[postgis-users] PostGIS case usages

[J Payne]

  Hi Regina,
    
    I'm an ecologist, and I do a lot of work with animal tracking.  The niche is small, but it's a very active area of work.  Typically, we are trying to figure out why animals go where they do.  That involves overlaying track data with polygon and raster layers; the rasters are often large (DEM's, rainfall, NDVI for vegetation, habitat types, etc.)  Time is always an important part of the analysis and a lot of the work tends to be one-off solutions to things like calculating angles between track segments or joining track positions with the closest pixel in a long time-series of 16-day composite satellite images, or interpolating the position of an instrument that only records times.  Scientific instruments produce more data every year; for example, the accelerometers we use  (which measure an animal's body movements) reported one reading every 4.8 minutes in 2013; today, they report 8 readings *per second*.  That goes for everything, even weather stations and the like.  The enormous increase in data volumes and the attendant need for speed makes database software essential.  
    
    Working with big satellite images is perhaps a need that ecologists have in common with other GIS users; it can be surprisingly difficult to navigate and to figure out simple things like how to properly attend to data quality issues in satellite images.  For example, there are often bands or layers that include quality measurements, but they have to be merged appropriately with the main image; otherwise clouds or other problems may introduce bias into your numerical summaries.  Of course MapAlgebra can be used, but the goal is to do the analysis fast, and tiling increases the complexity of the queries (BTW, I'd like more examples of how to write queries comparing tiled layers, especially in out-of-db cases; I've hit some subtle problems with that, especially where clipping is involved.  Correctly-timed use of ST_Union and ST_Dump seems usually to be part of the solution but I find it challenging to get right).  My work also involves a lot of standard distance-based queries, but those are mostly already well covered.
    
    Your first book was so rich with sophisticated and complex examples that I'm still going back to it repeatedly just to understand a bit more of each one, in an iterative way.  I'm more involved with online work than ever (Github, website-based databases, etc.) but am still feeling my way into that realm, and would welcome more examples how to do things.  As a simple example, I'd like to maintain a website with a database that is automatically updated through web scraping or an API as new data comes in to a source such as a government satellite feed.
    
    I feel a debt of gratitude to you for your generosity in sharing all of your superb work over the years; it has helped me tremendously, as I am sure it has helped so many of us.
    
    John