Changes in version 2.0.0 (2026-10-21)                  

TSGenerator 2.0 is a major release focused on current Copernicus HR-VPP
Seasonal Trajectories (ST) and Vegetation Phenology and Productivity
(VPP) workflows. The package has been reorganized around native R
acquisition, terra-based geospatial processing, explicit product-quality
handling, and reproducible time-series analysis.

Acquisition

  - Added native R access to WEkEO HDA through hdar; Python and
    reticulate are no longer required for ST/VPP acquisition.
  - Added download_st() for HR-VPP Seasonal Trajectories, including PPI
    and ST-QFLAG selection, tile/bounding-box filters, preview mode,
    structured results, and automatic splitting of long requests into
    service-compatible temporal windows.
  - Added download_vpp() with product-aware VPP selection, season
    handling, spatial/temporal filters, preview mode, and structured
    results.
  - Added check_wekeo() and check_wekeo_integration() for offline
    diagnostics and opt-in live integration checks.
  - Added bounded retry logic, transfer-size estimates, and
    skip-existing behaviour for more robust acquisition.
  - Retained Download.STPPI() and Download.HRVPP() as deprecated
    compatibility wrappers over the 2.0 acquisition API.

Geospatial processing

  - Migrated the active geospatial core from raster to terra.
  - Added geospatial_plan() for explicit validation of raster/vector
    inputs, CRS, polygon identifiers, and scaling choices.
  - Added extract_ts() as the unified polygon time-series extraction API
    with explicit feature IDs and mean/median/min/max/sum aggregation.
  - Added product-aware extract_vpp(), including VPP parameter
    interpretation, season/year provenance, official scaling/NoData
    handling, and date decoding before spatial aggregation.
  - Fixed extraction from real HR-VPP raster stacks whose layers share
    identical internal band descriptions by selecting extracted values
    positionally while preserving temporal attribution from filenames.
  - Original CLMS filenames are preserved for provenance; legacy VPP
    file renaming is no longer part of the recommended workflow.

Product quality

  - Added quality_info(), classify_quality(), summarize_quality(), and
    mask_quality() for current ST/VPP QFLAG semantics.
  - Product quality is explicitly separated from temporal completeness
    diagnostics.
  - Categorical QFLAG rasters are never bilinearly resampled
    automatically.
  - mask_quality() supports explicit quality thresholds/codes and
    requires compatible target/QFLAG raster geometry.

Time-series quality and analysis

  - Added temporal_plan() to define expected temporal coverage
    explicitly.
  - Added summarize_missingness() to distinguish absent expected dates,
    explicit NA values, and usable observations.
  - Added assess_ts_quality() for completeness-based whole-series or
    windowed assessment. It is the primary 2.0 replacement for the
    legacy quality.Series() and general.Quality() quality-classification
    workflows.
  - Added impute_ts() with explicit linear/Kalman methods, imputation
    provenance, optional expected-grid completion, long-gap protection,
    and safeguards against routine re-imputation of already processed ST
    series.
  - Added model_missingness() and plot_missingness() for GAM-based
    temporal missingness diagnostics, separating model fitting from
    plotting.
  - Retained selected 1.x functions as deprecated wrappers or migration
    guards where a safe transition path is available.

Unified graphical interface

  - Added the integrated TSGenerator2 Shiny application and made it the
    default interface launched by runTSapp().
  - The application follows the workflow Acquire → Prepare/Extract →
    Quality → Analyze → Export.
  - Acquisition, spatial extraction, product quality, temporal analysis,
    imputation, and export modules call the same public 2.0 R API used
    in scripts; scientific algorithms are not duplicated in the Shiny
    server.
  - Results & Export can provide available session outputs, acquisition
    provenance, reproducible R code, session information, and analysis
    bundles.
  - The ST and VPP acquisition mini-apps remain temporarily available
    for compatibility. Discontinued VI/QFLAG2 and standalone GAM
    interfaces provide migration guidance instead of silently running
    obsolete workflows.

Legacy and dependency changes

  - The discontinued VI/QFLAG2 workflow is no longer part of the
    active 2.0 core. Historical source remains under the build-excluded
    legacy/ directory for traceability.
  - Removed active dependencies on reticulate, raster, plotly, VIM, and
    doParallel.
  - Reduced the installed surface to current ST/VPP workflows, the
    unified Shiny application, and explicitly retained compatibility
    interfaces.
  - Legacy functions tied to discontinued products now act as migration
    guards where appropriate rather than executing obsolete processing
    chains.

Documentation, testing, and release engineering

  - Added CRAN-safe vignettes for getting started, ST/VPP workflows, and
    migration from 1.x to 2.0. Network-dependent examples are not
    executed during vignette builds.
  - Added offline tests for acquisition validation, geospatial
    extraction, quality handling, temporal completeness, imputation,
    missingness modelling, installed API availability, Shiny
    distribution, and package engineering.
  - Added regression coverage for malformed dates, explicit
    non-contiguous QFLAG codes, duplicate HR-VPP raster band
    descriptions, and release metadata.
  - Synchronized migration documentation so assess_ts_quality() is the
    documented primary replacement for general.Quality();
    summarize_missingness() remains the complementary diagnostic
    summary.
  - Prepared package metadata, documentation, build exclusions, and
    installed assets for the 2.0.0 CRAN release.