equally-sized data.frame error

[ianmseddy]

Hello, I am using climr@devl v 0.2.23 with a data frame of around 20K points. The id column is unique, but every id does not have a unique location (there are a few duplicated points). My initial attempts have produced the error shown below. This has run for me if I change the obs_years and obs_ts_dataset so I'm unsure if my request is the issue?

climr_reprex_data.csv

#> packageVersion("climr")
#[1] ‘0.2.23’

test <- read.csv("https://github.com/user-attachments/files/23940237/climr_reprex_data.csv")

vars <- c("AHM", "FFP", "CMD_an", "CMI_an", "SHM", "NFFD_an",
          "MAP","MAT", "MCMT", "MSP", "MWMT", "DD5_an", "DD18_an", "DDsub0_an")


test_obs_climate <- climr::downscale(xyz = test, obs_years = 1930:2022,
                                    obs_ts_dataset = "climatena", vars = vars)

Welcome to climr!
Downscaling in database...
Extracting [73] bands from refmap_climr
Extracting [3,348] bands from historic_climatena
Processed 64629793 groups out of 64629793. 100% done. Time elapsed: 502s. ETA: 0s..
Error in Ops.data.frame(climaterast[, ppt_offset], res[, match(labels[ppt_],  : 
  ‘*’ only defined for equally-sized data frames

[kdaust]

Hi Ian - thanks for this! Yep, definitely seems to be a bug there.

That being said - in this situation, where you're downscaling a large number of point, I would use local downscaling (i.e. db_option = "local"). This will we much faster than on-db downscaling, and also seems to fix the error you were having. For example, the following runs fine for me using 0.2.23:

library(data.table)
test <- fread("https://github.com/user-attachments/files/23940237/climr_reprex_data.csv")

vars <- c("AHM", "FFP", "CMD_an", "CMI_an", "SHM", "NFFD_an",
          "MAP","MAT", "MCMT", "MSP", "MWMT", "DD5_an", "DD18_an", "DDsub0_an")

bb <- get_bb(xyz = test)
pre_cache(bbox = bb)

test_obs_climate <- climr::downscale(xyz = test, obs_years = 1930:2022,
                                     obs_ts_dataset = "climatena", vars = vars,
                                     db_option = "local")

Note that I used the pre_cache function to speed up caching the high-res reference climatology. It should also work without doing that, but will take longer the first time.

Let me know if this works for you!

[ianmseddy]

Thanks Kiri! That was definitely faster, although now I have more NAs in the climateNA product. So.. I see the six NA near Haida Gwaii - clearly a coordinate typo because they're supposed to be on land - but then there's a random assortment of ~1300 more that are all NA in climateNA, for every year. Do you think this might be the result of corrupted download? The NAs seem to be tiled when I plot them

> test_obs_climate[is.na(MAT), .N, .(PERIOD, DATASET)]
     PERIOD    DATASET     N
     <char>     <char> <int>
  1:   1920  climatena  1376
  2:   1921  climatena  1376
  3:   1922  climatena  1376
  4:   1923  climatena  1376
  5:   1924  climatena  1376
 ---                        
306:   2018 mswx.blend     6
307:   2019 mswx.blend     6
308:   2020 mswx.blend     6
309:   2021 mswx.blend     6
310:   2022 mswx.blend     6

[kdaust]

Hi Ian - I just checked and it looks like we somehow excessively cropped the climatena obs_ts layers on the database. So they don't include Haida Gwaii, much of Van Isl, and a lot of the east coast. We'll work on getting that fixed this week. In the meantime, the other available obs timeseries should be fine still.

[ianmseddy]

Ahh okay, well that is a satisfying resolution. Thanks for the help!