Remote Sensing and GIS for Economists
Introduction
In 2021, when I was applying for PhD programmes, I highlighted my remote sensing training in my personal statement. I argued that it would allow me to conduct empirical research in regions where conventional data sources are scarce. While economics papers using satellite imagery had already existed for quite some time back then, they were few and far between. As a result, I felt confident that this skill would give me an edge in the application process.
Five years later, as I write this book, I am no longer so sure. The use of remotely sensed data in economics has come a long way since 2021. Not only are there far more papers using remote sensing as a primary data source, but many economists, particularly in climate economics, now regularly produce new data products as part of their research. Yet, remote sensing as a data source remains heavily underexplored in economics. Furthermore, even when these datasets are utilised, researchers often treat them as just another data source, rarely concerning themselves with how the data were actually constructed.
I am, therefore, writing this book to provide economists and other social scientists with a systematic introduction to remote sensing and GIS. As this is a book for social scientists rather than remote sensing specialists, it will not dwell on the nitty-gritty technical details of the remote sensing process. Instead, it will focus on the tools, techniques, and conventions used when working with satellite images and derived data. However, in Chapter 2, I provide a high-level yet thorough discussion of the physics and engineering behind remote sensing. You may start with Chapter 3 if you wish to dive headlong into tools and techniques, but if time permits, I urge you to spend time on Chapter 2 so that:
- You understand the physical constraints of the data, such as why certain atmospheric conditions or sensor angles might bias your results.
- You can critically evaluate the quality of derived data products and understand its limitations before incorporating them into your analysis.
- You understand the entire data construction pipeline and can begin to think about constructing data products that fit your research needs rather than solely relying on datasets produced by others.
Similarly, Chapter 3 introduces spatial data and the framework for understanding and manipulating data containing spatial information. Chapters 4 and 5 introduce raster and vector data, respectively, along with the tools to manipulate them. Chapter 6 deals with the methods used to combine raster and vector data, which is often necessary in social science research. Finally, Chapters 7 to 10 each present a case study to facilitate a hands-on learning approach.
This book uses the R programming language as its primary tool, introducing necessary packages and techniques as they are required. While this book is designed so that no prior experience with spatial data is necessary, a basic familiarity with R and tabular data will be useful.
Remote Sensing, The Science
There are many, and often sophisticated, ways of defining remote sensing (such as characterizing it as the detection of electromagnetic radiation from a distance), but I find it useful to think of remote sensing in terms of something we’re all familiar with—taking photographs. When we take a photo of an object with our cellphone or a digital camera, the light reflecting from the object enters the camera through the lens to a photosensor, and the sensor detects the amount of light coming from the object, records the value, and this value is reconstructed on our computers or phones to produce the photograph we see on our screen.1
-
The principle for old analog cameras is the same, only the photosensor, instead of being digital, is a photographic film which undergoes a different amount of chemical reaction based on the amount and type of light it receives. This chemically altered film can then be reconstructed into a photograph. ↩