Last week I attended the Upper Midwest GEOCON, which brought together GIS professionals from Minnesota, Iowa, and Wisconsin.
REST Best Practices
REST Services can get out of control quickly, and cause all sorts of performance issues. Mark McCart at Iowa DOT gave a talk on creating an inventory (in just a simple spreadsheet) of all the REST services they have, as well as collecting information on what maps/apps consume those services. They also came up with a schedule of updates and a change plan, who is in charge of the service, simplified symbology, table structure, and metadata.
One cool thing to learn is that you can get Esri to come out to perform a ArcGIS Server Health Check. They send an Esri staff person to be with you two days on site to evaluate your server architecture and infrastructure. Mark noted that this health check was the absolute most useful step they took to get started in creating better REST Service practices at Iowa DOT.
Paul Wickman at RESPEC lead a workshop on CartoDB, which is a web mapping platform that is a competitor to ArcGIS Online. It’s very analytics/visualization focused. If I want to make a really nice, pretty web map (especially one with a time scale), I wouldn’t hesitate to use CartoDB.
Paul Braun at Continental Maps (who was one of our past guest GIS bloggers) gave a fascinating talk on UAV elevation accuracy. His company is well known for providing quality LiDAR data. So, he did a test where he compared his extremely-high-confidence-in-accuracy LiDAR surface, to what the elevation results were from collecting on a UAV. And it was off by feet of difference in elevation, inconsistently too. After digging in, some of the reasons this might have happened is that the UAV elevation processing algorithm isn’t quite there yet and has a hard time with certain kinds of collection environments. Perhaps the software technology has a way to go before being mature enough to trust without extensive QA/QC.
Remote sensing is essentially the idea is that you can teach software visually what something looks on an aerial image so that it can find all the locations where the thing you are looking for occurs. For example, you could teach software to know what forest land looks like, and it will scan through the entire aerial to find all the locations on the aerial that match “forest land” criteria. Then, you can update your criteria as needed, and rerun until you get results that seem right.
What I learned from Cynthia Berlin at UW Lacrosse in her talk is that there is a sweet spot when it comes to utilizing this kind of software: at about 2-3 counties big. Otherwise, you are probably better off doing delineations by hand, because algorithms aren’t 100% and there’s always some clean up. The software used in the demonstration was Visual Learning System’s Feature Analyst (because it was easier to learn, and cheaper), but Cynthia plans on using E-Cognition in the future. It’s probably best used as a “first pass” for large areas.
All in all, a good conference. It was great to connect with professional across the Midwest to discuss GIS!