Create a function that returns the complete forecast metadata in JSON and XML formatted based on EFI standard.

When running devtools:test() locally with terra 1.1.13, I get the following error:

Error (test-validate.R:72:5): Validation has correctly formatted returns with multiple output
  comparisons
Error: no valid constructor available for the argument list
Backtrace:
  1. PoPS::validate(...) test-validate.R:72:4
  2. PoPS::configuration(config) /home/vpetras/Projects/pops/rpops/R/validate.R:165:2
  3. PoPS::initial_raster_checks(config$infected_file) /home/vpetras/Projects/pops/rpops/R/configuration.R:130:4
  5. terra::rast(x)
  6. terra:::.local(x, ...)
  7. SpatRaster$new(f, subds - 1, "", FALSE, 0[])
  8. methods::new(def, ...)
 10. methods::initialize(value, ...)
 11. .Object$initialize(...)
 12. Rcpp::cpp_object_initializer(.self, .refClassDef, ...)
 13. Rcpp:::new_CppObject_xp(fields$.module, fields$.pointer, ...)

> packageVersion("terra")
[1] ‘1.1.13’

However, I don't see the in the CI.

seems warnings are being thrown from RCPP so will probably not fix in the near future.

• spread_rates

• Quarantine stats

• exposed

• output timing

• pesticide treatments

• helpers
• auto_manage
• uncertainty_propogation
• over-population (individuals disperse longer distance in an area of high population)
• movements
• mortality
• deterministic spread

move weather coefficient functionality from https://github.com/ncsu-landscape-dynamics/weather-coefficient.

Currently, the ABC calibration is only set up to handle a single point in time comparison of model output to observed data. In some cases, we may want to calibrate multiple outputs at once.

Created by Chris via monday.com integration. 🎉

Allows testing on windows, Mac, and Linux.

Currently model only provides outputs to the user on a yearly interval. Update to output at a user-specified interval. Doing so should allow for the model to be shifted toward an iterative disease-forecasting system.

With a fresh dev version install of PoPS, (e.g. devtools::install_github("ncsu-landscape-dynamics/rpops") it will install terra 1.1-4 from CRAN if terra is not already installed. However, the dev version of PoPS now depended on a version of terra (1.1.7) only on github.

If terra was originally installed from github, it will prompt to update. If terra was originally installed from CRAN, the latest version of PoPS will fail to install. You can of course manually update terra from github, but I don't believe this is the desired behavior.

This is necessary for the pops model API for the website which uses rpops as a dependancy

This is mostly likely due to a parameter draw being outside of the reasonable bounds and easily fixable

Affects multirun and validate functions. Doesn't cause any failures but causes unintended reuse of same parameters.

Potentially a package loading issue in the test file or a system issue but happens both locally and in CI when run using R CMD Check but not when used locally.

Use last years output as the prior for the new year so we don't have to run the entire time series of data for calibration but can update it quickly as new information becomes available by only running the new data. Need to figure out how to weight the priors from previous years.

All Travis builds failing while passing locally and on Appveyor.

Terra processing is faster

Treatment should be able to be implemented at any point in the year as specified by the user.

Problem description

I'm not sure how exactly the terra versions are numbered, but there is couple of commits in Git which have 1.1-7 in the DESCRIPTION file. Just any of them is not enough. The first commit rpops now works with is rspatial/terra@cd4c35c.

Expected behavior, suggested solution

Notably, this will resolve itself over time (when the version we need is on CRAN) and I'm not sure what is the right for now.

The terra package is changing rapidly and we are currently using the development version from GitHub, so maybe using specific version in readme which is guaranteed to work?

devtools::install_github("rspatial/terra", ref="cd4c35c7818adc26be896d65d130070843fb61c5")

And/or using 1.1.8 (1.1-8) in DESCRIPTION file?

On the other hand, in the CI we are simply using the latest version/commit, so with active development in rpops (or a scheduled action we set up), we discover a potential incompatibility with terra here in the repo.

