Brickman
What we call the “Brickman” data, is a suite of outputs of model runs using the BNAM. Runs were made for “PRESENT” (1982–2013), 2055 and 2075. For the latter two parameters were set to account for two scenarios: RCP4.5 (intermediate case) and RCP8.5 (business as usual case). For all years and scenarios both annual mean and monthly mean projections were produced.
This is a link to the paper describing the model construction and outputs.
We have developed a small database of these scenarios and years, and we provide tools to allow you so easily read these into data objects that you can manipulate.
First we can read in a small table (aka database) that provides information about the files we have provided.
DB = brickman_database() |>
print()## # A tibble: 82 × 4
## scenario year interval var
## <chr> <chr> <chr> <chr>
## 1 PRESENT PRESENT ann MLD
## 2 PRESENT PRESENT ann Sbtm
## 3 PRESENT PRESENT ann SSS
## 4 PRESENT PRESENT ann SST
## 5 PRESENT PRESENT ann Tbtm
## 6 PRESENT PRESENT ann U
## 7 PRESENT PRESENT ann V
## 8 PRESENT PRESENT ann Xbtm
## 9 PRESENT PRESENT mon MLD
## 10 PRESENT PRESENT mon Sbtm
## # ℹ 72 more rows
Your might be wondering “what are these variables?” Some details can be
found by calling brickman_variables(),
brickman_variables("all")## # A tibble: 18 × 5
## name group longname description units
## <chr> <chr> <chr> <chr> <chr>
## 1 depth static Bathy_depth bathymetric depth "m"
## 2 mask static land_mask land mask ""
## 3 SST ann SST_ann annual SST "C"
## 4 SST mon SST monthly SST "C"
## 5 MLD ann MLD_ann annual mixed layer depth "m"
## 6 MLD mon MLD monthly mixed layer depth "m"
## 7 SSS ann SSS_ann annual SSS "psu"
## 8 SSS mon SSS monthly SSS "psu"
## 9 Tbtm ann Tbtm_ann annual Tbtm "C"
## 10 Tbtm mon Tbtm monthly Tbtm "C"
## 11 Sbtm ann Sbtm_ann annual Sbtm "psu"
## 12 Sbtm mon Sbtm monthly Sbtm "psu"
## 13 Xbtm ann Xbtm_ann annual Xbtm "unknown"
## 14 Xbtm mon Xbtm monthly Xbtm "unknown"
## 15 U ann U_ann annual U "m/s"
## 16 U mon U monthly U "m/s"
## 17 V ann V_ann annual V "m/s"
## 18 V mon V monthly V "m/s"
Read in model data by first filtering the database to just the files you want. It’s important to keep in mind that to keep things simple you’ll want to read in “shape” compatible groups. For instance, it is complicated to read into one R object a dataset that mixes single-layer static variables (like depth) with multi-layer variables (like monthly SST). So, we have put some restrictions on the groupings you can read in one go.
Let’s start by reading in the static variables.
db = DB |>
dplyr::filter(interval == "static")
static_vars = read_brickman(db)
static_vars## stars object with 2 dimensions and 2 attributes
## attribute(s):
## Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
## depth 5 60.2631 145.0126 923.3138 1703.552 4964.409 4983
## mask 1 1.0000 1.0000 1.0000 1.000 1.000 4983
## dimension(s):
## from to offset delta refsys point x/y
## x 1 121 -74.93 0.08226 WGS 84 FALSE [x]
## y 1 89 46.08 -0.08226 WGS 84 FALSE [y]
OK, we read two static variable: depth and mask. The printout shows
a statistical summary of each variable including how many NAs (missing
or “not available”) values. Note that mask has pixel values of either
1 or NA. It also shows something about the geometry/dimensions. In
this case our variables have just x (east-west) and y (south-north)
dimensionalities.
That’s it. Filter the database to what you want and then read the data.
And, now, we can do the usual stuff with the space-time arrays (aka
stars). Of course, these are static variables meaning they don’t
change in time - there’s no time-iness to them, but whatever.
plot(static_vars['depth'], axes = TRUE)
Let’s try some non-static variables by loading monthly data - say RCP85 for 2075. So, we filter the table and then read those variable.
db = DB |>
dplyr::filter(scenario == "RCP85",
year == 2075,
interval == "mon")
x = read_brickman(db)
