All notable changes to this project should be documented in this file.
The Unreleased section should become the release once the release is ready and the text can be used as part of the release description.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning with an additional rule that interfaces to the core library follow the same numbering although they are in separate repositories, so increasing version number there increases version here although there is no specific tag or release for that in this repository.
Final step as an option for output frequency in all functions (@petrasovaa, #163).
Added ability to use mean and sd for temp and precip coefficient. Uses 3 new parameters: weather_type, temperature_coefficient_sd_file, and precipitation_coefficient_sd_file. This is added to help understand uncertainty in model predictions due to weather drivers (@ChrisJones687, #168).
Added the ability to use different seeds for all processes within pops-core. Adds 2 new parameters: multiple_random_seeds (boolean for using this functionality) and random_seeds (set to NULL to allow internal model selection of kernels or pass a CSV with the number of rows being the number of model runs and columns being the kernels in order (@ChrisJones687, #168).
Added ability for model to simulate pathogen/pest survival in soil and then emergence/rain splash in an area (@ChrisJones687, #178).
Model can now write out forecast ensemble members (@ChrisJones687, #179)
Added ability for model to use multi-host pops-core API. The model now takes 2 new parameter tables for hosts. It also takes a list of rasters for each host allowing testing of hypothesis related to different host combinations being more conducive to spread on the landscape (@petrasovaa, @ChrisJones687, @wenzeslaus, #186 and #188).
Added the ability to calibrate, validate, and simulate from county level infection data (@ChrisJones687, #196).
calibrate can now use the Mathews Correlation Coefficient as the summary statistic of to keep or reject parameter sets (@ChrisJones687, #145).
validate, calibrate, pops_multirun, and pops now take use_survival_rates, survival_rate_month, survival_rate_day, and survival_rates_file (@ChrisJones687, #147).
validate, calibrate, pops_multirun, and pops now take network_movement as a parameter. This parameter controls how dispersal occurs along the network (@ChrisJones687, #147).
validate, calibrate, pops_multirun, auto_manage and pops no longer take network_min_distance and network_max_distance as these parameters are now passed in through the parameter_means and parameter_cov_matrix parameters and are calibrated as part of the calibration if network kernel is selected (@ChrisJones687, #140).
calibrate now calibrates the network_min_distance and network_max_distance parameters during calibration and they are now part of the parameter_means and parameter_cov_matrix that are exported from the calibration (@ChrisJones687, #140).
calibrate has more flexible success metric options removes use_distance, use_rmse, and use_mcc parameters and replaces it with the more flexible success_metrics parameter. Users can now select multiple combinations of different success metrics for the calibration (@ChrisJones687, #150).
pops_multirun removed the ability to write all simulations. (@ChrisJones687, #144).
the deterministic parameter has been renamed to dispersal_stochasticity. This was done to be more consistent with generate_stochasticity, movement_stochasticity, and establishment_stochasticity parameters (@ChrisJones687, #147).
validate, calibrate, pops_multirun, and pops now propogate uncertainty from host and initial conditions. This adds the parameters use_initial_condition_uncertainty and use_host_uncertainty. If use_initial_condition_uncertainty is TRUE the infected_file and/or exposed_file need to have 2 layers a mean and standard deviation. If use_host_uncertainty is is TRUE the host_file needs to have 2 layers a mean and standard deviation. For each model run a host and/or initial conditions are drawn from the mean and sd layers so that each run has a unique host and/or initial conditions this will allow for both propogation of uncertainty from these sources but also partitioning (@ChrisJones687, #151).
quantity_allocation_disagreement now takes in the use_distance parameter which is FALSE by default. This allows the model to commute the minimum total distance between observed and simulated infestations (@ChrisJones687, #130).
validate, calibrate, pops_multirun, auto_manage and pops now take in network_min_distance, network_max_distance, and network_filename parameters these are used when the anthropogenic_kernel_type = “network”. This allows directed spread along a network such as a railroad (@ChrisJones687 and @wenzeslaus, #131)
validate now takes the variable point_file and uses it to calculate statistics based on the point_file in addition to the raster file also calculates new measures of model performance accuracy, precision, recall, and specificity (@ChrisJones687, #124).
