North River Geographic Systems Inc

Spatial Problem Solving

postgis

20 minutes at the East TN TNGIC Meeting

· · Leave a Comment

I went through a year or so of not talking at conferences. For local ones I’d show up and do something. I taught a workshop at FOSS4GNA 2018. I did a workshop at FOSS4G 2017 in Boston (and it sorta sucked).  I need to get out of that mode and talk more – but I’m happier being quiet these days and working with clients.

Anyway – at FOSS4GNA 2018 I discovered something – the OSGEO Suite had disappeared from Boundless’s Website and was replaced with a github repository. It wasn’t long after Don Meltz put up an excellent explanation of the OSGEO Suite. The OSGEO Suite being gone isn’t a problem. It does increase the amount of talking I have to do spend explaining it’s not a problem. I could even compile the suite – but my clients can’t. They are smart people – but the time to compile the boundless suite won’t be happening.

One of the things I keep running into is the idea of an Open Source Server. The OSGEO Suite was well known and I get people asking “Hey – can you set us up a geo server”. They mean the opengeo suite. I then launch into an explanation of what makes up a “geo server” and they go a bit blank and answer me “Yeah – one of those”.

So for the next bit I’m gonna be doing something like this:

I want to explain how this works to people. We’ve gotten a bit deaf by hearing “ArcServer” at every conference in TN as this one monolithic thing you have to install to make a web map…which really don’t need since AGOL appeared but…you want SDE and it’s not called that and you have two people editing but you need SQl Server but you don’t….

Anyway,  QGIS/PostgreSQL/PostGIS I’ve installed multiple times and as of late I’m experimenting with geoserver and what it can/can’t do. I have no grand delusions of creating ‘Randy’s Open Geo Spatial Super Duper Server’.  I want to get to the point where maybe mapserver is an option in this. So much to learn…so little time.

I suck at github – but I’ve been more active as I try to develop a skill set and get over the angst of pushing and pulling things. I’ve built a vagrant box to install PostGIS/PostgreSQL/Geoserver so I can have something to point at plus it gives me a starting point for an upcoming class I’m in the middle of building. The vagrant box isn’t pretty – I expect to change it over the next bit as it’s mostly pieced together from a lot of other people’s work (Coleman McCormick, Dave Smith, etc). It works though and I have everything but pgadmin 4 running in one spot. Which – I don’t think this is anywhere near “install this for a server” because this is mostly just for talking, training, and testing.

So why vagrant? Why not docker? I can gloss over vagrant faster at theses talks than I can docker.  I have 16 minutes to draw diagrams and about 4 minutes to publish one piece of data: QGIS -> PostGIS -> Geoserver -> Leaflet.

Anyway – if you’re up for the possibility of a train wreck of a presentation come by the East Tennessee Meeting in Kingsport next week as I try to explain how this works and hopefully you get excited enough to install some or all of these components and do something cool with your data.

Adding Triggers to Geopackage

· ·

In between work I get a chance to play around with software. I’ve been digging around postgis more and more these days trying to get a better understanding of what it will do or won’t do. It’s a never ending learning experience for me. Sometimes it pays off – the never ending tutorial left me suggesting we dump a database to a geopackage.

I’m great at using geopackage in sentences like “USE GEOPACKAGE” and “Hey – what do you think about geopackage?”. Since I’m actually officially using it I decided to do some digging. It’s a SQLite container. Since it’s a SQLite container I should be able to do some things with it – like…..

spatialite gui

I can use the spatialite gui and throw SQL against it and change the tables.

While talking to another person the question started with ‘Well – why can’t we just do the data collection in geopackage?’.  Which we did some back and forth and in some cases this would be acceptable vs doing a full postgis install.

So my first question is: Can I make a trigger in a geopackage? The answer is yes. For those of you who don’t know what a trigger does – it responds to some action in a database. In postgresql, I put multiple triggers in a database for a client. Some triggers calculate lengths and some areas and some make a note of who edited a feature.

After some digging I found the syntax and wrote a trigger for SQLite/Geopackage in the form of:

CREATE TRIGGER update_esn AFTER INSERT
ON addresses
BEGIN
update addresspoints set “esn” = (select esn from polygon where st_within(new.geom, geom));
END;

I used the spatialte gui to add the trigger. So if I pass this off to a client they can edit an address and have it automatically pull in the emergency service number. Magic? Eh – close enough for me.

QGIS Screenshot

While it’s not going to work for a multi-user situation (or will it – I’ve not investigated) – it works. So now I’m adding widgets to QGIS to aid in data collection.

Rolling out PostGIS/QGIS Part II: Load your data

· ·

I did it – I’m at Part II. I actually did it. Wooooooo. For some of you – this isn’t news – but – for some it will be.

For the Record – Part 1.

So you’re sitting with a client and you’ve documented and asked “WHY” so much they really want you to leave – it’s time to start loading their file based geodatabase into an actual database. Here’s where it gets weird for about 5 minutes. In those 5 minutes where you’re getting ready to load data you really need to look at your client (hopefully sitting with you) and go “OK – you’re no longer calling yourself a technician – you’re a database person now”.  They flip. You attempt to calm them – but yes – database person. Which isn’t particularly appetizing but this pile of data you’ve been hoarding in flat files for a good bit is a geographic database.

