Niche Modeling

The Portal of the new WhyWhere 2.0 server available here. The algorithm has been rearranged to provide a streaming experience that works better for data mining large numbers of variables, interacting with web browsers, and embedding in other applications. This is the first predictive algorithm to be implemented in this way.

Some Resources

• Method: See David R.B. Stockwell, Improving ecological niche models by data mining large environmental datasets for surrogate models, Ecological Modelling 192 (2006) 188–196.
• Rationale: Advantages of WhyWhere
• Data: General description of datasets to date
• Data: All_Terrestrial.list
• Data: Marine.list
• Data: Metadata files
• Example: Predicting Global Warming
• Example: Predicting House Prces
• Example: Day, T and DRB Stockwell, (in press) A dynamic surrogate model of global landslide distribution. Earth Surface Processes
• Old: Tutorial introduction to WhyWhere

To use, read usage instructions below.

Algorithm

Instead of the stages - upload data, gis, predict, and explain, the last three occur simultaneously as the algorithm mines continuously through the chosen database and the information is streamed down to the browser. The best prediction to date is displayed as well as the response curve, and may update a number of times before completion.

Each variable is treated as an image. A simple surrogate model is developed by classifying the evironmental values into discrete classes using an image processing heuristic, determining the probability of presence in each class from the proportions of classes, and then altering the palette of the image to produce a graduation from red (high habitat suitability) to blue (low habitat suitability).

For details read method description at link above.

Issues

The multipart updates that provide the streaming are not always supported by browsers. If you have a problem at first, try another browser such as Mozilla, FireFox or Opera.

Usage

To test, paste the points below into the box provided in the WhyWhere serve test form and click submit. You should be rewarded with an updating map of the distribution predicted from the points. Experiment with different options.

-99.1166666666667 21.295
-104.866666666667 20.7666666666667
-96.45 15.9666666666667
-96.4666666666667 16.0666666666667
-103.9 20.7333333333333
-96.9833333333333 16.5166666666667
-105.98 23.5583333333333
-99.235 18.925
-96.575 17.2783333333333
-103.608333333333 19.5616666666667
-104.4 19.6966666666667
-104.323333333333 19.8
-92.65 15.0333333333333
-92.6583333333333 15.3183333333333
-99.2033333333333 23.0983333333333
-96.9283333333333 19.545
-99.3 19.045
-99.3083333333333 19.0483333333333
-99.1916666666667 23.0583333333333
-99.15 23.0333333333333
-99.7166666666667 17.5383333333333
-100.751666666667 19.69
-96.95 19.45
-98.6416666666667 19.2116666666667
-97.8333333333333 22.25
-99.1416666666667 19.3083333333333
-103.631666666667 19.6166666666667

For serious usage you must first register using the link at the lower right. A message in the top left will tell you if you have been successful and give you a user ID number. Record this number. Use it in the User ID field of the WhyWhere form. Good luck!

ToDo:

A great many enhancements. The following are high priorities.

1. Extend to 2,3 dimensions

2. Hardening

3. Replacing slow parts with faster heuristics

4. Shell out computation (parallelization)

5. Arc info output

6. Other methods