v1.1.0

ParticleDA v1.1.0

Diff since v1.0.1

Merged pull requests:

• Update GaussianRandomFields version and add additional tests for linear Gaussian models (#244) (@matt-graham)
• fix: Replacing Threads.@threads :static (#247) (@DanGiles)
• Bump to version 1.1.0 (#254) (@matt-graham)
• Remove explicit dependency on SciMLBase (#256) (@giordano)
• Allow HDF.jl v0.17 (#258) (@giordano)
• Reduce LLW2d model mesh dimensions + number of samples for test (#259) (@matt-graham)
• [CI] Add dependabot, use newer workflows, use concurrency groups (#260) (@giordano)
• Bump codecov/codecov-action from 3 to 4 (#261) (@dependabot[bot])
• Bump actions/checkout from 3 to 4 (#262) (@dependabot[bot])
• Ignore result of runs with nightly builds in CI job status (#263) (@matt-graham)
• Add extra links to README and improve description (#264) (@matt-graham)
• Run different CI jobs only when relevant files are touched (#266) (@giordano)
• Add compat bound for ChunkSplitters (#267) (@giordano)

Closed issues:

• Update GaussianRandomFields version in test dependencies (#243)
• Replace the use of Threads.@threads :static (#246)
• Request for code for Lorenz63 plot in paper (#250)
• Adjust number of samples and/or mesh dimension for optimal proposal unit tests with llw2d model (#257)
• Rename default branch to main (#265)

v1.0.1

ParticleDA v1.0.1

Diff since v1.0.0

Merged pull requests:

• Add DOI badge to README.md (#240) (@giordano)
• Update installation instructions (#241) (@giordano)

v1.0.0

ParticleDA v1.0.0

Closed issues:

• Implement particle filter (#2)
• Parameters shouldn't be hard-coded in the module (#4)
• Add noise to the model (#6)
• Add distributed parallelism (#7)
• Create test/Project.toml (#14)
• Initial condition in llw2d is obscure (#15)
• Add noise to the observations (#17)
• Add some documentation (#18)
• Add plotting function(s) (#19)
• Rename the package (#23)
• Change output format (#26)
• Benchmark suite (#34)
• Slow performance in Process Noise (#35)
• Timeseries outputs, ESS (#36)
• Don't get observations when PF is not called (#40)
• Redefine time stepping (#44)
• Redefine grid parameters (#45)
• Add variance to output (#46)
• Possible memory allocations improvements (#53)
• Represent states with structs (#58)
• Add integration test that tests particles (#62)
• Names bikeshedding (#64)
• New Initial State of Particles (#67)
• Error when changing nx, ny (#74)
• Simplify station setup (#75)
• Observations should be independent (#76)
• Plot to show estimated Standard Deviation instead of Variance (#81)
• Extra Plot to check effect of Background Noise (#82)
• Parellelisation (#83)
• Add station coordinates to output (#84)
• Provide station coordinates in an input file (#86)
• Consistency at colouring at plots of "true height" and "assimilated height" (#92)
• jupyter notebook for plotting (#96)
• Parallelise resampling (#97)
• Provide plot for initial condition of (the two) true Velocity Fields (not only of initial Height Field) (#101)
• add_random_field is not thread-safe after #94 (#105)
• Discontinuities in variance and mean (#108)
• Parallel performance (#115)
• Thread Scaling (#117)
• Separate particle filter from the model (#124)
• Parametrize initial condition in llw2d properly (#134)
• Optimal particle filter (#136)
• Different Background Noise Parameters for Each of Three State Variables (#175)
• Plotting tool in Julia (#182)
• Add SPEEDY climate model (#199)
• Precompiling ParticleDA Error (#202)
• Role of weight_std parameter not clear (#205)
• Generalise update_particle_proposal! (#208)
• OptimalFilter BoundsError when running with mpiexecjl (#209)
• Covariance function assumed in optimal proposal not guaranteed to match that used for process noise (#213)
• Fix TODO about covariance in comment (#214)
• Generalising the optimal proposal implementation (#215)
• Allow returning particles as well summary statistics and at multiple time steps (#220)
• Simplify interface for doing reproducible runs when using MPI (#221)
• Optimal proposal filter not using correct weights (#223)
• Observations currently generated within filtering loop (#224)
• Adding test cases for analytically tractable models (#225)
• Inconsistent conventions around spatial grid coordinates in LLW2d model (#229)
• Optimal proposal particle weights are being computed incorrectly (#230)
• Summary statistics computed on every time step when not outputted (#231)

