| title | tags | authors | affiliations | date | bibliography | ||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Open-Geo-Tutorial: a modern open-source toolkit for geospatial and remote sensing analysis in the earth sciences using python |
|
|
|
6 November 2019 |
paper.bib |
The python ecosystem for scientific computing, visualization, geospatial analysis, and remote sensing analysis is now stable, powerful, and largely open-source. This allows both totally new analyses to be conducted compared to traditional GIS and remote sensing applications and opens these incredible tools to an entirely new community. Additionally, thanks to tools such as Docker and Anaconda, these tools are now both more reproducible than ever and much simpler to run on any operating system.
While much of this material can be learned via the documentation for the individual packages we introduce here, we find scientists
early in their career, or unfamiliar with programming, are not initially comfortable with technical documentation and struggle to find
the appropriate tools in the massive python ecosystem. open-geo-tutorial is a series of labs introducing students initially to
fundamental remote sensing and GIS methodologies and gradually builds to more challenging and complex techniques such as raster
processing, vector analysis, image classification using machine learning, time series analysis, parallalization, linkages with Google
Earth Engine, and more all using modern open source software in Python (rasterio, shapely, GeoPandas, folium, xarray, etc).
The materials included here should take the student anywhere from 4-8 hours to work through, are designed to be worked through independently, and may serve as a reference for the advanced geospatial and remote sensing analyst.
Given the preponderance of academic programs in the earth and ocean sciences, geography, and even strong remote sensing programs that teach students graphical based geospatial and remote sensing analysis (often in expensive proprietary software applications) we found that students were often poorly equipped to analyze complex, highly dimensional, and large earth science datasets. Computer scientists coming into the earth sciences on the other hand often have the programming background, but not the intuition about geospatial data or an understanding of the data sources available and the domain specific analysis techniques. And neither of these groups are usually well versed in the modern tools that will be enable them to tackle challenging geospatial and remote sensing analyses as a part of a scientific team such as visualization highly dimensional datasets, machine learning, parallalization of tasks, and reproducibility [@Wilson2014].
This work is aimed at both of those groups, earth scientists looking to improve their python and ability to tackle much larger problems, and students with experience programming but little background in geospatial and remote sensing analyses. Our goal is to give a broad taste of the tools needed to work through challenging earth science problems, from downloading data from a NASA DAAC, to running code in a reproducible Docker container, to the python packages and code you'll need to run to explore the data. We think this work is especially timely as the earth science community begins to tackle the complex interconnectedness of the earth system, new remote sensing platforms begin delivering more data at incredible spatial and temporal resolution, and machine learning tools are enabling nuanced insights into massive data streams.
We acknowledge funding support from the North Carolina Space Grant Graduate Research Fellowship which permitted time to develop this open-source series and the Duke Bass Connections program for providing the space to test this series with undergraduate students.