Computers, Society, and Nature

This article gives us a detailed introduction about Geovisualization. The author started by giving out reasons about whys should we care about Geovisualization. In short, I think the explanation would be that people can get knowledge from it by transforming the geospatial referenced data into information and turn the information into knowledge by analysing it. One example would be the switching from data to paper maps, and then paper maps to web-based maps. The visualization is advancing overtime, so researcher can get more out of a geospatial dataset. How much can we get from the dataset are largely depends on the visualizing techniques.

Then, the author introduced some issues that still remains in the geovisualization field. These problems are representation, visualization-computation integration, interface design, and cognition-usability. One issue I noticed about representation is that in order to take full advantage of the information we want to give out in the geospatial dataset, we want to personalized representation as much as possible. One example I can think of is about Google Map. When I’m trying to find certain stores in a big shopping mall, I always find it hard to locate a certain floor and the direction since sometimes there is no detailed information about this represented on the map. However, I find some other map application gives about 3D navigations in shopping mall so user will find it very easy to locate a certain store. Obviously, the latter application gets more out of the mall database, and make the navigation process more personalized.

This article essentially introduced the emergency of GeoAI. The author gives out some detailed reasons about why we should use GeoAI, and he also briefly reviews the expert systems approaches, the use of heuristic search procedures, and the utility of neurocomputing based tools. At last, he predicted the future trends of GeoAI as a emerging new technology.

GeoAI is actually a new topic to me since I have never done a project using this technique. I think it would be very useful and convenient when we are facing a huge dataset and trying to analyse or model it. As far as I understand it, people basically just transfer their thoughts to the computer and let the computer to decide and calculate result. Then, there would be a point where GeoAI will be connected to spatial data uncertainty. Is it possible to train the data and let it decide the level of uncertainty in a dataset? Or it there any way to eliminate of reduce some uncertainty in a dataset?

Another aspect to think about the uncertainty problem in GeoAI would be the supervision of human. What I get from the article is that people can supervise the computer when they are doing analytic works using the algorithms researchers put in. Would this supervision process bring more uncertainty into the dataset, or it will help to reduce the error? These are thoughts that come to me when I’m reading the article.

In this paper, Maceachren and Kraak (2001) concluded the research challenges in geovisualization. The first thing that catches my eye is the cartograph cube, which defines visualization in terms of map use. The authors argue that visualization is not the same thing as cartography. Visualization, same as communication is not just about making maps, but is also using them.

While the authors highlight the importance of scale issue, integrating heterogeneous data also present a challenge for geovisualization, because of the different categorization schemes and complex semantics that are applied in data creation. Similar conditions or entities are often represented with different attributes or measured with varying measurement systems. Therefore, the heterogeneity raises questions when we use data from different data producers: How to assess heterogeneity? How to make decisions about whether data may be combined? How to integrate multiple data sets if the same semantics are used differently?

Further, the emergence of geospatial big data, such as millions of conversations via location-enabled social media, stretches the limits of what and how we map. The potential of using geospatial big data as a data source for geovisualization requires developing appropriate methodologies. While this paper mainly discusses geovisualization of quantitative data. I am also curious about how to visualize qualitative spatial data. (QZ)

This piece gives a very thorough background on geovisualization and its problems, especially its problems across disciplines.

A part of the piece that caught my attention was when MacEachren and Kraak said that “Cartographers cannot address the problem alone.” Through all the papers we have read in this class, there is a trending theme that there needs to be more cross-disciplinary communication in GIS to solve crosscutting problems. This article is better than other articles who just mention that more communication needs to happen; this article actually lists ways to better research cross-disciplinarily, in addition to listing short, medium, and long term goals. 

Although, I also feel that this article was written in a way that was very very generalized and vague, which made it a bit difficult to follow. This also gave their reasons less clout because it’s always easier to explain vague solutions as opposed to more specific ones. Some specific GIS examples would have also been very helpful!