Merged pull requests:

• Particle filter (#1) (@tkoskela)
• Add keyword argument to tdac to enable/disable verbosity (#3) (@giordano)
• Refactor parameters (#5) (@tkoskela)
• Add instructions on how to install the package in the README (#8) (@giordano)
• Fix copyright of the package in the LICENSE file (#9) (@giordano)
• Add GitHub Actions script for CI (#10) (@giordano)
• MPI Parallelisation (#11) (@tkoskela)
• Bug fix, initheight was called on the wrong state array (#12) (@tkoskela)
• First integration test (#13) (@tkoskela)
• Remove now unused submodule (#16) (@giordano)
• Update README.md (#20) (@tkoskela)
• Jupyter notebook for plotting output files (#21) (@tkoskela)
• Fix link in README.md (#22) (@tkoskela)
• Add gaussian random noise (#25) (@tkoskela)
• [skip ci] Do not fail fast in GitHub Actions (#27) (@giordano)
• Run tests with 1 and 2 threads (#28) (@giordano)
• Update GaussianRandomFields and fix lower/upper bounds for deps (#29) (@giordano)
• [skip ci] Add CompatHelper (#30) (@giordano)
• Clean up function argument lists (#31) (@tkoskela)
• Move output into HDF5 file (#32) (@tkoskela)
• CompatHelper: bump compat for "Distributions" to "0.23" (#33) (@github-actions[bot])
• Do not run MPI.Finalize() (#37) (@giordano)
• Add new get_params - function where user can pass dict of parameters (#38) (@tkoskela)
• Generate Gaussian random field with primes=true (#39) (@giordano)
• Do process noise and observations only when DA is done (#41) (@tkoskela)
• Read default parameters using multiple dispatch (#42) (@tkoskela)
• Define a custom type for the random field (#43) (@giordano)
• Update structure of the time stepping loop (#47) (@tkoskela)
• Set total domain size and derive cell size from it (#48) (@tkoskela)
• Add variance to output, collect arrays in structs (#49) (@tkoskela)
• Introduce timers with the @timeit_debug macro (#50) (@tkoskela)
• Fix timer on first run, add print on screen (#51) (@tkoskela)
• Use params.verbose to check verbosity setting (#52) (@giordano)
• Remove dependency on Distributed (#54) (@tkoskela)
• Use multi-dimensional arrays to represent states (#55) (@giordano)
• CompatHelper: add new compat entry for "TimerOutputs" at version "0.5" (#56) (@github-actions[bot])
• Preallocate the buffer for tsunami update (#57) (@giordano)
• Disable FFTW threads in tests (#59) (@giordano)
• Use preallocated buffer in add_random_field (#60) (@tkoskela)
• Use a buffer array to generate the random numbers in gaussian field (#63) (@giordano)
• Move initialisation of the field buffer to init_tdac (#65) (@giordano)
• Fix distribution of random numbers for the Gaussian random field (#66) (@giordano)
• Update particle initial state (#68) (@tkoskela)
• Add colorbars to contourf plots (#69) (@tkoskela)
• Mean and Var need to be calculated for IO after initial state (#70) (@tkoskela)
• Add particle weights to output + plotting (#71) (@tkoskela)
• Add missing argument from list (#72) (@tkoskela)
• Make output of MPI test visible (#73) (@giordano)
• Fix issue of non-default parameters not getting passed to init_gaussian_random_field_generator (#77) (@tkoskela)
• Add test for different grid and time step size (#78) (@tkoskela)
• Update to weight and observation noise calculation (#79) (@tkoskela)
• Modify to plot standard deviation instead of variance (#85) (@tkoskela)
• Update README.md (#87) (@tkoskela)
• Fix x-axis in ESS plot to start from 1 (#88) (@tkoskela)
