Use of Sentinel 2 imagery to estimate vegetation height in fragments of Atlantic Forest |
| |
| Institution: | 1. Federal University of São Carlos - Graduate Program in Environmental Sciences (UFSCar - PPGCAm), Hwy. Washington Luiz, 235, São Carlos, SP, Brazil;2. Federal University of São Carlos - Environmental Sciences Department (UFSCar - DCAm), Hwy. Washington Luiz, 235, São Carlos, SP, Brazil;3. University of Sao Paulo - Faculty of Philosophy Sciences and Letters of Ribeirao Preto - Department of Biology (USP - FFCLRP - DB), Ave. Bandeirantes, 3900, Ribeirão Preto, SP, Brazil;4. National Institute of Science and Technology in Interdisciplinary and Transdisciplinary Studies in Ecology and Evolution (IN-TREE), Brazil;1. Department of Renewable Resources, University of Alberta, Edmonton, AB T6G 2H1, Canada;2. Department of Computer Science and Engineering, IIT Hyderabad, Sangareddy, Telangana 502285, India;3. Department of Mechanical Engineering, University of Alberta, Edmonton, AB T6G 1H9, Canada;1. Instituto Iberoamericano de Desarrollo Sostenible (IIDS), Unidad de Cambio Climático y Medio Ambiente (UCCMA), Universidad Autónoma de Chile, Temuco, Chile;2. Département des Sciences Géomatiques, Université Laval, Québec G1K 7P4, Canada;3. Departamento de Ciencias Biológicas y Químicas, Facultad de Recursos Naturales, Universidad Católica de Temuco, Casilla 15-D, Temuco, Chile;4. Núcleo de Estudios Ambientales, Universidad Católica de Temuco, Casilla 15-D, Temuco, Chile;5. Grupo de Investigación Botánica Sistemática y Aplicada, Campus de Excelencia Internacional Agroalimentario (ceiA3), Universidad de Córdoba, Spain;6. Proyecto Aysén, Forestal Mininco SpA, Coyhaique, Chile |
| |
| Abstract: | Information on vegetation height can be used in a variety of applications, but the high cost to obtain it in large areas using field sampling and the latest remote sensing technologies is still a barrier for low-income countries and organizations. In an attempt to overcome these limitations, we explored the possibility to estimate vegetation height in fragments of Atlantic Forest (São Paulo - Brazil) based on Sentinel 2 imagery, using LiDAR (Light Detection And Ranging) and field data as reference. The initial results showed that only wet season images appear to be related to the vegetation height, especially band 5 (red-edge) and related vegetation indices (VIs). Predictions made with Sentinel 2 and LiDAR data showed that vegetation height can be estimated with a root mean square error (RMSE) close to 3 m, with simple linear models outperforming random forest algorithms. It's also shown in a variety of validation tests, that although better results are obtained if the models are applied to the same images they were trained in, they are still able to reasonably predict vegetation height when applied to other images and locations if the right predictive variables are used. The results agree with recent studies made in other biomes and show that Sentinel 2 imagery can be used to estimate vegetation height in the Atlantic Forest as well. We conclude that vegetation height estimates with linear models can be used as a simple low cost alternative for future applications in this environment. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
