate, there is a way to use them to approximate values at any location. This is called interpolation. The two most popular ways to interpolate a map of sparse data points are called inverse distance weighting (IDW) and Krigging. IDW calculates a weighted average between the points with values already assigned to give approximations for what the values at any given point on the map should be. Krigging is slightly more complicated and uses a Gaussian regression process to calculate an approximation for each value. Nonetheless, the map of the sparse data points I derived in the last post can easily be interpolated in ArcGis with an interpolation model that I built from ArcGis tools.
.png)
Application of Mapping
Loading the saved shapefile from the previous post into the IDW tool, which I had happened to call gw_pomkha_ based on the fact that is showed the change in groundwater storage in the post-monsoon Kharif season over the past ten years. The output of this model is the following map:
Application of Mapping
If you compare this map with the point plot from before (pictured below), you can see that the areas with less data points are more similar looking because they are simply averages of the few points around them.
An overall observation that can be taken away from this map is that the red areas represent the areas where the groundwater depletion is greatest. This is a result derived from a dataset sourced from an online governmental agency. The information to understand the effects of climate change is available, and I hope that some of these tools will enable you to do so.