In this old paper, Openshaw (1992) calls attention to the potential of artificial intelligence (AI) methods in relation to spatial modeling and analysis in GIS. He argues that GIS with a low level of intelligence has only little changes to provide efficient solutions to spatial decision-making problems. The application of AI principles and techniques may provide opportunities to meet the challenges encountered in developing intelligent GIS. One thing which draws my attention is that the author mentions it is important to “discover how best to model and analyse what are essentially non-ideal data”. But I didn’t see a definition or explanation of non-ideal data in this paper. Does the non-ideal data refer to less structured data or unreliable data? AI can use less structured data such as raster data, video, voice, and text to generate insights and predictions. However, every AI system needs reliable and diverse data to learn from. Very similar data can lead to overfitting the model, with no new insights.

Further, Openshaw demonstrates the usefulness of artificial neural networks (ANNs) in modeling spatial interaction and classifying spatial data. But he didn’t mention how to transfer data from the GIS to the ANN and back. The most widely used ANNs requires data in raster form. However, the spatial data used to produce an interpretive result in GIS is most efficiently managed in vector form. Therefore, I am wondering if there is an efficient methodology to transfer information between the GIS and the ANN.

As of now, GIScience is not new to AI. For example, the most well-known application of AI is probably image classification, as implemented in many commercial and open tools. Many classification algorithms have been introduced to clustering and neural networks. Also, recent increases in computing power have made AI systems efficiently deal with large amounts of input data. I am looking forward to learning more about the current uses of AI in GIS.

There’s a very important component of geovisualization missing from this article: aesthetics. Koua et al cover a great number of factors important to geovisualization, in particular test measures, effectiveness, and usability. However, they only briefly mention users’ “subjective views” towards a geovisualization and “compatability (between the way the tool looks… compared with the user’s conventions and expectations).” This omission is noticeable, since geovisualization is, as its name implies, a very visual aspect of any cartographic scheme, and the aesthetics of any visualization are almost always inherently important. However, it may not be entirely surprising, since this paper focuses on the “usability and usefulness of the visual-computational analysis environment.” The authors have implied through this omission that aesthetics do not relate to the usability and usefulness of geovisualization; however, I would disagree with that assumption. Maps are, at their core, a visual way of displaying data; how they look, not just how they show data, matter. Therefore, however subjective aesthetics are (and they are quite subjective), they must relate to the effectiveness of a map. A map could have great data to show, but if the colors are oversaturated, or the water isn’t blue, this could distract the eye of whoever is looking at it and take away from the map’s findings. If the map user can’t pull the important information from it, then what’s the point of having a map at all? I understand why the authors may have decided to omit aesthetics, considering it’s such a subjective factors compared to everything else they discuss; however, including aesthetics would have made this discussion on visualization usability more robust and complete.

This article is extremely relevant to the independent study I’m conducting this year, and I think I’ll be able to use it for some of the methods I’m conducting. My study is looking at different 911 calls in Baltimore over the last 7 years (assaults, overdoses, car accidents, person found not breathing, and shootings). I have both the address where the call was placed as well as the time, down to the minute. This could be considered a “health” study of sorts, since bodily harm is a health outcome, and I’d like to find factors that correlate to the calls’ times and locations. Therefore, the geoAI described in this article would be great for my study. It tackles big data, which I have (over 6.5 million 911 call times and locations), geoAI that produces high-resolution exposure modeling, which is what I’m looking for. The example given about the study that developed a method to predict air pollution was particularly appealing to me. I’d like to do something similar with my own data, inputting as much spatial data available (demographic indicators, built environment factors, etc) to see which variables predict calls in space and time. I’m glad that this article discusses “garbage in, garbage out computer science,” as that was an issue I was concerned about while reading this piece: that treating geoAI like a black box or ignoring data quality because advanced methods are applied to them may result in flawed results. These are factors I’ll have to keep in mind in my own study, and I’ll have to research further than this article on the proper data, methods and contexts to conduct such an exposure analysis.