Most of the database admins I know are actually weirder than I am – which says a lot. One of my most spectacular “butt chewings” as a federal employee was the one time I did something to an oracle database and broke it. The admin had a fit. So much so when the yelling stopped several people told me “it’s not you – he’s on medication for this type of thing – you know – oracle DBAs and all – it happens so don’t take it personally”.

If you search the internet using Google for “Load Data into PostGIS” here is what happens.

google results not using qgis database manager

To spare you some reading – the first return mentions shp2pgsql. Actually most of the returns on the search will mention this. So much so you might think “I’m going to dump this file based geodatabase out to a shapefile and load it”. Don’t. For the sake of not over explaining this part – open QGIS and  use your browser to navigate to your file based geodatabase and QGIS will read it. Same with shapefiles. Same with about anything. If you’re leaving the land of file based geodatabases, as my client was, here we sit with a way out. Add that data to your map display in QGIS.

qgis screen

If you open the Database Manager in QGIS you can do quite a bit with the data. If you have PostgreSQL/PostGIS running (as we do) you can connect to it and import data. You don’t have to make a jump to shapefile then load your data. You can do it here. Reproject, define primary keys, and apply spatial indexes all from a user interface. With one click your data gets pushed into postgresql/postgis. For a lot of my clients and potential clients this is unknown magic. Your data will change just a hair. Lower case your columns for the love of all that’s good in the world. Apply spatial indexes. Document if something significant changes.

So go back to my second paragraph where I went “you are now a database person”. You’re taking all these flat files or flat tile structured things and making them interact with each other via  Spatial SQL. Which – not that big of a deal if you’ve got a full enterprise commercial roll out of Something Something Server Enterprise. It is a huge deal if you’ve never been able to do that. You get a large set of tools at your disposal you didn’t have before – plus the tools with QGIS you’ve got something very flexible in the making. Go ahead and order those books you’ve been eyeing. Join a listserve or three and ask questions. I still suck at SQL – but less than I did a few years ago.

There are still a pile of things to worry over like ESRI’s subtypes and domains and projections. Maybe you’re projections aren’t right. Maybe there is raster loaded in this File Based Geodatabase.   What about permissions? Editing? You’ve just upgraded your Geographic Information System from a bunch of files to a database. With great power comes great responsibility – so you need to jump back into learning mode.

………and where does that lead me – to Part III and “Hey why did my attributes turn into numbers you said this was just pushing a button?”

 

 

Rolling out QGIS/PostGIS Part 1: Document

· ·

I hated putting Part 1 up there. I start a lot of Part 1 posts and never get to Part 2…..I’m planning on this being a multipart post. After a while of screaming ‘hey why don’t we put QGIS everywhere’ I had a client decide to replace ArcGIS as their primary GIS. We were working along with QGIS/PostGIS/Fulcrum/ArcGIS and it finally got to a point to just switch. I’ve done a lot of QGIS installs and a lot of ArcGIS installs and enough PostGIS/PostgreSQL installs It doesn’t worry me – what does worry me? Data.

Way back in 1995 I was working for the Federal Government doing commercial mapping and the one thing I took for granted at the end was “the data dictionary”. We would work a project and at the end one of the guys would sit down and crack open a spreadsheet program and document every layer we had made and why we had made it. I guess sometime in 1996 or close to that we had FGDC metadata roll out and to be honest I don’t remember if we continued on with the Data Dictionary or dove into the FGDC Standard.

The client had a file based Geodatabase they had been working out of for a good while. It was functional. It had domains and subtypes and all the things you would hope a properly built ESRI file based database would have. The one thing it didn’t have were “reasons”. As in “why is this field here and not here and why is this here?” There were a lot of domains. Not so many subtypes though.  Apparently there had been somewhat regular revolving door of consultants. Everyone added their spin on things and here I sat….adding my spin…or should I say changing it all. So that started a whole list of questions which ends with me writing everything down in Google Docs. Yes – 22 something years later I built a data dictionary to describe all the attributes and columns.

Domains
Domains in ArcGIS

If I look back at my time for data conversion and banging my head on the wall – 60% was asking “Why?”. Then documenting the responses. In some cases we just had to accept “that’s just the way it is” and in others I was able to trim some bloat out of the database. I still think there needs to be a second round of trimming on this database.

So I’m going to spend some time talking about this move. Then talk about QGIS 3.2 because that came out – and there are things in that release that make me want to upgrade them off the LTR right now. For those of you not familiar with QGIS once a year you get a Long Term Release. It’s awesome because “things don’t change” – it’s stable. So we have it in enough places it’s going to sit until October at least.

Part 1 for all of it’s rambling ends…or starts right here: Document the data. Sit with the GIS person and ask why to the point you’ll get on his or her nerves. Your default answer to everything should be “delete”. Except you don’t. Use  ArcGIS because there are going to be a few things you aren’t going to find talking to anyone.  Once you’ve written everything down you can think of then you can start. Take the goofiest layer you have and move it into PostGIS. Check it. Double check it. Check it again. Check Field Lengths. Check Date Fields. Check to make sure someone hasn’t “shapefiled” your data and you’re left with “descriptio” vs descriptions.

