Time Series and Multi-temporal Classification in Forest Studies

Lecture given by Dr. Premysl Stych and Dr. Josef Lastovicka

The Perks and Pitfalls of Automation in Geoinformatics

I was quite excited to hear this talk as it aligned nicely with a project I was doing in another class on the use of remote sensing in damage detection. For the project I had decided to focus on the newly developed Disaster Vegetation Damage Index (DVDI) which is supposed to help better quantify damage to vegetation cause by natural disasters or extreme weather events. As such, I had anticipated that the topic of my reflection would be related to the content of their talk (forest degredation). However, while listening to their talk I found myself reflecting not so much on the content but rather more so the methods the described as part of their research. These reflection largely fell into two categories: the benefits and drawbacks of automation in geoinformatics and an appreciation for their thoughtfulness.

It came as no surprise to me that, as Drs. Stych and Lastovicka highlighted, there has been a move towards automation, programming, and cloud computing in geoinformatics as a result of changes in our computational ability due to an increase in the availability of long-term satellite data and exponential growth in computing power and processing methods. I have seen for myself how the field has turned towards the data science realm even during the short time I have been studying the topic. Despite knowing this and seeing this transition myself, I was still shocked to see the level of automation achieved by some of the processing tools the doctors introduced during their lecture. I was amazed at how accessible the Sentinel-Hub EO browser made many remotes sensing processes. Different tools in the Hub could automatically apply indices or calculate time series or filter out cloud cover all just with the click of a button, no knowledge of how the processes work or the complicated math behind the transformations necessary. As amazingly powerful as that is, I was left with the same mixed feeling and questions I usually have when it comes to the automation of GIS and EO processes. On the one hand, platforms that make vector and raster data and analysis accessible in terms of ease of data selection and the use of tools go a long way in making geoinformatics more widely accessible and comprehensible which is a great thing since spatial analysis can be such a powerful tool. However, I think we run an increased risk of a user not truly understanding the data they are using or what the results they receive mean which can have negative consequences on the conclusions they draw from their results. I continually grapple with the question of how much knowledge is truly necessary to draw correct conclusions from an analysis. I do not have the answer but this is one topic that I am always excited to broach with experts in the field and it is a question I imagine I will grapple with for a long while considering my passion and the focus of my studies is on geocommunication and accessibility.

It is now time for me to get off my soap box and turn to the second category of my reflections on the methods used Drs. Stych and Lastovicka: and appreciation for their thoughtfulness. In the final part of their presentation of their content, the doctors broke down their results by different focus questions. I am used to focus questions being related to trying to understand a phenomenon but the questions the doctors were asking were more along the lines of trying to identify which methods, validation option, or data source was best used for their specific problem. In essence they were focused on analyzing the usefulness of their methods. This really excited me as I truly believe this is one of the most important, and often neglected, steps in the research process. Methods, data sources, models, and processes have proliferated as our computing power increases. This is can be a great thing as it introduced new ideas and perspectives that can be useful in solving a problem, but this can also lead to a lot of noise. And it can become dangerous when we don’t take a moment to reflect on how or why we have selected the data, methods, or models we did. Reflection is key to impactful research. It gives us a moment to step back from our work to really understand, evaluate, and gain perspective on what we are doing and this allows us the space and to pivot in our research if need be. With reflection comes comprehension, perspective, and flexibility which breeds impactful research and it was exciting to see that built into Drs. Stych and Lastovickas’ work. I came into this talk expecting to be focused on analyzing their content and drawing comparisons between my courses. To my surprise and delight, instead I left with many existential reflections and unanswered questions I am looking forwards to pursuing.

Twitter  facebook  Email