Advantages of opensource GIS to improve spatial environmental modelling
Formal Metadata
Title 
Advantages of opensource GIS to improve spatial environmental modelling

Title of Series  
Number of Parts 
45

Author 

License 
CC Attribution  NoDerivatives 3.0 Germany:
You are free to use, copy, distribute and transmit the work or content in unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. 
Identifiers 

Publisher 

Release Date 
2002

Language 
English

Content Metadata
Subject Area 
00:00
Integrated development environment
Software
Open source
Telecommunication
Connectivity (graph theory)
Model theory
Open set
Freeware
Bounded variation
00:43
Equaliser (mathematics)
View (database)
Source code
Model theory
Polygon
Virtual machine
Price index
Cartesian coordinate system
Variable (mathematics)
Computer programming
Number
Crosscorrelation
Integrated development environment
Lecture/Conference
Personal digital assistant
Different (Kate Ryan album)
Pattern language
Figurate number
Error message
03:19
Game controller
Multiplication sign
Execution unit
Water vapor
Parameter (computer programming)
Mereology
Number
Crosscorrelation
Causality
Different (Kate Ryan album)
Representation (politics)
Energy level
Software framework
Traffic reporting
Distribution (mathematics)
Multiplication
Graph (mathematics)
Software developer
Model theory
Mathematical analysis
Variable (mathematics)
Category of being
Uniform resource locator
Word
Computer animation
Integrated development environment
Pattern language
Right angle
Game theory
Table (information)
Window
07:06
Computer virus
Building
Execution unit
Mereology
Food energy
Computer programming
Mechanism design
Different (Kate Ryan album)
Error message
Partition (number theory)
Vulnerability (computing)
Predictability
Area
Mapping
Curvilinear coordinates
Sampling (statistics)
Variable (mathematics)
Measurement
Category of being
Autocorrelation
Arithmetic mean
Data model
Angle
Text editor
Right angle
Point (geometry)
Interpolation
Computer file
Image resolution
Virtual machine
Mass
Rule of inference
Number
Frequency
Crosscorrelation
Lecture/Conference
Average
Helmholtz decomposition
Term (mathematics)
Energy level
Representation (politics)
Model theory
Content (media)
Variance
Generic programming
System call
Integrated development environment
Formal grammar
Window
12:20
Area
Group action
Algorithm
Graph (mathematics)
Execution unit
Model theory
Line (geometry)
Computer font
Variable (mathematics)
Number
Medical imaging
Uniform resource locator
Lecture/Conference
Representation (politics)
Partition (number theory)
Resultant
Gradient descent
14:06
Group action
Building
Presentation of a group
Multiplication sign
View (database)
Execution unit
Source code
Set (mathematics)
Water vapor
Function (mathematics)
Disk readandwrite head
Mathematics
Different (Kate Ryan album)
Software framework
Endliche Modelltheorie
Area
Predictability
BayesEntscheidungstheorie
Software developer
Sampling (statistics)
Sound effect
Maxima and minima
Bit
Variable (mathematics)
Category of being
Data management
Arithmetic mean
Website
Figurate number
Reading (process)
Statistics
Functional (mathematics)
Image resolution
Patch (Unix)
Rule of inference
Number
Lecture/Conference
Term (mathematics)
Authorization
Energy level
Representation (politics)
Analytic continuation
Distribution (mathematics)
Graph (mathematics)
Validity (statistics)
Model theory
Polygon
Variance
Integrated development environment
Vertex (graph theory)
Object (grammar)
Window
19:42
Mereology
Condition number
20:05
Point (geometry)
Building
Group action
Computer file
Similarity (geometry)
Regular graph
Mereology
Computer programming
Mathematics
Meeting/Interview
Different (Kate Ryan album)
Hierarchy
Representation (politics)
Energy level
Error message
Area
Addition
Graph (mathematics)
Software developer
Moment (mathematics)
Model theory
Sampling (statistics)
Line (geometry)
Right angle
Reading (process)
Reverse engineering
24:02
State observer
Distribution (mathematics)
Presentation of a group
View (database)
Model theory
Computer simulation
Function (mathematics)
Fehlererkennung
System call
Integrated development environment
Meeting/Interview
Authorization
Representation (politics)
Error message
Resultant
00:09
electron again to the variations and today the conference and I would like to open the connection graph opensource and free software where 3 talk about issues related to grant being opened up there and the relationship environmental modeling the because I would like to invite you got data from the Department of Geography university of Toronto you would prevent advantages of open knowledge to
00:43
improve patient environmental models 1 of which with of them were here they were they let what I heard about yesterday I was making a dangerous step of providing my talk about finding minute and I read it although maybe it would have helped them the 3 of my figures don't interview compatible with the machine to accommodate important 1 of them portable will be open I'm going to be talking about a number of tools that we use in our last and research for me trying to gloss over the application that and talk about the tools and why we need them 1 of the tools of something that I I'm just the user are not involved in programming and the other 1 I think that I've inherited the source code and trying to update by day and figure out what to do next my the overall being at the
02:22
end of what we're trying to do that than to indicate environmental model GIS and and figuring out how to deal with heterogeneity I recommended that year on the left of typical polygon view world where you can divide everything happens to equal polygons and everything is uniform in the bullet in the middle thank show being you but there is there's variability within each polygon except that in that case the variability is the same in every polygon and then over here the idea that there might be more to the pattern there might be different variability in different places they was a correlation and might not be effective program so that we can kind of directionality that in calling on down here that here to remind mentioned aggregation error to small you
03:21
