Box Geometry Model (BGM) files

Michael D. Sumner

2020-04-12

Box Geometry Model files

Box Geometry Model files are text files that store a spatially-explicit model domain used by the Atlantis Ecosystem Model.

The model domain consists of boxes (polygons) composed of faces (line segments) and an overall boundary that delimits the domain of the model.

The boundary will include boundary polygons that are not boxes in the model, but some box edges are on this boundary, for example some models bounded by land do not have boundary boxes on this edge.

The rbgm package includes tools to read the BGM format, returning tables of all X/Y vertices and tables of indices for the boxes and faces.

library(rbgm)
library(dplyr)
mfile <- bgmfiles::bgmfiles()[1]

bgm <- bgmfile(mfile)

names(bgm)
## [1] "vertices"         "facesXverts"      "faces"            "facesXboxes"     
## [5] "boxesXverts"      "boxes"            "boundaryvertices" "extra"

Box and face attributes, and model topology

Both boxes and faces have attributes stored row-wise in the respective table.

These are the box attributes for this example

print(bgm$boxes)
## # A tibble: 35 x 9
##    label nconn  botz        area vertmix horizmix  insideX  insideY  .bx0
##    <chr> <int> <dbl>       <dbl>   <int>    <int>    <dbl>    <dbl> <dbl>
##  1 Box0      7 -53.7 9254788445.       1        1 3430169. 3073000.     0
##  2 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1
##  3 Box2    120 -55.7 3922348664.       1        1 3521055. 3050400.     2
##  4 Box3    105 -53.4 3576072582.       1        1 3523085. 3077300.     3
##  5 Box4     83 -30   1672921591.       1        1 3522699. 3102469.     4
##  6 Box5     12 -24.1  411048577.       1        1 3496968. 3124359.     5
##  7 Box6     16 -10.2  447377917.       1        1 3520947. 3133926.     6
##  8 Box7     18   0    402661184.       1        1 3523883. 3122610.     7
##  9 Box8     13 -15    705100219.       1        1 3562646. 3115917.     8
## 10 Box9      3  -5.4   93312196.       1        1 3581060. 3127380.     9
## # … with 25 more rows

and these are the face attributes.

print(bgm$faces)
## # A tibble: 824 x 7
##    cosine  sine  left right length  .fx0 label
##     <dbl> <dbl> <int> <int>  <dbl> <dbl> <chr>
##  1  0.112 0.994    34    32 29876.     0 face0
##  2  0.506 0.862    32    33 17891.     1 face1
##  3  0.116 0.993    34    33 10169.     2 face2
##  4  0.117 0.993    34    33   155.     3 face3
##  5  0.869 0.494    33    27  9391.     4 face4
##  6  0.118 0.993    34    27 12818.     5 face5
##  7  0.121 0.993    34    27 21316.     6 face6
##  8  0.276 0.961    32    33 25394.     7 face7
##  9  0.514 0.858    32    33  8425.     8 face8
## 10  0.136 0.991    32    30 14366.     9 face9
## # … with 814 more rows

By using relations between these tables we can reconstruct the geometry appropriately. (This is pretty painful, but shows that we have all the information required.)

bgm$boxes %>% 
  filter(label %in% c("Box1", "Box11"))  %>% 
  inner_join(bgm$boxesXverts, ".bx0") %>% 
  head()
## # A tibble: 6 x 10
##   label nconn  botz        area vertmix horizmix  insideX  insideY  .bx0  .vx0
##   <chr> <int> <dbl>       <dbl>   <int>    <int>    <dbl>    <dbl> <dbl> <int>
## 1 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1   157
## 2 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1   178
## 3 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1   222
## 4 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1   219
## 5 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1   218
## 6 Box1     81 -48.9 1996056923.       1        1 3498567. 3029308.     1   217

Convert to Spatial objects

To avoid constructing boxes and faces manually, rbgm provides several helper functions to create the boxes, faces, coordinates, and boundary as Spatial objects from the sp package.

Since Spatial objects cannot store the full connectivity of the BGM structure, it is best to consider these as one-way conversions, keeping all the details in the raw tables as above.

To generate an object with just “Box1” and “Box11”, use the function boxSpatial and then subset on box labels.

boxes <- subset(boxSpatial(bgm), label %in% c("Box2", "Box16"))

To determine which faces belong to these boxes, first create all faces and then join.

allfaces <- faceSpatial(bgm)
faces <- allfaces[match(bgm$facesXboxes$iface[which(bgm$facesXboxes$.bx0 %in% boxes$.bx0)], allfaces$.fx0), ]

Plot the two layers together.

plot(boxes, main = "boxes only")
text(coordinates(boxes), lab = sprintf("%s\nnconn: %s", boxes$label, boxes$nconn))

plot(boxes, main = "boxes and faces overlaid")
text(coordinates(boxes), lab = sprintf("%s\nnconn: %s", boxes$label, boxes$nconn))
plot(faces, col = c("firebrick"), lwd = 1, add = TRUE)

(Future releases will provide simpler tools to do this kind of matching).