forked from carpentries-incubator/geospatial-python
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
89aefee
commit 64d33d2
Showing
3 changed files
with
90 additions
and
2 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,87 @@ | ||
# Updates for Introdcution Geospatial python | ||
|
||
**Context** | ||
The original course exists of the following episodes: | ||
|
||
|episode|summary| | ||
|----|----| | ||
|0. Summary and Setup | setup for the software | | ||
|1. Introduction to Raster Data | theory about Raster data format | | ||
|2. Introduction to Vector Data | theory about Vector data format | | ||
|3. Coordinate Reference Systems | theory about project coordinate systems| | ||
|4. The Geospatial Landscape | explanation about various tools and why python would be a good idea to use. | | ||
|5. Access satellite imagery using Python | STAC catalogue and data retrieval using python. Based on a xy coordinate an overview is generated on the available satellite images | | ||
|6. Read and visualize raster data | A series of processing and calculations are provided on the satellite image that is accessed in step 5 | | ||
|7. Vector data in Python | Data about water wells and crop plots are provided on which a couple of vector analyses are performed. Basically, a point in polygon analysis is performed to see which wells are in a crop plot [7.1]. Furthermore, a flipped dataset is corrected but nothing is done with it.| | ||
|8. Crop raster data with rioxarray and geopandas | the vector dataset from 7.1 is cropped and clipped with the dataset from step 5. | | ||
|9. Raster Calculations in Python | NDVI on raster file that was obtained in step 5, resulting in a classification on vegetation density. | | ||
|10. Calculating Zonal Statistics on Rasters | Zonal statistics performed to relate the crop type from brp to the classified NDVI dataset. Before that a vector to raster conversion is performed. | | ||
|11. Parallel raster computations using Dask | parallel raster computation is performed to remove noise from the original image | | ||
|
||
After a thorough evaluation of the existing material the following observations were made: | ||
|
||
- The various components in the course are good, however not very much related to each other. It lacks coherence. | ||
- The various steps in the workshop can also be done in a GUI (i.e. QGIS, ArcGIS pro etc.), thus GIS experts that would join it would probably assess that 99% of the steps in the workshop could already be done in a GUI (which also has the advantage of being able to explore the data much easier by zooming in and out and being able to very intuitively rearrange the layers in the table of content). The workshop material should thus be updated in such a wy that it shows that strength of coding; rerunning the script with a different configuration (a different region, different time period etc.) | ||
- The used vector data in the workshop it too specific, since it is a subset from a specific Dutch dataportal (i.e. [BRP](https://www.pdok.nl/introductie/-/article/basisregistratie-gewaspercelen-brp-). | ||
|
||
Based on these observations the following starting points for the update have been formulated: | ||
|
||
1. Show the strength of python in terms of reproducibility of an analysis / process. The learners should develop a script / jupyter notebook which they in the end need to rerun with different configuration steps (different settings/ different location). This will allow them to open their eyes that that is much more effective compared to the click and go tools in a GUI. Also, since we are using libraries outside the ecosystem of the GIS software this goes beyond the model builder tools | ||
2. Develop a strong narrative and explain why certain steps are taken. This will allow the learners to better understand what they are doing and also why they are doing so. Introducing a tool or an analysis without a purpose is less effective when teaching something. | ||
|
||
This led to the following restructuring. | ||
|
||
# Updates | ||
|
||
For the narrative we chose to focus on wild fires. Results of large fires, like forest fires or fires on savannah landscapes can be identified on sattelite images. Scorched land leaves a particular signature on satellite images that can automatically be identified. By combining that with data about where people live we can estimate the impact of these large fires on our world, which in theory could also lead to insights on where to take certain mitigation matters. Based on the study of | ||
[Zhai et al., 2023](https://doi.org/10.3390/f14040807) we also decided to include a calculation of slope classes in our workshop, since they identified a correlation slope and forest fires. As for the study area, we decided to focus a the island of Rhodos around July 2023, when large wild fires occured and around 19.000 people needed to be [evacuated](https://en.wikipedia.org/wiki/2023_Greece_wildfires). | ||
|
||
## Restructering | ||
|
||
**Episode 0** (keep) | ||
Setup | ||
|
||
**Episode 1-4** (keep) | ||
For the moment we decided to keep chapters 0 to 4 the way they are (although some updates on the theory would have to be processed at a later state). | ||
|
||
**Episode 5** Access satellite imagery using Python (keep) | ||
After the setup and theory we start with the access to the satellite images and DEM episode. This is the same as episode 5 from the orginal setup. Here we focus at the island of Rhodos and in particular to July 2023. The satellite image we can get does not cover the whole island. The participants will see that the area scorched and will directly see the impact of the wildfire on the island. As a nice to have we can lateron update the material to do some merging / mosaicing. | ||
|
||