can feel about that men and believe that the that the the framework for what I'm going to be talking about what we're trying to do environment modeling are coming up with a way to simple bag representation of environmental problems and there's some kind of relationship between spatial pattern on the ground and the properties that are operating in the environment and we need to try and represent the spatial patterns somehow in the modeling environment in the room probably have your graph you're probably you use data and the and of then grafted great representational worldly but all the data the but for modeling usually we want to that a causal modeling unit so we've got a bunch of tools we'll will move back and forth and deal with the different levels of the visual after a while talking about I would give you a universal moratorium they should be analysis of the ultimate quickly loaded Boulevard out there that we've been using for many years and and then the American examples of water and not quite clear example of tool that you could use to you we model on a graph or Hill spatial data or not you were mine I will be talking about anyone I'm 0 1 unit that on your model representation of the world and you can make over the ball harder example sampling to all that we used it but the idea is that we want to be able to characterize the data that by sampling their locations and there again in all these examples will be that need know the level of the parallel development we talked about yesterday there always other ways to do that the men certain needs without a certain time and then our last and then may or may not be used for the future given that the other tools that could do the job and but what we're actually tool that you at by the user like well that already that operate by well with different part hierarchical being I'm worried have adapted grids so we're going to have 1 game where your sampling every the number of parameters so for random and playing with it and leave about random word got a random to create like random location and then added to the location of a variable window report what was found in either 1 or more layers I want to put that on the that's right distribution that is found within that window of control and and the variability or
06:38
if there are multiple layers involved with the correlation of ordering the between the layers and then the resulting in the table but I would do because of the 2 important are for the analysis of the spatial locations and we also have the a window that was used you can have multiple windows and different increment between window size and then in the case of
07:07
reporting the averages within each of those windows and then variant of our 3 again to protect that simple grapple with the overly simple example I just created using our government purpose I normally distributed mass and energy you with different levels of correlation and of course the window size is going up on the back and what was so of what was found at the sample but that land but of the for a real example you could characterize the gendered and exploratory data will
07:49
characterize landscape we use it for example to compare interpolation of the reminded example but the mother work lot with them that is up modeling environment in different ways changes the uncertainty in your prediction and so very quickly and reasonable of course when of putting up on modeling unit that aggregating finer level finer detail data you're doing some kind of aggregation and depending on the relationship that your modeling that may or may not introduce aggregation error in your prediction and 1 kind of curvilinear relationship in properties but when we're aggregating on an angle tendency the you will introduce you so this is bigger maybe my and some of you are modeling is taken from about the paper virus letters always but the point here is that you're going to have some kind of aggregation and modeling give so what you want to try and do not just throw up your hands in and give up to try and the 1st example of something didn't work these machines for some reason because of the pretty picture of a different kind of partitions that are commonly used in the the watershed almost all up here I've got a regular grid of another example here a map from 1 kind of knowledge of the the fight with my other example of education today mechanism and then down here is that 1 the 9 1 to be mapped out all the talking about a little more so we want we can on that on a generic Windows machine at home has nothing to the following on and of our quadtree who has a variable resolution data model so it's not arbitrary shaped like thereafter great but the plight of the gradient variable across the and let you concentrate put more or more resources in terms of resolution and the areas where you need and what we don't need after the 1st step of the program is to build up here and a very weak people representation pyramid of all possible representations of and then you can have different decompositions of being building of different possible representation using either I will number of units or the and now the residual variance that is not explained editor criteria where you stop decompose here look I was worried that the building of the pyramids of the world that will already use data that is part of everything right taking the mean of all the that's what I mean by theorem and ready to be In the numbers a lot of work them and then that he might be accommodating meaning building up all of our representation both the compositional rules and starting at the top of call them back on the 1st we look around this is the period of the the mean value and and there's also a pyramid of maybe some measure of the variance in each each of these files and goes to look at here which has smaller variance and the following content made with the war and look at all the the common variance with 1 of the most variance underneath it along with that or you got building variable grammars do you have to again bottom you reach a certain number of this is quite terminology the file them or you can use to bring about variability but there are some quick example to I thought you were the same thing that showed you put it on the example of a random thing on autocorrelated if you have a moving window and this is what we will be given in the eye the
12:22
eye image for online my mind but location but when algorithm on your quickly I and the 8 a bipolar that I wanted to come up with the partitioning that 128 I hope the 128 units to represent the area beneath the data font group and using the descent algorithm on the variability it come up with 1 possible representation both the yellow line on top of the original data that showing I would divide bacteria and for that could be used for modeling unit the graph below probably will remain the house where the the remaining variability so now you increase the number of units the were not much more the I don't really blame much variability and quickly and from the summary and it can only the other way of course that you
13:33
could do it what do you think I don't have a big number of unit but I think that you're only willing to live with a certain amount of variability and then we get the result of the whole on the really move but I don't know With only 25 per unit you are correlated with the increase in the 190 there random that really can't do much with random but work
14:07
