Vector data cubes for changing geometries

Data cubes and structures
Spatial Data Science across Languages
Prague, Sept. 2024

Lorena Abad
lorena.abad@plus.ac.at

University of Salzburg


Data structures in R

Data frames

Data cubes

A spatial object

polygons
Simple feature collection with 25 features and 2 fields
Geometry type: POLYGON
Dimension:     XY
Bounding box:  xmin: -0.2974337 ymin: -0.00297557 xmax: 0.9730806 ymax: 1.153558
Geodetic CRS:  WGS 84
First 10 features:
   gid   datetime                       geometry
1    a 2020-10-01 POLYGON ((0.5474949 0.80889...
2    b 2020-10-01 POLYGON ((0.2791708 0.83373...
3    c 2020-10-01 POLYGON ((0.2807462 0.62779...
4    d 2020-10-01 POLYGON ((0.7650701 0.47444...
5    e 2020-10-01 POLYGON ((0.3825692 0.35378...
6    a 2020-10-02 POLYGON ((0.4961102 0.87283...
7    b 2020-10-02 POLYGON ((0.3298312 0.76120...
8    c 2020-10-02 POLYGON ((0.328914 0.568743...
9    d 2020-10-02 POLYGON ((0.7217233 0.52617...
10   e 2020-10-02 POLYGON ((0.3101455 0.31689...

A spatial object

plot(polygons)

Raster data cubes

Vector data cubes

Vector data cubes

Summary geometries

Summary geometries in {post}

Changing geometries

Union

Centroid

Bounding box

Minimum rotated rectangle

Convex hull


VDCs in array form

VDCs as post_array

as_post_array(polygons)
stars object with 2 dimensions and 1 attribute
attribute(s):
         geometry  
 POLYGON      :25  
 epsg:4326    : 0  
 +proj=long...: 0  
dimension(s):
         from to     offset  delta refsys point
geom_sum    1  5         NA     NA WGS 84  TRUE
datetime    1  5 2020-10-01 1 days   Date FALSE
                                                           values
geom_sum POINT (0.647816 0.9018588),...,POINT (0.4690683 0.17772)
datetime                                                     NULL

VDCs in tabular form

VDCs in tabular form

VDCs as post_table

as_post_table(polygons) |> 
  face_spatial()
# cubble:   key: gid [5], index: datetime, nested form, [sf]
# spatial:  [0, 0.18, 0.75, 0.9], WGS 84
# temporal: datetime [date], geometry [POLYGON [°]]
  gid       long   lat                 geom_sum ts          
* <chr>    <dbl> <dbl>              <POINT [°]> <list>      
1 a      0.648   0.902     (0.647816 0.9018588) <sf [5 × 2]>
2 b     -0.00254 0.840 (-0.002543201 0.8403816) <sf [5 × 2]>
3 c      0.390   0.541    (0.3897167 0.5412363) <sf [5 × 2]>
4 d      0.753   0.354     (0.752585 0.3538271) <sf [5 × 2]>
5 e      0.469   0.178      (0.4690683 0.17772) <sf [5 × 2]>

VDCs as post_table

as_post_table(polygons) |> 
  face_temporal()
# cubble:   key: gid [5], index: datetime, long form
# temporal: 2020-10-01 -- 2020-10-05 [1D], no gaps
# spatial:  long [dbl], lat [dbl], geom_sum [POINT [°]]
   gid   datetime                                                       geometry
 * <chr> <date>                                                    <POLYGON [°]>
 1 a     2020-10-01 ((0.5474949 0.8088912, 0.4874872 0.8838477, 0.5520055 1.032…
 2 a     2020-10-02 ((0.4961102 0.8728385, 0.4361026 0.947795, 0.5006209 1.0966…
 3 a     2020-10-03 ((0.5578801 0.8616378, 0.5623908 1.085448, 0.8520717 0.8981…
 4 a     2020-10-04 ((0.5652241 0.872057, 0.5685463 1.036898, 0.7819022 0.89897…
 5 a     2020-10-05 ((0.6063791 0.8304178, 0.5463715 0.9053744, 0.6108898 1.054…
 6 b     2020-10-01 ((0.2791708 0.8337388, -0.2847997 0.6540191, -0.2825181 0.6…
 7 b     2020-10-02 ((0.3298312 0.7612067, -0.2341393 0.5814869, -0.2318577 0.5…
 8 b     2020-10-03 ((0.3796448 0.767853, -0.1843257 0.5881332, 0.09777458 1.05…
 9 b     2020-10-04 ((0.3642467 0.779727, -0.1182199 0.6259799, 0.1231118 1.026…
10 b     2020-10-05 ((0.2665368 0.8649996, -0.2974337 0.6852798, -0.2951521 0.6…
# ℹ 15 more rows

Applications









Research supported by the Austrian Research Promotion Agency (FFG) through the project ReHIKE (“Analysing Glacier Retreat Effects on Alpine Hiking Infrastructure using Earth Observation”; project number: FO999911954).

Applications

Svalbard glaciers

Minimum rotated rectangle

Centroids

Applications

Time series analysis

What’s next for me?

  • Set-up functions for raster-vector data cube interaction with time-matching
    • st_extract(rdc, vdc)
    • st_crop(rdc, vdc)
  • Work on a concept to group changing geometries when group IDs are unknown
  • Application cases in my RG projects

What could be next?

  • Find out ways to save post objects to file
    • zarr files are an option, and in general something to look into for vector data cubes
    • {xvec} issue
  • Test implementations for Python (and Julia?)

Credits

Cover photo by Etienne Girardet on Unsplash

Pasterze glacier photo by Karsten Würth on Unsplash

References

remotes::install_github(“loreabad6/post”)

König, Max, Christopher Nuth, Jack Kohler, and Moholdt. 2014. “A Digital Glacier Database for Svalbard.” In Global Land Ice Measurements from Space, 229–39. Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-540-79818-7_10.
Pebesma, Edzer, and Roger Bivand. 2023. Spatial Data Science with Applications in R. Geographical Analysis. 1st ed. Chapman & Hall. https://r-spatial.org/book/.
Wickham, Hadley. 2019. Advanced R. CRC Press.