EXERCISE 10 – In too Deep

1. Geostatistical Analyst

We've played around a bit with interpolations before using Spatial Analyst. There's another extension, Geostatistcal Analyst, that opens up vistas in the world of interpolation that you might not have imagined existed. If you've taken Ed Green's class, you might know enough to really get into trouble with the interpolations options available. Today we're just going to do a casual exploration of them, so that you can see how different they can be. Don't worry if you don't understand all of the options available, neither do we (and neither do the majority of even advanced ArcGIS users).

Activate the Geostatistical Analyst extension and its toolbar. The toolbar doesn't seem to have much to offer, even when you drop down the menu, but that's misleading. Under Explore Data are a number of very interesting interactive data exploration tools that can tell you quite a bit about your data if you are satistically savvy. The Geospatial Wizard allows us access to a slew of interopolation methods, each of which has many different options.

Load climate and climun from the \intgeo\ClassWork\spatial directory. Under Geospatial Analyst->Explore Data, choose Histogram. This option provides us with a histogram of our data (you might want to be sure you're looking at one of the temp or the precip fields), along with summary statistics that provide us information about the distrubtion of our data. This information is very important for choosing appropriate statistical methods or pre-analysis data transformations. The other exploratory tools provide information about spatial autocorrelation, among other things.

Now open the Geostatistical Wizard. The methods you see are different interpolation methods. With climate as your input data and TMAX as your attribute, choose Inverse Distance Weighting. After clicking next, you'll see the IDW interactive interface. The map window shows what points will be used for the interpolation for any given location on the map. They are color coded by the amount of influence they have. The Sector type options allows you to have mulple sectors (each with the minimum number of neighbors you specify. Note that you can preview both the points used and the interpolated surface. Play with the options to see how they impact both.

Once you're done playing around, click Next. The next screen shows the error associated with your interpolation method and parameters, calculated by comparing the value of the interopolated surface (the predicted value) to the actual value (the measured value) at each point. Click Finish, then OK. By default the results are added to your display. Note that the output is rectangular, and one of the glitches in the Wizard seems to be that it ignores any mask you might have set., so we'll have to deal with the block. Note also that the output is not a "real" data layer, it's Geostatistical Analyst Output Layer. In order to save it as raster, you'll need to go to Properties->Data->Export to Raster. You will need to do this if you want to further manipulate the output, say in the Raster Calculator, for instance. Take a moment and explore the options for the other .

Assignment: Using either the precip or one of the temperature fields, perform 3 interpolations using different methods, one using IDW, one using global polynomial interpolation (set your power higher than 1!) and one using kriging. Using the raster calculator, subtract the GPI results from the IDW results and then the kringing results from the IDW results. This will show where and how much those results differ from the IDW results. Make a map (8.5" X 11") showing your results and how they differ from the IDW. On a separate piece of paper, do your best to describe how they differ from the IDW and hypothesize why based on what you know about the procedures (which admittedly isn't much).

Hand it in Monday, April 21st.

2. Projections

At this point in your project, you should have just about all the data you need for your project. You may have come across some datasets that just don't display correctly on top of other data for your study area. This is most likely a problem of the datasets in question not having their projection defined in a way that ArcGIS can use to make them work nice with other data in diferent projections. You'll need to define projects for all your data and reproject them into a common projection. Even if all your data does seem to be working nicely together, you should probably convert all your data to a single projection before beginning significant analysis. How do we resolve these issues?

The first thing we need to do is figure out what projection each dataset is in. This might be as simple as looking at any files that were included in the data directory or download. They might not be official metadata files, but they might mention what projection the data is in. If the projection information isn't included this way, then we need to do some sleuthing. The format of the coordinate is a big clue as to what projection your data is likely to be in. It helps if you can narrow your likely projection choices down, which we can do in most cases. For this example, we'll assume that the data set is either in geographic coordinates (lat/long), NJ State Plane or UTM Zone 18 North. If your data is for another state, you can include their state plane system and UTM zone as among the likely candidates.

The first step is to take a look at the coordinates the data layer is using. In a completely new, clean, fresh, never before had data in it data frame, add one of your suspect layers. In data view, mouse over an area covered by your data and look at the bottom right hand corner of the ArcGIS window. You'll see the coordinates at the position of the mouse cursor. Lat/Long are generally easy to recognize, especially if you have a general idea of what the lat/long for your study area should be. Lat/long coordinates do come in different formats though, and might be in degrees/minutes/seconds, decimal degrees, or degrees/decimal minutes. If the coordinates are not lat/long, they might be state plane or UTM. In NJ, these coordinates can look similar, but there's one big difference that helps distinguish between them. NJ State Plane Coordinates are comprised of either a (6-digit, 5-digit) pair of numbers of a (6-digit, 6-digit) pair (ignoring numbers the right of the decimal point). UTM Zone 18 coords (in New Jersey at least) are always a (6-digit, 7-digit) pair, again ignoring the part to the right of the decimal point.

Once you have an idea of what projection your data is in, it's time to project to match your other data. This is a two step process. You need to use the Define Projection tool (under Data Management->Projections and Transformations) to tell ArcGIS what projection the data is in. This creates a small file containing the projection information and associates it with the other files of the layer. If the data layers is in geographic coordinates, you should probably select Geographic Coordinate Systems->North America->North American Datum 1983. If NJ state plane, use Projected Coordinate Systems->State Plane->NAD 83 (Feet)->NAD 1983 StatePlane New Jersey FIPS 2900 (Feet). If UTM Zone 18 N, use Projected Coordinate Systems->UTM->Nad 1983->NAD 1983 UTM Zone 18N.

Once you've defined a projection you can then reproject it. The easiest way to do this is open up a new data frame. Find a layer of the projection you want to end up with (make sure it's defined) and add it to the new data frame before doing or adding anything else. Then add the layer you want to reproject to the data frame. Right-click on that layer's name and go to Data->Export Data. Note that one of the options is to the export the layer using data frame's coordinate system. Select this option and export your layer into the output data layer. That layer will now have the same projection as the layer you originally added to the data frame. You can also use the Feature or Raster Projections tools to perform the projection. If you do so, it's usually a good idea to Import the projection from a data layer you know to be in the projection you want. If you want to find out if a layer has a defined projection, just look at its Source tab in the layer's properties If you do this and features seem to be off by a few hundred feet, it might be case that the original data set had a datum of NAD27, not NAD83. You can use the Define Projection Tool to redefine to the appropriate projection using NAD27 and see it works.

This basic process can help identify the most likely projections you're going to come across. If it doesn't correctly identify your projection, then your remaining options are limited. You can try defining various projections for the data set and see if you find one that allows the data to line up appropriately with data in the projection you are working with..