calibrate no longer uses success_metrics and checks parameters as these are both handled internally by auto updating if the values are too far of in the first generation. This makes for a simpler and faster user experience (@ChrisJones687, #130).
validate no longer uses success_metrics parameter but now has the added parameters use_rmse and use_distance that are used in quantity_allocation_disagreement. This allows for a more intuitive user interface (@ChrisJones687, #130).
quantity_allocation_disagreement variable configuration changed to use_configuration to be more consistent with variable names. (@ChrisJones687, #130).
pops_multirun can now use variable write_output = “all_simulations” to write out susceptible, exposed, and infected rasters from all simulations. This will allow for a future update where simulations can easily be started from previous simulation outputs. (@ChrisJones687, #133)
New dispersal kernels added: Uniform, Power-law, Deterministic neighbor, Hyperbolic-Secant, Gamma, Weibull, Normal, and Logistic (@ChrisJones687, #73).
natural_dispersal_kernel, anthropogenic_dispersal_kernel).pops, pops_multirun, auto_manage, calibrate, and validate).Overpopulation module added to pops-core (@wenzeslaus, #83).
Spatial Index to increase computational speed (@ChrisJones687, #67)
Exposed and Infected populations can both be present at the start of a simulation (@ChrisJones687, #92).
Internal functions for data handling have switched from the raster package to the terra package (@ChrisJones687, #79).
Mask parameter now used in pops-multirun for post processing data for visualization (@ChrisJones687, #104).
Calibration function now takes verbose parameter (@nfkruska, #99).
All functions now support vrt data types (@ChrisJones687, #97).
Raster files can now be read from S3 buckets (@ChrisJones687, #75).
Outputs can now be saved with write_outputs and output_folder_path parameters (@ChrisJones687, #111).
Host map now used as initial mask for post processing and validation calculations (@ChrisJones687, #112).
Movements now moves exposed, resistant, and mortality tracked populations (@ChrisJones687, #118).
Mortality can now occur at various timesteps not just yearly (@ChrisJones687, #118).
mortality_frequency and mortality_frequency_n.Validation now exports the statistics for each output and the cumulative statistics for each year (@ChrisJones687, #121).
Treatments now update total_hosts (@ChrisJones687, #122)
Version 1.0.0 of the PoPS Core C++ library and its interfaces: rpops R package and r.pops.spread GRASS GIS module. The release of rpops includes:
Susceptible-infected (SI) and susceptible-exposed-infected (SEI) host phases (model_type, latency_period).
Host mortality tracking (mortality_rate, mortality).
Host removal and pesticide application treatments (treatments, treatment_date, pesticide_duration).
Host resistance based on pesticide application treatments (pesticide_duration > 0).
Treatments applied only to a ratio of hosts (treatment_application).
Yearly pest removal based on lethal temperature (lethal_temperature, lethal_month).
Two different dispersal kernels (natural_dispersal_kernel, anthropogenic_dispersal_kernel).
Cauchy and exponential radial dispersal kernels use Von Mises distribution.
Seasonal spread (seasonality in months).
Multiple stochastic runs (pops_multirun).
Parallel execution of multiple runs (number_of_cores in pops_multirun).
Output of average and standard deviation of infected hosts across multiple runs and for a single stochastic run (simulation_mean, single_run, simulation_sd).
Average and standard deviations for output averages (number_infecteds, infected_areas).
Infection probability output in percent (probability).
Spread rate measurement in 4 cardinal directions (west_rate, east_rate, south_rate, north_rate ).
Distance to quarantine in 4 cardinal directions (north_distance_to_quarantine, south_distance_to_quarantine, east_distance_to_quarantine, west_distance_to_quarantine ).
Probability of quarantine escape (escape_probability).