Algorithm to identify point connection

[mercus]

The dataset contains geodesic points, even spaced, which show elevations for each point. The goal identify the number of islands in the dataset. Say for a dataset which represents a map of the united kingdom there would be 2 (major islands) + a number of smaller island as the result.

What is a good way to algorithmically acheive this?

[Oralloy]

mercus,

You will have to use an adaptive algorithm that finds disjoint sets. The algorithm is recursive...

Algorithmically, I would try the following:

Create three sets - non-sampled points, water, and islands. 
non-sampled points and water are sets of raw data.
islands is a set of sets, each entry will be an island. 
 
Find-Water
For each point in non-sampled, if elevation is zero, move to water set. 
 
Find-Island: 
Select a point from non-sampled points, this will be part of a unique island.  
Remove this point from non-sampled points 
Add this point to temporary-set of island-points 
For each adjacent location, do Extend-Island(point, island-points)
 
Extend-Island(point, island-points): 
If point is in non-sampled points Then
  Remove point from non-sampled points
  Add point to island-points
  For Each adjacent location, do Extend-Island(point, island-points)
Endif
 

Cheers,
Oralloy