unit so at least a little bit better than the full resolution from an even more from you 18 will become very quickly to mind laughter then the rule a little bit different here I'm going to show what we do to build up a spatial data set for a particular environment model and I read it to you that there are a whole view of the world by world and you example 1 little bit about the active and based on that the variance in the paper about the meaning of you may not realize that the hierarchical not that in in our area that is within 1 of the whole within 1 of the big and you can go that using whatever layers of that makes sense for your site and the property that you're modeling reason it is modular with respect to which the model used within the well the time to go into that but for example top 1 often or regarding the water that and that it has a bunch of functions and that impact what makes sense in terms of your building of the world it was just the whole territory that will get going than I mean so rather the world of the tool that take graph data layers and build up all building representation of the world or the real model over the again it will be I've been getting back book department opened 1st advantage is that it relies on the previous by here was actually built for a different views and we were able to get source code to it and then bring it in make work the application and and again and you just get the time to read the idea people all possible representations of the world where model make a predictions and compare the effect of different levels of resolution or different now and will detail and so on and there's another tool used to analyze that calling that all I I realized through our object today that have the graph are integrated maybe another way to do more correctly made the and better purposes and after that was developed by our model statistics showed up in the graphic unit that somebody almost identical so that the person them all of parallel development now this is really important and come up the but these are all I article like that and they cannot monitor get the thing thing and a polygon layer you can calculate it said that such as mean and variability and so on of of a continuous layer underneath polygon for each of my modeling union for each of my patches that I decided to represent the world like and say well how much aggregation aredoing underneath that how much aggregation of elevation the mean and variance validation and the head of different window different that number of patches that were chosen to represent the world of changes that distribution and then they can do the same thing on the output looking at aggregated output prediction on and the figure it is not so bad that this figure in the paper that the I'm going to have a lot more examples like 1 there are a lot of them I with the idea that handling heterogeneity in environmental modelling groups because you know what you're going to get anyway but back to try and get me main how much of a problem for you and then you can try to minimize all I would like you open for managers exchange of ideas and conference where and of that all of the minimum of the fact that you can see how other people all the problems in their 1st that for yourself and we're able to it forgot remember example vertical read that properly credit the author the bad news is that in our in our sample and I'm and experimenting with the fact that we were behind these modeling units and we want them look with new ways of representing data we have the framework in which we can do these experiments change how the world is represented and not have to be constrained by the Bayesian models that are not in the murder of yeah I could work there thank you and thank you for this interesting presentation other questions underneath summary against which all
19:35
the names of the model that you didn't read to you from the you will need more money according to the
19:44
modern conditions that were not included in that part of the the of the year of it and on word because of did it and you're
20:23
on the data and the various levels of development you would want to work on graph right on the on the radio build great graph data running on our out of working in and and that is that they were going to happen with changes in that at the here and and maybe we need we evaluated in America and Europe as well as in the ground and like all I think you're what we call the whole thing and that were you grab right well but it varies from 1 part of the fact at different is that single across multiple layers of just 1 layer the similarity both of them are there more going to interact with the point and that's the 1 that gave may the best conventional are and it is easier to use he applied on are not quite the reverse of what the representations are red and the moment it is if that program that we built you using pyramid about competition at all through reaches will and I have here a bigger or whether it be that will go on and on the right you have all the data files that are not right by are not group and all that and we will reason on and and grasp the world finite is very good the running the read model but that will form in our graph data layer and coming up with the huge and of the the and what it is that the different error hierarchical representation world because of that the Broncos I have a question we got in Europe and in the line specially on a regular sampling design can you In addition to light based on if I understood you
22:44
correctly you told In this and being design in the sense that you repeating gap at the end because of the that is correct that I mean that approach needed to be and take that the dependence between the selected because of the following doesn't have to be in the early years of so I would like to ask you about the model is more important for you can never but there is no like every and because of that and that of the between them so that people who the VM occurred in the area that of kind of the primary way wanted it where it's not like they were going on during sampling on the ground all and ecology or whatever it's not the same kind of regular at end up this is what you wanted for that particular thing so again with the research so run together with the purpose and I I can more true regular sand and with all of the various I would like to ask you if you have
24:14
you been call representation of data in from the environment model because the presentation will problem of sampling and really kinda error control view of the of the of the in my own work and mean that at the beginning of the from all 3 of the authors own the uncertainty act and looking at for the example I put out there is just going up and showing how the results change the from the output from the model is just a descriptive qualitative and making it from people do you observations that we do have a lot something I've done similarly integrated application and with another model was using our MonteCarlo simulations to build up a distribution of possible outputs and so I can build up an error predict rich the uncertainty doing multiple runs but this while the debate and so long was really beautiful the big 1 2 days just to do 1 things but there aren't any further questions think thank you again