Public case · open data · Atacama Region, Chile

Bomarea ovallei: when four of its populations become active

The lion's claw is an endemic, endangered geophyte of the coastal Atacama Desert. It emerges and flowers only in years with enough rain. We reconstructed with Sentinel-2 the 2017–2026 series of the vegetation around the four populations sampled by Rozas-Lazcano et al. (Plants, 2025) between Totoral and Llanos de Challe, and compared it with each season's rainfall. It is the population-scale companion of the regional study The Desert That Awakens.

Latest scene: … · Open data: DOI 10.5281/zenodo.23887276 (CC BY 4.0) · Versión en español

Active years
4 of 10
2017, 2022, 2025 and 2026, the same in all four populations. In 2018–2021 and 2023–2024, none.
Separation
0.051 · 0.105
The largest rise of a non-event year against the smallest rise of an event year. Nothing in between: the +0.08 threshold falls in the gap.
Apr–Sep rainfall
≥ 45 mm
In every active year according to ERA5; years without activation had ≤ 28 mm. Coastal and Huasco-valley stations confirm the ordering.
Observations
2,103
Valid Sentinel-2 dates over the four populations, 2017–2026. Full series downloadable with a DOI.

What the data show, and what they do not

What they show

  • The four populations become active in the same years (2017, 2022, 2025 and 2026). The result does not depend on the threshold: it is identical between +0.06 and +0.10.
  • Activation follows rainfall. Every active year had 45 mm or more between April and September (ERA5), and every other year 28 mm or less. That is the 30–50 mm range the literature associates with the species' emergence. At the four coastal and Huasco-valley stations (Freirina, Freirina Nicolasa, Vallenar and Caldera), every active year with data was wetter than every year without activation.
  • 2025 was a moderate active year, with the NDVI peak between 14 and 26 September in all four populations. It matches that year's public reports: CONAF anticipated the event after early-August rain (CONAF, Aug 2025) and the administrator of Llanos de Challe National Park reported flowering B. ovallei plants (El Desconcierto, 4 Nov 2025).
  • 2026 was, with 2017, the most intense event of the series: at QSN it even exceeded it (+0.222 vs +0.188).

What they do not show

  • That NDVI measures B. ovallei. It measures the whole plant community in a 600 m window. An isolated geophyte is not resolved at 10 m: activation shows the site had event conditions, not that the species emerged or flowered.
  • A relationship between activation and genetic diversity. With four populations it cannot be inferred.
  • That more rain means more activation. 2026 had 163 mm according to ERA5 and an activation similar to 2017, with 51 mm: the response saturates.
  • A threshold in millimetres. The ordering of the years is robust; the value is not: Vallenar reached 34 mm in 2020 with no activation and Caldera 16 mm in 2025 with activation. At the two Copiapó-valley stations (La Copa, Copiapó UDA), inland, the coastal active years were not the wettest.

Open

  • In which of the four populations did B. ovallei emerge or flower in 2025? That it flowered in the area is public; in which population is not. That is the field observation that would validate whether the series correctly identifies each population's reproductive years.
  • How much rain does each ravine need? The nearest station is 44 km away. A rain gauge in the ravines would resolve what ERA5 and distant stations cannot.
  • Versions. The dataset accepts new versions on Zenodo: with field observations, a second version would cross the satellite series with each population's actual phenology.

The sites

The four populations of the study by Rozas-Lazcano et al. (2025), with the 600 m window used by the series. Elevations from that work: QCA 39–41 m, QH 83–93, QSN 84–91 and TOTO 153–162 m a.s.l. Locations are shown rounded to 0.001° (about 100 m). B. ovallei is a protected species: collection requires a CONAF permit.

Activation by population, year and 10-day period

Median NDVI in the 600 m window. ▼ = first observation of the season whose 3-observation rolling median reaches 0.20. Grey = no clear scene.

NDVI heatmap by population, 2017–2026
Clouds missed by the mask: the light cells inside the 2026 peak at QSN and QH correspond to the 21 Sep 2026 scene, covered by cloud or fog that the scene mask classified as bare soil. Across the whole series there are 9 such observations (isolated NDVI drops), flagged in the dataset; they include the 25 Aug 2026 pass at QSN, TOTO and QCA. Clouds can only lower NDVI, never raise it, and recomputing without those observations leaves the baseline and all yearly activations unchanged.