More info

Currently, you can get things like:

> devtools::load_all()
...
Warning message:
object ‘patches’ is not exported by 'namespace:terra' 
> packageVersion("terra")
[1] ‘1.1.7’

Discovered when writing automated tests. Will be fixed with a minor change to pops.cpp.

Add configuration setup to reduce redundant code across functions.

Add the extra setup for auto management to the configuration function and use configuration function in auto management.

Update all functions that use pops_model to take the vector of means and the cov_matrix from the abc calibration to draw from the current parameter distribution. This should will work even if the user wants to not use the distribution as they can pass in a cov matrix of all 0's and then each draw will just be the mean.

When running the model using timestep of days, yearly outputs, the model returned exposed but not other metrics. 2 problem spots come to mind

• export statement in pops_model function

• scheduler functionality

Here is the example code that doesn't work it is in abc calibration but the issue is calling the pops_model function.
'''
infected_years_file <- ffIn("purelines_7weeks.tif")
number_of_observations <- 68 ### This is the number of infected cells - just make sure it's consistent across years
prior_number_of_observations <- 0
prior_means <- c(0, 0, 0, 0, 0, 0) ### leave as 0 for now, means that you are giving them no weight
prior_cov_matrix <- matrix(ncol = 6, nrow = 6, 0)
params_to_estimate <- c(T, T, T, T, F, F) ### 1st: reproductive rates, 2nd: natural distance, 3rd: percent natural, 4tH: anthropogenic distance, 5th Natural Kappa, 6th anthropogenic kappa
number_of_generations <- 6
generation_size <- 50
checks = c(60, 20000, 1000, 10000)
infected_file <- ffIn("Pure_infected.tif")
host_file <- ffIn("Pure_host.tif")
total_plants_file <- ffIn("Pure_host.tif")
temp <- FALSE
temperature_coefficient_file <- ""
precip <- FALSE
precipitation_coefficient_file <- ""
model_type = "SEI"
latency_period = 14
time_step <- "day"
season_month_start <- 1
season_month_end <- 12
start_date <- '2003-01-01'
end_date <- '2003-02-13'
use_lethal_temperature <- FALSE
temperature_file <- ""
lethal_temperature <- -30
lethal_temperature_month <- 1
mortality_on <- FALSE
mortality_rate <- 0
mortality_time_lag <- 0
management <- FALSE
treatment_dates <- c('2003-01-24')
treatments_file <- ""
treatment_method <- "ratio"
natural_kernel_type <- "cauchy"
anthropogenic_kernel_type <- "cauchy"
natural_dir <- "N"
natural_kappa <- 0
anthropogenic_dir <- "NONE"
anthropogenic_kappa <- 0
pesticide_duration <- c(0)
pesticide_efficacy <- 1.0
mask <- NULL
success_metric <- "residual error" ### keep this the same
output_frequency <- "year"
movements_file = "" ## ignore - for pigs
use_movements = FALSE
percent_natural_dispersal <- 1.0
anthropogenic_distance_scale <- 0.0
'''

Add spatial index for all suitable cells.

Adding more arguments to the C++ pops_model function causes too many arguments in foreign function call. Currently the function has 62 parameters. Adding 4 (e.g., for the overpopulation movements in #83) causes the error as the limit of 65 is hit. This is according to the documentation (the foreign function here is our pops_model):

arguments to be passed to the foreign function. Up to 65.

The best fix seems to be embrace the concept of config more. For #83, I'm working on a solution along these lines which will be a good prototype for the overall fix. Although it fixes this problem, I'm not sure yet if it works 100%.

The call:

rcmdcheck::rcmdcheck(args = c("--no-manual", "--no-multiarch"), error_on = "warning")

Gives:

N  checking dependencies in R code (3.1s)
   Namespaces in Imports field not imported from:
     ‘igraph’ ‘rgdal’ ‘rlist’
     All declared Imports should be used.
   Missing or unexported object: ‘PoPS::pops_model’

Incorporate aws.s3 read into the configuration function to handle data passed from the model API and case study API—also, export configuration as a function usable in both of those functions.

