Incorrect Rounding on getTerraClim Palmer Results?

[pjgoodling]

Hi Mike,

I've noticed that the results from getTerraClim seem to be truncated to 1 decimal place, but do not have the correct rounding when compared with TerraClim data retrieved from the Climate Engine. Here's an example:

#Setting up the lat/long point to extract data from
pts <- data.frame(ID = 1,
                  lat = 38.31250,
                  lon = -75.14583)

#Extracting the data for a few months in 2001
getTerraClim(AOI=pts[1,2:3],param=c('palmer'),startDate = "2001-06-01",endDate = "2001-10-31")

#Output 
    source     lat       lon    date palmer
1 terraclim 38.3125 -75.14583 2001-06    1.2
2 terraclim 38.3125 -75.14583 2001-07    1.5
3 terraclim 38.3125 -75.14583 2001-08    1.9
4 terraclim 38.3125 -75.14583 2001-09    1.4
5 terraclim 38.3125 -75.14583 2001-10   -1.1

The palmer results have 2 significant figures.

When I go to The Climate Engine and retrieve the same dataset using the "Make a Graph" function, I get the following results:

#Variable: PDSI
#Data Source: TERRACLIMATE 4000 m (1/24-deg) monthly dataset (University of Idaho)
#Missing Value: -9999
#Time Period: 2001-6-01 to 2001-10-01
#Point (Lat,Lon): 38.3125N,75.1458W
#Date(yyyy-mm-dd) PDSI
2001-06-01 1.2200
2001-07-01 1.5200
2001-08-01 1.9700
2001-09-01 1.4400
2001-10-01 -1.1600

It seems like for August and October 2001, the rounding was incorrectly applied. A few other spot checks shows that the data is truncated rather than rounded. I have only looked through the palmer results, and haven't checked any of the other TerraClim variables.

I was looking through the "fast download" code, but couldn't trace the issue.

Thanks!

Phillip

[mikejohnson51]

Hi Phillip,

Thanks for the detailed outline of the issue. This appears to go back to how the PDSI values have been aggrageted on the Northwestern Server. Please see below:

library(ncdf4)
# Point to the Aggregated Dataset
url = 'http://thredds.northwestknowledge.net:8080/thredds/dodsC/agg_terraclimate_PDSI_1958_CurrentYear_GLOBE.nc?PDSI[521:1:525][1240:1:1240][2516:1:2516]'

# Open stream to netcdf file
nc = ncdf4::nc_open(url)

# Extract PDSI values
ncdf4::ncvar_get(nc, "PDSI")
#> [1]  1.2  1.5  1.9  1.4 -1.1

# Point to the yearly file datasets:
url2 = 'http://thredds.northwestknowledge.net:8080/thredds/dodsC/TERRACLIMATE_ALL/data/TerraClimate_PDSI_2001.nc?PDSI[5:1:9][1240:1:1240][2516:1:2516]'

# Open stream to netcdf file
nc2 = ncdf4::nc_open(url2)

# Extract PDSI values
ncdf4::ncvar_get(nc2, "PDSI")
#> [1]  1.2  1.5  1.9  1.4 -1.1

^{Created on 2019-11-25 by the reprex package (v0.3.0)}

In both cases the data truncation/rounding have been done on the data preparation end. ClimateR is just reading the data from these resources. I can try emailing them again on this front.

Sorry I cant be more help!

Mike

[pjgoodling]

Thanks for that quick response, Mike! Its frustrating it's on their end, but reassuring it isn't a problem with ClimateR!

Phillip

[mikejohnson51]

@pjgoodling

Its been a while but I am confirming this is still the case:

# Point to the Aggregated Dataset
url = 'http://thredds.northwestknowledge.net:8080/thredds/dodsC/agg_terraclimate_PDSI_1958_CurrentYear_GLOBE.nc?PDSI[521:1:525][1240:1:1240][2516:1:2516]#fillmismatch'

# Open stream to netcdf file
nc = RNetCDF::open.nc(url)

# Extract PDSI values
RNetCDF::var.get.nc(nc, "PDSI", unpack = TRUE)
#> [1]  1.2  1.5  1.9  1.4 -1.1

# Point to the yearly file datasets:
url2 = 'http://thredds.northwestknowledge.net:8080/thredds/dodsC/TERRACLIMATE_ALL/data/TerraClimate_PDSI_2001.nc?PDSI[5:1:9][1240:1:1240][2516:1:2516]#fillmismatch'

# Open stream to netcdf file
nc = RNetCDF::open.nc(url2)

# Extract PDSI values
RNetCDF::var.get.nc(nc, "PDSI", unpack = TRUE)
#> [1]  1.2  1.5  1.9  1.4 -1.1