x## stars object with 3 dimensions and 8 attributes
## attribute(s):
## Min. 1st Qu. Median Mean 3rd Qu.
## MLD 1.01127517 5.188425422 15.077325821 19.203468161 28.110147476
## Sbtm 22.72784424 32.119906425 34.454338074 33.564421845 34.941211700
## SSS 15.61822033 30.504541874 30.948341370 31.406357837 31.939316273
## SST 0.84497827 7.999758005 13.929839611 13.630253277 19.051631451
## Tbtm 0.74468768 4.001685858 7.768758297 7.369697369 9.249178886
## U -0.22471914 -0.010182445 -0.002215136 -0.009042294 0.001126237
## V -0.19030032 -0.009814290 -0.002245136 -0.007339959 0.001396833
## Xbtm -0.04662703 0.001123656 0.002673889 0.007115667 0.006559871
## Max. NA's
## MLD 111.33813477 59796
## Sbtm 35.70283508 59796
## SSS 36.40741730 59796
## SST 27.80369568 59796
## Tbtm 25.78357315 59796
## U 0.09090719 59796
## V 0.08733927 59796
## Xbtm 0.21393159 59796
## dimension(s):
## from to offset delta refsys point values x/y
## x 1 121 -74.93 0.08226 WGS 84 FALSE NULL [x]
## y 1 89 46.08 -0.08226 WGS 84 FALSE NULL [y]
## month 1 12 NA NA NA NA Jan,...,Dec
Whoa! Eight variables and now with a third dimension (“month”). Things
get complicated in a hurry, but no worries. Let’s plot the time series
for the sea floor temperature (Tbtm).
plot(x['Tbtm'], axes = FALSE)
We provide a small dataset of select observational buoys from the Gulf of Maine. The data provides simple metadata and location information. For any of these buoys you can use the NERACOOS Mariner’s Dashboard to click on a buoy such as Buoy E01 located off of the midcoast area.
buoys = gom_buoys()
buoys## Simple feature collection with 6 features and 3 fields
## Geometry type: POINT
## Dimension: XY
## Bounding box: xmin: -70.4277 ymin: 42.3233 xmax: -65.9267 ymax: 44.10163
## Geodetic CRS: WGS 84
## # A tibble: 6 × 4
## name longname id geometry
## * <chr> <chr> <chr> <POINT [°]>
## 1 wms Western Maine Shelf B01 (-70.4277 43.18065)
## 2 cms Central Maine Shelf E01 (-69.3578 43.7148)
## 3 pb Penobscot Bay F01 (-68.99689 44.05495)
## 4 ems Eastern Maine Shelf I01 (-68.11359 44.10163)
## 5 jb Jordan Basin M01 (-67.88029 43.49041)
## 6 nec Northeast Channel N01 (-65.9267 42.3233)
plot(x['Tbtm'] |> dplyr::filter(month == "Jan"), axes = FALSE, reset = FALSE)
plot(sf::st_geometry(buoys), col = "orange", pch = 20, add = TRUE)
text(buoys, labels = buoys$id, adj = c(1,1), col = "blue", cex = 0.8)
Now we can extract the data for each month (12) and each variable (8) for each buoy (6). There are two shapes of results you can request. “long” form (the default) in which each row holds one variable for one month for each point. It will have 12 x 8 x 6 = 576 rows. It may feel awkward to work with long form data at first, but this form is super handy when manipulating data. The other form, “wide”, has more human-eye appeal because each row has the data for one point at one month but has all of the variables. “wide” form results will have 12 x 6 = 72 rows. We’ll extract each for below to compare.
First the default “long” form.
long_values = extract_brickman(x, buoys)
long_values## # A tibble: 576 × 4
## point name month value
## <chr> <chr> <chr> <dbl>
## 1 p1 MLD Jan 28.2
## 2 p1 MLD Feb 26.2
## 3 p1 MLD Mar 14.4
## 4 p1 MLD Apr 7.63
## 5 p1 MLD May 4.15
## 6 p1 MLD Jun 2.45
## 7 p1 MLD Jul 2.16
## 8 p1 MLD Aug 2.74
## 9 p1 MLD Sep 6.85
## 10 p1 MLD Oct 14.8
## # ℹ 566 more rows
Each row holds just one measurement for one variable, and, as predicted, we have 12 x 8 x 6 = 576 rows.
Now let’s extract the same data, but this time to a “wide” form. Note that we end up with the predicted 12 x 6 = 72 rows. Now we have all 8 variable measurements per point per month on a single row.
wide_values = extract_brickman(x, buoys, form = "wide")
wide_values## # A tibble: 72 × 10
## point month MLD Sbtm SSS SST Tbtm U V Xbtm
## <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 p1 Jan 28.2 32.5 31.1 5.38 7.18 -0.00114 0.000563 0.000373
## 2 p1 Feb 26.2 32.2 31.2 3.94 6.23 0.000150 0.00149 0.000659
## 3 p1 Mar 14.4 32.0 30.8 3.72 5.61 0.000291 -0.00401 0.00141
## 4 p1 Apr 7.63 32.0 30.2 5.62 5.77 0.00138 -0.00406 -0.000471
## 5 p1 May 4.15 32.0 30.0 9.45 6.40 0.00124 0.000674 0.000503
## 6 p1 Jun 2.45 32.2 29.8 14.0 6.86 0.00120 0.00152 0.000690
## 7 p1 Jul 2.16 32.4 30.0 17.2 7.10 0.0000825 0.00151 0.000380
## 8 p1 Aug 2.74 32.4 30.3 18.5 7.24 0.0000935 0.000438 -0.0000370
## 9 p1 Sep 6.85 32.4 30.5 17.5 7.42 -0.000390 -0.00102 0.000375
## 10 p1 Oct 14.8 32.4 30.7 14.2 8.06 0.000132 -0.000916 0.000586
## # ℹ 62 more rows
Just what you want to do with the data may help you determine which form you prefer.