Example – this database had two fields that completely baffled me: Location and Place. Location was for “where” the data resided as in an address. Place was a placeholder for notes. I had no clue. Luckily the client was patient as I asked more and more questions. Note the screen shot.

QGIS Screenshot

The other reason I’m writing this all up here – I think this is about to be a service: Migrate over to a free and open source server that is commercially supported for clients that can’t afford all the licensing that comes with an ESRI option. Of course I don’t want to call it an ‘open geo suite’ since that name has been used. I don’t plan on making downloads. My one big heartburn was I didn’t load geoserver for them – I couldn’t justify the overhead and banging my head on Windows Server more plus they had a bare bones server. Being able to edit the data was more important than pushing services out. Maybe later – but for now – They are editing and making maps and making decisions based off where we are.

 

 

Last Year’s Hurricane Season in the USVI

· ·

So I’ll talk about this finally. I usually don’t talk about “work” unless it’s in generalities about software and processes. I probably should talk about work more.

I’ve never been in a hurricane. I think so far I’ve been in about a handful of earthquakes, stood on one volcano, and I think I’ve seen one tornado. Not that I’m looking to be in a natural disaster – I just don’t have a frame of reference for what happens during a hurricane or a tornado.

If you flash back to early September 2017 the Caribbean was gearing up for a hurricane and. Turns out two hit close together. I had the chance to work in the US Virgin Islands on an addressing project with AppGeo and Spatial Focus several years ago. It’s been a “wait and see” type of deal. We did a little and watch as red tape and things slow it down. I called a few friends down there before the hurricanes hit and wished them well. Overall – there wasn’t a ton of fear but a lot of concern. I was concerned because the 911 system isn’t great as the addressing isn’t wonderful in the USVI.

My memory is getting fuzzy – but it seems like both hurricanes hit and it was a while before anyone could call/email me. Maybe it was late September or October before the phone rang. They needed help to figure out where the tarps needed to go.  FEMA/USACE was headed down to bring tarps and supplies – maybe they were already there. A sneaker net of GIS data had cranked itself up on the island as thumb drives were passed around. I started trying to figure out “what made sense with the base data I had available”. I started digging into what USACE and What FEMA would do if your roof was damaged. More or less if you had 50% of the structure of the roof left they could put a tarp on it. I came up with a plan and contacted AppGeo and they tossed a person my way.

Structures were categorized:

  • 1 – 25% Roof Damage
  • 26 to 50% Roof Damage
  • 51 to 75% Roof Damage
  • 76 to 100% Roof Damage
  • Structure Destroyed
  • No Damage
  • No Imagery

NOAA flew imagery of the area. The imagery was a bit rough around the edges but it worked. Actually – let me back up – for what it was it was awesome. Granted it didn’t lend itself to any sort of automatic classification as there was a lot of building lean and some missing imagery. It worked though. That was a great free resource by NOAA.

Hey that’s a lot of points. Yes – 48380 points covering three islands. Granted I didn’t get every house as I had nothing much to start out with except OSM and some USDA NAIP imagery. I would guess I’m within 10% of an accurate building count. Some people have offices/live in shipping containers down there. WHICH – when I was there made me go “awesome” but when looking at the imagery at a shipping container blowing a half mile across a field – not so awesome.

It did give me flash backs to my old remote sensing days with the government. Although this was less refined – it was a game in the afternoons to see how many points I could collect in an hour. Some neighborhoods were easy. Some not so much as there wasn’t much left. The other nice thing that happened was sometime around December I switched over to QGIS 3.0 and started seeing what I could/couldn’t do with this almost released software. Several bug reports were filed so more good came out of this.

The data started making it’s way out to one agency after another. My big fear then was “what if someone uses this beyond ‘roofs'”. I wasn’t doing a full blow damage assessment because I was seeing some evidence that maybe the roof survived but the bottom floors didn’t. Plus I wasn’t on location.

It took me a bit longer to finish than I wanted – Volunteer projects tend to be the last thing I work on most days.

I did start pulling numbers out of some of the work – this is for St John Island:

Count Classification Percent
1487 1 – 25% Roof Damage 41.42
326 26 to 50% Roof Damage 9.08
131 51 to 75% Roof Damage 3.65
126 76 to 100% Roof Damage 3.51
1007 No Damage 28.05
17 No Imagery 0.47
496 Structure Destroyed 13.82
3590 Total 100

So 25% of the houses had no roof damage. 60% had something happen. More or less 15% were leveled. I’ve heard ancillary stories where almost all houses had water coming in them and in a warm tropical climate you get mold. Mold would make the house unusable for me.

Anyway – I worry about the 2018 Hurricane season that is upon us. Maybe it will be quiet – maybe it won’t be. AppGeo has Mapgeo for the islands. It worked out well. I’m going to try to “build” a database here so in case something happens and I’m not starting from scratch in case I get to help out again. I hope I don’t help as nothing happens.

 

Contact