Merge pull request PecanProject#4012 from anshul23102/fix/align-met-rep-each-length

fix(align.met): use length(stamps.hr) instead of stamps.hr as the rep() each argument

Author: infotroph

modules/data.atmosphere/NEWS.md

@@ -1,3 +1,9 @@
+# PEcAn.data.atmosphere 1.9.1.9000
+
+## Fixed
+
+* `align.met()` now correctly calculates output timesteps in cases where the source data is at a coarser time step than the training data and `align == "repeat"` (#4012, @anshul23102).
+
 # PEcAn.data.atmosphere 1.9.1
 
 ## Added

modules/data.atmosphere/R/align_met.R

@@ -428,7 +428,7 @@ align.met <- function(train.path, source.path, yrs.train=NULL, yrs.source=NULL,
           dat.tem <- ncdf4::ncvar_get(ncT, v)
 
           if(align=="repeat"){ # if we need to coerce the time step to be repeated to match temporal resolution, do it here
-            dat.tem <- rep(dat.tem, each=stamps.hr)
+            dat.tem <- rep(dat.tem, each=length(stamps.hr))
           }
           df.tem <- matrix(rep(dat.tem, n.src), ncol=1, byrow=F)
 

modules/data.atmosphere/tests/testthat/test.align_met.R

@@ -0,0 +1,79 @@
+## Tests for align.met()
+##
+## The bug fixed in this PR (line 431, ensemble source path):
+##   rep(dat.tem, each = stamps.hr)
+## where stamps.hr is a numeric *vector*.  R silently takes stamps.hr[1]
+## and truncates to an integer.
+##
+## For hourly training data, stamps.hr = c(0.5) (the first centred hour stamp).
+## Truncating 0.5 to 0 makes rep() return an empty vector, so dat.source
+## ends up with zero rows -- silently discarding all source data.
+##
+## Fix: rep(dat.tem, each = length(stamps.hr))
+## The single-time-series source path (line 304) already used length() correctly.
+
+## Helper: create a minimal single-variable NetCDF with a given number of
+## time steps placed in outdir/filename.  Time dimension is in fractional days
+## since 2001-01-01, matching the step size implied by n_time.
+make_align_nc <- function(n_time, outdir, filename) {
+  time_dim <- ncdf4::ncdim_def(
+    name  = "time",
+    units = "days since 2001-01-01",
+    vals  = seq(0, by = 1 / (n_time / 365), length.out = n_time)
+  )
+  temp_var <- ncdf4::ncvar_def(
+    name    = "air_temperature",
+    units   = "K",
+    dim     = list(time_dim),
+    missval = -9999
+  )
+  nc <- ncdf4::nc_create(file.path(outdir, filename), vars = list(air_temperature = temp_var))
+  on.exit(ncdf4::nc_close(nc), add = TRUE)
+  ncdf4::ncatt_put(nc, 0, "description", "synthetic data for align.met test")
+  ncdf4::ncvar_put(nc, temp_var, vals = seq(280, length.out = n_time, by = 0.01))
+  invisible(file.path(outdir, filename))
+}
+
+test_that("align.met ensemble source path produces non-empty source data (bug: each=0 from stamps.hr truncation)", {
+  train_dir  <- withr::local_tempdir()
+  source_dir <- withr::local_tempdir()
+
+  ## Hourly training (8760 steps) produces stamps.hr[1] == 0.5, intentionally
+  ## chosen so that any truncation of the step count to integer yields 0.
+  make_align_nc(n_time = 8760, outdir = train_dir, filename = "2001.nc")
+
+  ## Source: daily (365 steps), placed inside an ensemble subfolder.
+  ens_dir <- file.path(source_dir, "ens001")
+  dir.create(ens_dir)
+  make_align_nc(n_time = 365, outdir = ens_dir, filename = "2001.nc")
+
+  result <- align.met(
+    train.path  = train_dir,
+    source.path = source_dir,
+    n.ens       = 1,
+    seed        = 20260602
+  )
+
+  ## 8760 hourly training rows; 365 source rows (one per day, repeated once each).
+  expect_equal(nrow(result$dat.train$air_temperature), 8760)
+  expect_equal(nrow(result$dat.source$air_temperature), 365)
+})
+
+test_that("align.met single-series source matches training row count when already aligned", {
+  train_dir  <- withr::local_tempdir()
+  source_dir <- withr::local_tempdir()
+
+  ## Both training and source at the same 3-hourly resolution (2920 steps for 2001).
+  make_align_nc(n_time = 2920, outdir = train_dir,  filename = "2001.nc")
+  make_align_nc(n_time = 2920, outdir = source_dir, filename = "2001.nc")
+
+  result <- align.met(
+    train.path  = train_dir,
+    source.path = source_dir,
+    n.ens       = 1,
+    seed        = 20260602
+  )
+
+  expect_equal(nrow(result$dat.source$air_temperature),
+               nrow(result$dat.train$air_temperature))
+})