This will need to be in version 2.0 as it will be a major change to the API. The idea is to create 2 configuration creation functions (one for current parameter layout and one that works with a JSON format) that write out a Rdata file that has the config file ready to run for the function selected (e.g. calibrate, validate, multi-run, or auto-manage). The idea is that it simplifies the addition of new parameters which will be especially relevant with the addition of multiple networks and support for multi-host systems. The setup should save users time and make for more easily reproducible workflows.

Unlike rasterToPoints, terra::as.points does not have fun and spatial parameters. The current code:

  suitable_points <- terra::as.points(suitable,
                                    fun = function(x) {
                                      x > 0
                                    },
                                    spatial = TRUE)

produces:

Error: unused arguments (fun = function(x) {
    x > 0
}, spatial = TRUE)

with devtools::test().

Discovered in #81. Terra introduced in #79.

useful when data on the intensity of infection is available.

Once the website version 1.0.0 launches add links to website for tutorials, FAQ, etc.

Once public-facing database API is available put the link and how to use it in the readme. Additionally, add functionality for pulling directly from the API.

I would like to run the calibration and other functions with .vrt files. Currently limited to ".grd", ".tif", ".img".

Add exposed file so can have starting position of the model have both infected and exposed individuals at the start of the simulation.

sf package is faster and less memory intensive.

Here are two possibly related problems with auto_manage code:

1. R CMD check reports possible problems for the auto_manage function. Here are the results:

checking R code for possible problems ... NOTE
  auto_manage: no visible binding for global variable ‘infected’
  auto_manage: no visible global function definition for ‘stack’
  auto_manage: no visible binding for global variable ‘host’
  auto_manage: no visible global function definition for ‘nlayers’
  auto_manage: no visible binding for global variable ‘num_cells’
  auto_manage: no visible binding for global variable ‘buffer_cells’
  auto_manage: no visible binding for global variable ‘random_seeds’
  auto_manage: no visible binding for global variable ‘exposed’
  auto_manage: no visible binding for global variable ‘total_populations’
  auto_manage: no visible binding for global variable ‘mortality_tracker’
  auto_manage: no visible binding for global variable ‘mortality’
  auto_manage: no visible binding for global variable ‘quarantine_areas’

2. Function is ignoring config or needs some explanation: The auto_manage function takes parameters, adds them to config list, but then config is not used to set parameters in the pops_model call. If this is an intention, it should be documented in a comment. If not, it should be fixed.

Combine abc_calibration function with calibrate function and abc_validate function with validate function. This reduces the amount of code that is duplicated across these and should be easily handled with a selection between the methods of calibration and thus the way they export the parameters and uncertainties.

1: TIFFAppendToStrip:Maximum TIFF file size exceeded. Use BIGTIFF=YES creation option. (GDAL error 1)
2: C:/Users/nfkruska/AppData/Local/Temp/RtmpKO6IQ2/spat_VDJGg7EJVrX8RAq.tif: WriteEncodedTile/Strip() failed. (GDAL error 1)

I'm posting here, although this may be an issue with terra. When the data files are too large, PoPS will fail with the above warnings from R. This is likely the weather data file. It appears when the file is being copied to the TEMP location it is being written without the BIGTIFF flag set to YES which allows a .tif file to exceed 4GB. My current weather file is ~2.88GB, but it is compressed. It may be we need to ensure the .tif profile is adjusted when replicating the data (to YES or IF_SAFER), or this may fall to terra not offering that option. This is not an issue with available RAM.

Chris, this is for the 250m CA data.

Add a function to run the multi-run version of pops multiple times to get calculate sensitivity and specificity for model outputs compared to data. Currently being used in PEDv case study.

1. Add variable ranges or values

• Calibrate
• Validate
• PoPS
• quantity allocation
• pops_model