**Episode 6** Read and visualize raster data *(refactoring of episode 6)* | ||
Keep in in the way it is, however instead perform it on the DEM (which makes more sense) | ||
|
||
**Episode 7** Vector data in Python *(refactoring of episode 7)* | ||
Once the participants have explored the area, we are going to introduce them to vector data in python. We will let them obtain the geometry from Rhodos and use that as a mask to clip the satellite image with. We can ~~either get the data from the build-in datasets in geopandas or~~ get them from [GADM](https://gadm.org/download_country.html) we can choose to get the [geopackage](https://geodata.ucdavis.edu/gadm/gadm4.1/gpkg/gadm41_GRC.gpkg) or the [shapefile](https://geodata.ucdavis.edu/gadm/gadm4.1/shp/gadm41_GRC_shp.zip) (level-3). With this data they will have to select a polygon by it's attribute. | ||
|
||
Next, we will download data from OpenStreetMaps, for this various options are available see [https://wiki.openstreetmap.org/wiki/Downloading_data](https://wiki.openstreetmap.org/wiki/Downloading_data) of which getting it from [Geofabrik](https://download.geofabrik.de/) is the easiest one. We might lateron consider to include an episode here where we dive into the API in which we configure the download exactly to our area of interest. In the narrative we explain that we want to relate the damage of the wild fires to build-up areas i.e. infrastructure and residential areas. We will either ask them to download the whole of [Greece](https://download.geofabrik.de/europe/greece.html) or prepare a selection of that dataset. | ||
|
||
**1. Roads and Railways:** Since roads are stored as line in our dataset we are going to put a buffer of 10 meters around them. The dataset we will be focussing at is *gis_osm_roads_free_1.shp*, since this dataset also contains hiking paths, we will focus at lines with the values: motorway, primary, secondary, tertiary and residential (we can put in as a disclaimer that the dataquality of OSM can be debated and that it contains all kind of errors, however for the purposes of this workshop working with only these datasets is ok (we can relate this to the whole primary and secondary data capture discussion as well)). In other cases *gis_osm_railways_free_1.shp*, would also be an option, however Rhodos does not have any ;-) . | ||
|
||
**2. Buildings:** The information about the buildings should be constructed by combining the layers *gis_osm_buildings_a_free_1.shp*, *gis_osm_landuse_a_free_1.shp* and *gis_osm_transport_a_free_1.shp*. | ||
2.1 For the buildings we again do a buffer of 10 meter creating a new layer. | ||
2.2 Since not all buildings have been vectorized in OSM we also need to use *gis_osm_landuse_a_free_1.shp* and make a selection on the attribute fclass with the values: | ||
allotments, commercial, industrial, residential (again, a disclaimer should be made about the quality of OSM here!) this will result in a new layer. | ||
2.3 Airports and other transport related areas are in *gis_osm_transport_a_free_1.shp* | ||
|
||
(in this episode we thus drop the point in polygon analysis and the flipping of the coordinates) | ||
|
||
Next, we will need to merge these layers into one layer containing all the valuable assets. Lateron we are going to use this dataset to perform the zonal statistics. | ||
|
||
**Episode 8** Crop raster data with rioxarray and geopandas *(refactoring of episode 8)* | ||
Create subsets from the DEM and the satellite images based on the shape of Rhodos (collected in episode 6). These subsets are then used for the next episodes. | ||
Use `reproject_match` to align DEM and image data. | ||
|
||
|
||
**Episode 9** Raster Calculations in Python *(refactoring of episode 9, including the statistics part originally in EP6)* | ||
Keep this one as much the way it is, however instead focus at the NDVI ranges for scorched areas see: (https://custom-scripts.sentinel-hub.com/sentinel-2/burned_area_ms/). We can use the values provided in that source. For now we leave the SWIR part out. (would be an option to do add that at a later stage). This episode should result in raster with categories of scorched areas (or even a binary raster). | ||
|
||
**Episode 10** Parallel raster computations using Dask *(refactoring of episode 11)* | ||
Keep as much as possible, but convert to calculating the slope-class of a raster. However it seems that the calculation of the slope is so optimised in Xarray/Xarray-spatial that for the area considered parallelization has a negligible effect. It does however appear that there are a couple of errors in the DEM data, as identified by Francesco Nattino. Let us convert this episode to error processing followed by calculating the slope (in order to keep it in line with the narrative). | ||
|
||
**Episode 11** Calculating Zonal Statistics on Rasters *(refactoring of episode 10)*| | ||
Zonal statistics on: | ||
1. Burned areas in relation to buildings and infrastructure | ||
2. Burned areas in relation to slope classes. For this we first need to reclassify the slope classes into 5 categories like [Zhai et al., 2023](https://doi.org/10.3390/f14040807) have done. 0%–5%, 5%–10%, 10%–15%, 15%–25%, and above 25%. Next we perform the zonal statistics. We do not dive into the meaning of this, just showing how it is done. | ||
|
||
**Episode 12** Apply the script to a different time period! | ||
Throughout the whole workshop the participants will developed a script (jupyter notebook). The strenght of this is that they can now reuse it for a different area or a different time stamp. Since we are not using the OSM API applying this script for another area it too laborious. A different time period would be possible though. For this we only need to change the setting for the time stamp. We need to seek for a nice satellite image from before the forest fire. |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters