TSGenerator
'Copernicus' HR-VPP Vegetation Time-Series Processing
Provides tools for acquiring, extracting, quality-assessing, and analysing vegetation time series derived from 'Copernicus' High Resolution Vegetation Phenology and Productivity (HR-VPP) products. Supports Seasonal Trajectories (ST) and Vegetation Phenology and Productivity (VPP) products through native R access to 'WEkEO', geospatial processing, product-quality assessment, temporal completeness analysis, missingness modelling, and integrated graphical workflows. Selected functions from 'TSGenerator' 1.x are retained to support migration and legacy workflows.
TSGenerator
TSGenerator 2.0 is an R package for acquiring, extracting, quality-assessing, and analysing Copernicus HR-VPP Seasonal Trajectories (ST) and Vegetation Phenology and Productivity (VPP) products.
The 2.0 architecture is built around three explicit components:
- Acquisition: native R access to WEkEO HDA through
hdar. - Geospatial processing:
terra-based extraction and product-aware quality handling. - Time-series analysis: temporal completeness, optional imputation, and missingness diagnostics.
Python/reticulate, raster, the discontinued VI workflow, and the old QFLAG2 preprocessing chain are not part of the 2.0 core.
Unified Shiny application
TSGenerator 2.0 includes a single integrated graphical workflow for users who prefer not to program directly:
library(TSGenerator)
runTSapp()
The application follows Acquire → Prepare/Extract → Quality → Analyze → Export and calls the same public R API used in scripts. Results & Export records available session outputs, acquisition provenance, reproducible R code, and session information. Selected historical mini-apps remain temporarily callable for compatibility, while discontinued VI/QFLAG2 interfaces are blocked.
Installation
install.packages("devtools")
devtools::install_github("avalero92/TSGenerator")
A WEkEO account is required only for online ST/VPP searches and downloads. These operations also require an active Internet connection and depend on the availability of the external WEkEO service, which is outside the control of TSGenerator. Credentials can be supplied to hda_client() or managed by hdar through ~/.hdarc. Offline processing, local diagnostics, and package checks do not require WEkEO access.
Quick start
1. Check WEkEO
library(TSGenerator)
check_wekeo()
# For a live authenticated diagnostic:
# check_wekeo(online = TRUE)
2. Preview or download ST
st_search <- download_st(
start = "2020-04-01",
end = "2020-04-30",
tile_id = "30TXM",
product = c("PPI", "QFLAG"),
download = FALSE
)
# Set download = TRUE and output_dir = "HRVPP_ST" to download.
download_st() uses the current HR-VPP ST dataset and supports PPI and ST-QFLAG. Long requests are divided into service-compatible temporal windows.
3. Preview or download VPP
vpp_search <- download_vpp(
start = "2020-01-01",
end = "2020-12-31",
tile_id = "30TXM",
product = c("SOSD", "MAXD", "EOSD", "LENGTH"),
season = "s1",
download = FALSE
)
VPP metadata and scaling are handled by product. Date parameters such as SOSD/MAXD/EOSD are not treated like spectral or productivity values.
Geospatial extraction
ST time series
st_ts <- extract_ts(
x = "HRVPP_ST/PPI",
polygons = parcels,
id_col = "Parcel_ID",
fun = "median"
)
The result is a tidy tsg_time_series with ID, Date, Layer, and Value. extract_ts() validates polygon IDs and CRS and preserves the source raster grid.
VPP parameters
vpp <- extract_vpp(
x = "HRVPP_VPP",
polygons = parcels,
id_col = "Parcel_ID",
fun = "median"
)
The result identifies Year, Season, Product, Unit, and source Layer. Date products populate Date; numeric products populate Value.
Product quality
Copernicus product quality is deliberately separate from temporal completeness.
quality_info("ST")
quality_info("VPP")
q_summary <- summarize_quality(qflag, type = "ST")
st_masked <- mask_quality(
x = st_raster,
quality = qflag_raster,
type = "ST",
min_quality = 4
)
Quality flags are categorical. TSGenerator does not bilinearly resample QFLAG rasters.
Time-series completeness
plan <- temporal_plan(
st_ts,
expected_step = 10
)
missing <- summarize_missingness(plan)
quality <- assess_ts_quality(plan)
summarize_missingness() distinguishes absent expected dates from explicit NA values. assess_ts_quality() classifies completeness; its classes are TSGenerator diagnostics and must not be confused with Copernicus QFLAG classes.
Optional imputation
completed <- impute_ts(
user_series,
method = "linear",
series_type = "user",
complete_grid = TRUE,
expected_step = 10,
max_gap = 2
)
ST is already temporally processed, so additional ST imputation is blocked by default. It requires the explicit allow_processed = TRUE override when scientifically justified.
Missingness modelling
fit <- model_missingness(
missingness_data,
year_col = "Year",
doy_col = "DOY",
value_col = "Value"
)
plot_missingness(fit)
The 2.0 API separates the GAM model from plotting. Year is categorical by default and DOY can use a cyclic smooth.
Main 2.0 API
| Module | Functions |
|---|---|
| WEkEO | hda_client(), check_wekeo(), check_wekeo_integration() |
| Acquisition | download_st(), download_vpp() |
| Geospatial | geospatial_plan(), extract_ts(), extract_vpp() |
| Product quality | quality_info(), classify_quality(), summarize_quality(), mask_quality() |
| Temporal quality | temporal_plan(), summarize_missingness(), assess_ts_quality() |
| Imputation | impute_ts() |
| Diagnostics | model_missingness(), plot_missingness() |
| Apps | runTSapp() |
Migrating from TSGenerator 1.x
The principal replacements are:
| TSGenerator 1.x | TSGenerator 2.0 |
|---|---|
Download.STPPI() |
download_st() |
Download.HRVPP() |
download_vpp() |
get.Series.mean() |
extract_ts(..., fun = "mean") |
get.Series.median() |
extract_ts(..., fun = "median") |
get.Series.VPP() |
extract_vpp() |
QFLAG2.Mask() |
mask_quality() for current ST/VPP QFLAG |
quality.Series() |
assess_ts_quality() |
general.Quality() |
assess_ts_quality() |
count_missing() |
summarize_missingness() |
TsImpute() |
impute_ts() |
GAM.missing() / GAM.plot() |
model_missingness() / plot_missingness() |
The old VI/QFLAG2 workflow (Download.VI(), get.Stack(), get.Clean.IV(), renames.image.IV()) is legacy-only because it targets discontinued products/workflows. Historical source is retained under legacy/ for traceability, not as the recommended 2.0 workflow.
See vignette("migration-1x-to-2x", package = "TSGenerator") after installation for the detailed migration guide.
Validation before release
The development branch includes offline tests and an opt-in live WEkEO integration harness. Maintainers should run:
devtools::document()
testthat::test_local()
devtools::check(document = FALSE, manual = FALSE)
check_wekeo(online = TRUE)
z <- check_wekeo_integration(download = FALSE)
stopifnot(z$ok)
A controlled real-download smoke test should also be completed before a release candidate.
Vignettes
vignette("getting-started", package = "TSGenerator")— architecture and core workflow.vignette("st-vpp-workflows", package = "TSGenerator")— acquisition-to-analysis examples.vignette("migration-1x-to-2x", package = "TSGenerator")— detailed 1.x → 2.0 migration.
Authors and affiliations
-
Alexey Valero-Jorge — creator and maintainer. ORCID: 0000-0002-5993-7346.
Departamento de Sistemas Agrícolas, Forestales y Medio Ambiente (Unidad asociada a EEAD-CSIC Suelos y Riegos), Centro de Investigación y Tecnología Agroalimentaria de Aragón (CITA), Avda. Montañana 930, 50059 Zaragoza, Spain. -
Mª Auxiliadora Casterad Seral — researcher. ORCID: 0000-0003-4458-6966.
Departamento de Sistemas Agrícolas, Forestales y Medio Ambiente (Unidad asociada a EEAD-CSIC Suelos y Riegos), Centro de Investigación y Tecnología Agroalimentaria de Aragón (CITA), Avda. Montañana 930, 50059 Zaragoza, Spain. -
José-Tomás Alcalá Nalvaiz — researcher. ORCID: 0000-0001-7549-8825.
Departamento de Métodos Estadísticos, Instituto Universitario de Investigación en Matemáticas y Aplicaciones (IUMA), Universidad de Zaragoza, Zaragoza, Spain.
Contact
Alexey Valero-Jorge — [email protected]
Citation and DOI
Please cite the specific TSGenerator 2.0.0 software release when using this version in research:
Valero-Jorge, A., Casterad Seral, M. A., & Alcalá Nalvaiz, J.-T. (2026). TSGenerator 2.0.0: Tools for Copernicus HR-VPP Vegetation Time-Series Processing (Version 2.0.0) [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.23265767
The DOI above identifies the archived 2.0.0 release. The scientific publications below are separate research outputs and should be cited additionally when relevant.
Scientific publications
TSGenerator has been presented and applied in scientific research involving Copernicus HR-VPP products and vegetation time-series processing.
TSGenerator software publication
Valero-Jorge, A., Casterad, M. A., & Alcalá, J.-T. (2026). TSGenerator: librería R de código abierto para el procesado integrado de productos fenológicos HR-VPP de Copernicus. Congresos UEx, Actas de Congresos, 2. https://doi.org/10.17398/3101-7177.2.123
Research using TSGenerator
Valero-Jorge, A., Casterad, M. A., & Alcalá, J.-T. (2025). Evaluating the Influence of Missing Data from the Crop Vegetation Index Time Series on Copernicus HR-VPP Phenological Products. Engineering Proceedings, 94(1), 4. https://doi.org/10.3390/engproc2025094004
Funding
TSGenerator was created within the LAIKcA I+D+i project PID2021-124029OR-I00, funded by MICIU/AEI/10.13039/501100011033 and FEDER/EU. Alexey Valero Jorge acknowledges grant PRE2022-102328 funded by MICIU/AEI/10.13039/501100011033 and FSE+.
License
MIT. See LICENSE.