Skip to content

Latest commit

 

History

History
73 lines (37 loc) · 3.76 KB

README.md

File metadata and controls

73 lines (37 loc) · 3.76 KB

Geospatial Analysis with High-performance computing (HPC)

This course is under development by the GIS Science for Sustainable Transitions (GIST) Lab at Aalto University. The HPC available is provided by the IT Center of Science (CSC).

Lessons

1. Cell Tower aggregation by country worldwide

This lesson is focused in using Parallel Computing resources using Dask-Geopandas. By using a global dataset of Cell Tower's locations we will attribute the country to every tower using Spatial Join (Overlay) and estimating the performance of the parallel computing and and single-core computing.

Open the lesson here 👉 Lesson1-Notebook

The result:

map1

2. Shortest Path analysis in the Helsinki Region - Home to city center

This lesson is focused in computing the Shortest Path (parallel in 16 cores) from every available OSM residential buildings in the Helsinki Region to Rautatieasema. The notebook contain a step-by-step guide of the Shortest Path process using the available cores mainly in Finding the closest nodes, Computing the Shortest Path, and From nodes to path creation.

Open the lesson here 👉 Lesson2-Notebook

The result:

map2

3. Land cover classification using Random Forest - Lapland

This lesson is focused in processing point data and Earth Observation (EO) layers in regions of Finland. The case example is processed in Lapland province. This exercise was developed in cooperation with SYKE and their tool pointEO that made the process handy.

Open the lesson here 👉 Lesson3-Notebook

The result:

map3

4. Merging Tiled Vector Data into country-wide layers

This lesson is focused in a simple method to split a problem into smaller sub-problems that can be solved in parallel using Slurm scheduler (sbatch). It uses tile vector information from Paituli and the Topographic Data Base (TDB) of the National Land Survey (NLS). As a Geoportti service the tiles have been merged back into country-wide layers which are available in Paituli as well. In this exercise we go through one way this can be achieved.

Open the lesson here 👉 Lesson4-Notebook

The result:

map4

5. OvertureMaps Data - From Local to Global analysis with Arrow

In this Lesson, we will fetch Overture Maps data from the local level in Helsinki and the national level of Finland using tags like Buildings and Points of Interest (POI). Then, we will escalate the fetching process at the global level using grids and storing data in parquet format on our local disk. We will do an count analysis using a global dataset. This exercise can help downloading data in parallel and reading big data with in-memory libraries like Arrow.

Open the lesson here 👉 Lesson5-Notebook

The result:

map4

Attribution

GIST (2023). Geospatial Analysis with High-performance Computers using Core-parallelization. Aalto University. Website gistlab.science

Contact

Please, contact the personnel on charge if you have questions about this material:

- Developer: Bryan Vallejo