Published September 30, 2025 | Version v4

CLDynamicLandCover

  • 1. Facultad de Ingeniería y Ciencias, Universidad Adolfo Ibañez
  • 2. ROR icon Center for Climate and Resilience Research
  • 3. ROR icon Data Observatory Foundation
  • 4. Doctorado en Ciencias de Recursos Naturales, Universidad de la Frontera
  • 5. Department of Civil Engineering, Faculty of Engineering and Sciences, Universidad de la Frontera

Description

  Description
Dataset CLDynamicLandCover
Resolution 30 m
Area Central Chile (latitudes -31.61° to -43.50°; southern Coquimbo – southern Los Lagos)
Years covered 1990, 1999, 2004, 2009, 2013, 2018
Temporal resolution Five-year intervals
Development Semi-automatic algorithm, initially for south-central Chile, later expanded
Data sources Landsat (spectral reflectivity), SRTM (topography), auxiliary cartographic info
Training data 15 land cover classes, photointerpreted & georeferenced in 2018
Feature selection Machine learning algorithms + expert criteria
Accuracy Overall accuracy: 0.894–0.950; Kappa: 0.877–0.943

Abstract

The CLDynamicLandCover is a 30 m spatial resolution dynamic land use dataset for Central Chile, covering latitudes -31.61 to -43.50. It provides land cover maps on a five-year scale from 1990 to 2018, created by a semi-automatic algorithm initially developed for the coastal sector of south-central Chile. This method was later expanded to cover areas from the southern Coquimbo region to the southern Los Lagos region.

The dataset uses a supervised classification approach with satellite data (Landsat and SRTM) to represent reflectivity and topography, and auxiliary cartographic information. It includes training points for 15 land cover classes, photointerpreted and georeferenced in 2018, to create a spectral signature for each class. Predictive variables are selected using machine learning algorithms and expert criteria.

The 2018 endmember is used to estimate the Jeffries-Matusita Distance for points from other years (1990, 1999, 2004, 2009, 2013), determining whether a pixel's land cover class changed or remained constant. A separability cutoff threshold optimizes class discrimination between years. Points below the threshold are retained (indicating no change), and those above are eliminated (indicating change). These points train a Random Forest classification model to generate land cover maps for each year.

Land trajectories were corrected based on ecological transitions and economic cost criteria, with a new classification for harvested plantations. The product shows high accuracy, with values ranging from 0.894 to 0.950 and Kappa coefficients from 0.877 to 0.943 across the years.

Resumen

El CLDynamicLandCover es un conjunto de datos dinámico de uso del suelo con una resolución espacial de 30 m para Chile Central, que cubre las latitudes -31.61 a -43.50. Proporciona mapas de cobertura de suelo en intervalos de cinco años entre 1990 y 2018, creados mediante un algoritmo semiautomático desarrollado inicialmente para el sector costero del centro-sur de Chile. Posteriormente, este método se amplió para cubrir áreas desde el sur de la región de Coquimbo hasta el sur de la región de Los Lagos.

El conjunto de datos utiliza un enfoque de clasificación supervisada con información satelital (Landsat y SRTM) para representar la reflectancia y la topografía, complementado con información cartográfica auxiliar. Incluye puntos de entrenamiento para 15 clases de cobertura de suelo, fotointerpretados y georreferenciados en 2018, con el fin de generar una firma espectral para cada clase. Las variables predictoras fueron seleccionadas mediante algoritmos de aprendizaje automático y criterios de expertos.

Se usa la Distancia Jeffries-Matusita de 2018 para los puntos de otros años (1990, 1999, 2004, 2009, 2013), determinando si la clase de cobertura de un píxel cambió o se mantuvo constante. Un umbral de separabilidad optimiza la discriminación de clases entre años: los puntos bajo el umbral se conservan (sin cambio) y los puntos sobre el umbral se eliminan (cambio). Estos puntos entrenan un modelo de clasificación Random Forest para generar los mapas de cobertura de cada año.

Las trayectorias de cambio de cobertura fueron corregidas en base a transiciones ecológicas y criterios de costo económico, incorporando además una nueva clasificación para las plantaciones cosechadas. El producto muestra una alta precisión, con valores que van desde 0.894 a 0.950 y coeficientes Kappa entre 0.877 y 0.943 a lo largo de los distintos años.