• Upgrades to stations (#89) (@tkoskela)
• Update README.md (#90) (@tkoskela)
• Add threads to process noise (#91) (@tkoskela)
• Enforce same color scale in true and avg height plots. Fix unit of var (#93) (@tkoskela)
• Define a thread-safe RNG (#94) (@giordano)
• Separate copying of state vectors from resampling. (#95) (@tkoskela)
• Use in-place version of var (#98) (@giordano)
• Get rid of unnecessary flattening and reshaping (#99) (@tkoskela)
• Use allocation-free var everywhere (#100) (@giordano)
• Revert state split (#102) (@tkoskela)
• Add options for particle initial state (#103) (@tkoskela)
• Add tests for particle state (#104) (@tkoskela)
• Write velocity fields to output and plot them (#106) (@tkoskela)
• Move IO functions to separate file (#107) (@tkoskela)
• CompatHelper: bump compat for "MPI" to "0.15" (#109) (@github-actions[bot])
• Move hard-coded parameters from llw2d.jl to parameters.jl and Fix boundary calculation (#110) (@tkoskela)
• Parallel resampling (#111) (@tkoskela)
• Pack all model matrices in a single struct (#112) (@giordano)
• Implementation of custom MPI reduction for mean and variance (#113) (@tkoskela)
• Add scripts to run scaling tests on CSD3 and notebook to plot results (#114) (@tkoskela)
• Rename functions for readability (#119) (@tkoskela)
• Simplify initheight (#120) (@tkoskela)
• Update README.md (#121) (@tkoskela)
• Do not force FFTW threads to 1 (#122) (@giordano)
• Rename TDAC.jl -> ParticleDA.jl (#123) (@giordano)
• Add disclaimer to README (#125) (@tkoskela)
• Run CI on more versions of Julia (#126) (@giordano)
• Add benchmarks (#127) (@tkoskela)
• Separate model from particle filter (#128) (@giordano)
• CompatHelper: bump compat for "Distributions" to "0.24" (#129) (@github-actions[bot])
• Fix settings for deployment of documentation (#130) (@giordano)
• CompatHelper: bump compat for "MPI" to "0.16" (#131) (@github-actions[bot])
• CompatHelper: bump compat for "HDF5" to "0.14" (#132) (@github-actions[bot])
• Implement optimal particle filter from Alex (#133) (@tkoskela)
• Add SSH key to CompatHelper (#135) (@giordano)
• Update README.md (#137) (@tkoskela)
• Refactor llw2d.initheight (#138) (@tkoskela)
• Remove observation noise from model observations (#139) (@tkoskela)
• Add new filter type argument for dispatch (#140) (@giordano)
• CompatHelper: bump compat for "HDF5" to "0.15" (#141) (@github-actions[bot])
• Move filter initialisations into a function and group arrays into a struct. (#142) (@tkoskela)
• Update documentation to mention new argument to run_particle_filter (#143) (@giordano)
• Change initial value of bathymetry, pointed out by Devaraj (#144) (@giordano)
• Update model interface for optimal filter (#145) (@tkoskela)
• Add two new functions to implement in the model (#146) (@giordano)
• Slightly simplify test project file (#147) (@giordano)
• Fix bug in update_reference_data.jl (#148) (@tkoskela)
• Restore compat bounds for test dependencies (#149) (@github-actions[bot])
• Add GitHub Action workflow to run benchmarks on CI for pull requests (#153) (@giordano)
• Add more benchmarks (#154) (@giordano)
• Pass filter type to run_particle_filter and use it to dispatch (#155) (@tkoskela)
• Increase required Julia version to v1.5 (#156) (@giordano)
• Split get_grid_size into three functions (#157) (@tkoskela)
• More generic initialisation of boostrap filter (#158) (@giordano)
• Precompute FFTW plan (#159) (@giordan...