Evidences of drought-induced forest decline and tree mortality in the hygrophilous forests of central Chile

Abstract

This dataset compiles field measurements, high‑resolution UAV imagery products, satellite‑derived vegetation indices, and precipitation records to characterize vegetation condition, canopy structure, and long‑term productivity patterns across forested sites in central Chile. The first component consists of phytosociological field data collected over a total sampled area of 5,375 m² across seven protected areas. These records include species identity, diameter at breast height (DBH ≥ 5 cm), and canopy cover for all measured individuals. Tree vitality was classified into three categories, Healthy, Stressed, and Highly Stressed/Dead, based on the proportion of live canopy cover, following a simplified version of the criteria proposed by Dobbertin (2005). These field measurements provide ground‑based information on forest structure and condition. The second component includes multispectral UAV imagery products acquired during the 2022–2023 summer season over two forest sites. The flights produced 8 cm spatial resolution orthomosaics with six spectral bands (Blue, Green, Red, Red Edge, Near‑Infrared, and NDVI). Using supervised classification (maximum likelihood algorithm), five land‑cover classes were mapped: alive canopy, leafless canopy, naked soil, shade, and additional background categories. The classification was trained using manually interpreted polygons and evaluated with a confusion matrix. The resulting land‑cover products quantify defoliated and non‑defoliated canopy areas at fine spatial resolution. The third component contains satellite‑based vegetation productivity products derived from the Harmonized Landsat–Sentinel (HLS) collection. These data include monthly NDVI composites at 30‑m resolution from 2013 onward, produced from atmospherically corrected and cloud‑masked reflectance images. Each NDVI time series was converted into standardized anomaly products based on long‑term means and standard deviations, enabling comparisons of vegetation condition across years. Complementing these, the dataset also includes MODIS NDVI products at 250‑m resolution, covering the period 2002–2025. Monthly NDVI values were used to compute seasonal cumulative NDVI (cNDVI) for each growing season, and these values were standardized relative to a 2002–2007 baseline. The resulting z‑standardized cNDVI (zcNDVI) products provide long‑term regional indicators of vegetation productivity and include extracted time series for the sites where mortality was observed. Finally, the dataset provides precipitation records from weather stations located within 15 km of the affected forest sites. Annual precipitation data from 2002-2024 were transformed into standardized anomalies using the same 2002-2007 reference period. Only stations with complete observations for that baseline were included, resulting in a final set of 18 stations. Supplementary tables list all selected stations and the total number of reporting stations per year.