Notes

  •  Management of global change impacts on hydrological extremes by coupling remote sensing data and an interdisciplinary modeling approach (Chi)2, Award: PCI ANID Chile-NSFC China NSFC190018.
  • The catchment's memory: understanding how hydrological extremes are modulated by antecedent soil moisture conditions in a warmer climate, Award: ANID Fondecyt 1212071.
  • Improving forest water yield and productivity quantification at the catchment scale by mapping root depth and ecophysiological thresholds with remote sensing and water transfer modeling, Award: ANID Fondecyt 1210932.
  • Center for Climate and Resilience Research, Award: CR2, CONICYT/ FONDAP/15110009.
  • You can find: raster .tif files for land cover classification for the years 2018, 2013, 2009, 2004, 1999, and 1990; a technical report describing the main methodological aspects; and a land cover map for 2018.

Technical info

ID Class Name Description RGB Code
1 Water bodies and
channels
Natural and artificial inland water bodies. Lakes, lagoons, waterholes and wide/accumulation sectors of rivers and streams. Includes agricultural and mining reservoirs/ponds. 0, 92, 230
2 Beachs, dunes and
sandbanks
Coastal and riverine beaches and dunes; inland in abandoned and/or transported dunes, sandbanks in clogged rivers and other sandy sectors at altitude. 255, 235, 190
3 Mediterranean sclerophyll Native Forest Mediterranean sclerophyll forest sectors with mature native trees and dense thickets (woody plants greater than two meters in height), or with a coverage of over 80% withr espect to the Landsat grid. 56, 168, 0
4 Temperate Native Forest Temperate forested sectors with mature native trees (woody plants greater than three meters in height), or with a coverage over 90% with respect to Landsat grid. 38, 115, 0
5 Broad leaf plantation Forest plantations of exotic broad leaf species (mainly Eucalyptus sp.) 255, 0, 197
6 Fruit trees Fruit tree plantations (fruit growing). Class generated from training points obtained from the fruit tree cadastre (vineyards, avocado, citrus, grapefruit, etc). 200, 0, 0
7 Glacier and snow Snow-covered surfaces and eternal ice. 255, 255, 255
8 Riparian vegetation Flooded surfaces (permanent and seasonal) covered with vegetation, adjacent to watercourses, bodies of water and water outcrops. 0, 255, 197
9 Shrub and  bush
vegetation
Sectors covered with woody plants less than 3 meters high. Canopy projection covering from 5% to 80% with respect to Landsat grid. 255, 120, 0
10 Pynophyte plantation Forest  plantations  of exotic acriculate and imbricate/scaly/spiral-eaved species (mainly Pinus radiata). 115, 40, 0
11 Grasslands, pastures and annual crops Annual natural (un-managed) grasslands and pastures and annual irrigated and rained (non-fruit) agricultural crops. Herbaceous cover with high seasonal variation in early spring vs. late summer. 255, 255, 0
12 Meadows and evergreen grasslands Managed and unmanaged evergreen grasslands and pastures. More common in southern Chile, associated with animal grazing areas. Herbaceous cover with low seasonal variation in early spring vs. late summer. 50, 230, 0
13 Bare soil and non-vegetated areas Decapitated soils, excavations, landfills, roads and mine stockpiles, rocky outcrops, faults, landslides and rock slides and sectors above the altitudinal limit of vegetation. 199, 215, 158
14 Peatland Soils covered by peatlands. 104, 104, 104
15 Impermeable surface Surfaces covered with impervious human infrastructure: Buildings, paved roads, housing and industrial sectors. 0, 0, 0
16 Harvested plantation Estimated harvested areas of tree plantations 186, 85, 211

Files

Figure_1.png

Files (763.5 MB)

Name Size
md5:9d629d9f7126e2e16fbc492924f7d40e
89.9 MB Preview Download
md5:5439379acf1f770d928adffe892f9e44
130.5 MB Preview Download
md5:76a58a33863a019e194865381297d3c2
131.6 MB Preview Download
md5:7785c2a221907b4447b953773801363a
133.3 MB Preview Download
md5:0eacd4cccf879ebdd817141f41c87344
135.7 MB Preview Download
md5:1f128323f09c618a44067c3b4b8d06d2
139.2 MB Preview Download
md5:cc4ad1b6977e88313c335ebed569cbd5
2.6 MB Preview Download
md5:ada0377429c6c78d417b60cee3178f2e
653.4 kB Preview Download

Additional details

References

  • Zhao, Y., Feng, D., Yu, L., Wang, X., Chen, Y., Bai, Y. & Gong, P. (2016). Detailed dynamic land cover mapping of Chile: Accuracy improvement by integrating multi-temporal data. Remote Sensing of Environment, 183, 170- 185