Rainfall explains the active years

Bars: accumulated April–September rainfall (ERA5, 0.25° cell nearest to the four populations, centre 28.0° S 71.0° W). Dots: each population's activation, measured as peak NDVI minus the 2019–2021 dry baseline. Figure labels are in Spanish: "Lluvia abr–sep" = Apr–Sep rainfall, "Activación" = activation, "año activo" = active year.

ERA5 rainfall and activation by year
Limit: ERA5 is a reanalysis of about 28 km, and in the coastal desert it underestimates or displaces events. It serves to order the years, not to set a threshold in mm. Station validation is in the next section.

Activation (ΔNDVI) by population and year. In green, active years (≥ +0.08). Hover to see the peak date and the number of observations. *2026: series through the latest available scene; the season is past its peak.

2026First observation with rolling median ≥ 0.20 (previous → first)PeakPeak date

Dates as MM/DD. The raw value of a single observation may be below 0.20: the criterion is the centred 3-observation rolling median.

Validation with surface stations

Rainfall measured by rain gauges of the Chilean Meteorological Directorate network (including INIA and FDF stations published by the DMC), summed from April to September. These are the six stations in the Atacama Region with monthly totals since 2016, between 44 and 86 km from the nearest population. Hatched bars are years with missing months: in an active year that total is a floor. Gold bands mark the active years according to the satellite.

April–September rainfall at six DMC stations by year
Result: at the four coastal and Huasco-valley stations, the driest active year exceeds the wettest year without activation (Vallenar 63 vs 34 mm · Freirina 49 vs 23 · Freirina Nicolasa 79 vs 31 · Caldera 16 vs 11). At the two Copiapó-valley stations, inland to the north-east (La Copa and Copiapó Universidad de Atacama), it does not hold: 2021 and 2024 were wetter there than 2025. Limit: the station nearest to the ravines (Freirina) is 44 km away. The DGA stations at Totoral and Canto de Agua, which would be ideal, only have published data through 2019. Source: climatologia.meteochile.gob.cl (stations 280024, 280010, 280021, 270008, 270021 and 270009).

True colour: what the satellite sees

2 × 2 km Sentinel-2 crops centred on QCA and TOTO, with the same stretch: a dry year (2023), 2025 and 2026.

Sentinel-2 true-colour crops of QCA and TOTO
At QCA and TOTO the greening appears on the slopes in 2025, barely, and is clear in 2026.

Open data

The full series and metrics of this panel are published as an open dataset (CC BY 4.0), with a DOI and a Hugging Face mirror. What is published is the data obtained: series, metrics, rainfall and definitions.

DOI 10.5281/zenodo.23887276   MIRROR Hugging Face

  • serie_ndvi_poblaciones_2017-2026.csv — 2,103 observations: population, date, median NDVI, % vegetated, % clear, Sentinel-2 scene, undetected-cloud flag.
  • metricas_anuales_poblaciones.csv — per population and year: peak, date, activation, active year, first observation with rolling median ≥ 0.20.
  • lluvia_era5_abr-sep_2017-2026.csv and lluvia_estaciones_dmc_abr-sep_2017-2026.csv — seasonal rainfall by year.
  • poblaciones.csv, resumen.json, README.md (ES/EN), CITATION.cff.

How to cite: Copernicus Argentina; Kovacevic, G. R. (2026). Bomarea ovallei (Atacama, Chile) — activación satelital de cuatro poblaciones, 2017–2026: series NDVI Sentinel-2 y lluvia estacional (v1.0) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.23887276

Definitions. Sentinel-2 L2A (EU Copernicus Programme), bands B04 and B08 at 10 m, from January 2017 to the latest scene. Valid date: the least cloudy scene of the day, with at least 90 % of the window with data and 80 % free of cloud, shadow, snow and water according to the scene mask. Median NDVI in a 600 × 600 m window centred on each population. Centred 3-observation rolling median. Activation = April–November peak minus the dry baseline (2019–2021 median). Active year: ≥ +0.08. Rainfall: ERA5 (Copernicus Climate Change Service / ECMWF, via Open-Meteo, CC BY 4.0) and monthly station totals published by the DMC, summed from April to September.

Sources