Openshaw’s paper, written in the early 1990s, gives us an interesting glimpse into the early days of GIS. It was very interesting to hear about the problems of then as compared to the problems of today; for instance, he advocates that computers should be utilized more in analyzing GIS data, which is in comparison to today, where computers are ubiquitous in GIS analysis. This paper focuses on greater AI involvement because he believes the combination of non ideal data that GIS provides and the increasing complexities in GIS research make some analysis too complex for people to understand.

Openshaw believes that we must shift our mindsets from prioritizing the conservation of computer energy to the utilization of computer’s “endless” energy. He expands on this mindset by listing some examples of computer modeling, such as genetic optimization, and neurocomputing, that he believes GIS should start to utilize more (and are in popular use today).

A part of this piece which leaped out at me was about how people’s sentiments towards computers have changed over time. At one point, a computer’s use was something that must be conserved, and now computers are used everywhere for everything. Everyone assumes that you are connected to the internet all the time; however, this constant connection is also creating new problems concerning digital privacy. I wonder how this constant connection affects GIS data, is it still “non-ideal”, according to Openshaw? This is hard to answer as he never exactly explains how GIS data is “non-ideal” in the paper.

The author introduces the network data structure, the theoretical basis of Network analysis in graph theory and topology. It also discusses that networks are an alternative way of representing the location and the difficulties in the network location problems.
I am particularly interested in the discussion of the 3 types of data models in the evolution of network analysis. The author proposes that it is important to preserve the topological properties of the data whereas these properties also impose difficult constraints on network analysts. The author gave the scenario in which a vertex must exist in the crossing whether or not a true intersection exists, which is problematic when modeling bridges or
tunnels. This reminds me of the project that I’m working on. I plan to use a supervised machining algorithm to extract the road network from satellite images and turn it into a road network. However, the preservation of the topological properties such as the identification of if a crossing is a true intersection or a bridge is extremely difficult. The author then discusses the pure network data structure that is currently widely used in GIS. I think this is a useful topic for me to look into, although I am still not so clear about what is the difference between these two structures and what is planarity requirements that the author mentioned in discussing the two data models.

This article gives an outline of data formats, data representation, data sources, data mining approaches, related tools, and issue and challenges. The author concluded that “spatial data mining is the analysis of geometric of statistical characteristics and relationships of spatial data”, and it can be used in may fields with different applications.

In terms of spatial data mining tasks, I think other than spatial classification, spatial association rule, spatial clustering, and trend detection, terms like local statistics, spatial autocorrelation, point pattern analysis, etc. should be mentioned or included.

The author also mentioned that the issue and challenges of spatial data mining is data integration and mining huge volume of data. But the author did not explain why this two are challenges. This would be the first things that I felt confused after reading this article.

Then the author provided an architecture of the solution to this two issues and challenges. With no further explanation in detail, I find it really hard to understand why this architecture can solve the above challenges.

Mennis & Guo put together a brief, somewhat easy to read literature review on spatial data mining, a good starting place for those with only nominal knowledge of the topic (like me). They go over spatial data mining methods and processes as an intro and follow it up with recent (as of 2009) developments in the field and connect the methods to the reviewed articles. They end with the growing area of spatial data mining  and its expanding “frontier”.

With the exception of the first bit of Section 2.2, I found the paper pretty easy to follow. I was never good at statistics but I remember enough to get through it, and the examples included in each short paragraph were helpful. The most intuitive parts to me were the mentions of remote sensing, which is not something that I would have immediately associated with spatial data mining or Big Data, but once the connections were there it made more sense and helped clarify some other things mentioned.

In the conclusion, Mennis & Guo say “Data mining is data-driven but also, more importantly, human-centered”. This is basically the foundation of Critical GIS, that GIS isn’t a set of tools to be used in a vacuum but to incorporate human knowledge/social theory, and that products of GIS and their interpretation are always influenced by some personal epistemology or ontology. It isn’t surprising that they made this conclusion, but it is an interesting inclusion in an otherwise very technical and by-the-book type of article.