Disturbances in vegetation detected with BFAST in the Purapel fluvial catchment

This dataset contains the results (69 TIFF files) of seasonal disturbances detected in vegetation in the Purapel catchment (southern Chile) for the period from 2002 to 2019. These disturbances were obtained by applying the Breaks for Additive Season and Trend (BFAST, Verbesselt et al., 2010) algorithm to 745 Landsat 5, 7 and 8 imagery. We used Collection 2 Level 2 surface reflectance products and applied the CFMask algorithm (Foga et al., 2017) for cloud masking before utilizing the BFAST algorithm. The BFAST algorithm detects changes in the NDVI time series of each pixel. To determine which event was considered a disturbance, we used the same intensity thresholds as in Cabezas and Fassnacht (2018). We then filtered the results to keep just the disturbances with areas greater than 1 hectare, eliminating noisy data. Except for 2 big wildfires (2015 and 2017) it was assumed that all of the disturbances were clear cuts, since forestry is the main productive activity in the region. This was confirmed by validating the data with 35 manually drawn polygons that were randomly distributed across the catchment. Then, we performed an accuracy assessment, obtaining a confusion matrix with a balanced accuracy of 0,86 and a F1 score of 0,69. Each TIFF file is a binary grid with “zeros” representing no disturbance and “ones” representing a disturbance in the season that the name of the file indicates. A GIF file is also included, which contains the time series of the disturbances for easier graphical purposes. References J. Cabezas and F. E. Fassnacht. Reconstructing the Vegetation Disturbance History of a Biodiversity Hotspot in Central Chile Using Landsat, Bfast and Landtrendr. In IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, pages 7636–7639. IEEE, 7 2018. ISBN 978-1-5386-7150-4. doi: 10.1109/IGARSS.2018.8518863. S. Foga, P. L. Scaramuzza, S. Guo, Z. Zhu, R. D. Dilley, T. Beckmann, G. L. Schmidt, J. L. Dwyer, M. Joseph Hughes, and B. Laue. Cloud detection algorithm comparison and validation for operational Landsat data products. Remote Sensing of Environment, 194:379–390, 6 2017. ISSN 00344257. doi: 10.1016/j.rse.2017.03.026. J. Verbesselt, R. Hyndman, G. Newnham, and D. Culvenor. Detecting trend and seasonal changes in satellite image time series. Remote Sensing of Environment, 114(1):106–115, 1 2010. ISSN 00344257. doi: 10.1016/ j.rse.2009.08.014. URL http://linkinghub.elsevier.com/retrieve/pii/S003442570900265X.本数据集包含2002年至2019年间智利南部普拉佩尔流域（Purapel catchment）内植被季节性扰动的检测结果，共计69个TIFF文件。 上述扰动结果通过对745景Landsat 5、7和8影像应用加法季节与趋势断点（Breaks for Additive Season and Trend, BFAST）算法（Verbesselt等，2010）得到。研究使用了Collection 2 Level 2地表反射率产品，并在调用BFAST算法前，借助CFMask算法（Foga等，2017）完成云掩膜预处理。 BFAST算法用于检测每个像素的归一化植被指数（Normalized Difference Vegetation Index, NDVI）时间序列变化。为明确符合标准的扰动事件判定规则，研究采用了与Cabezas和Fassnacht（2018）一致的强度阈值。随后对检测结果进行滤波处理，仅保留面积大于1公顷的扰动斑块，以剔除噪声干扰数据。 除2场大型野火（2015年与2017年）外，研究假设其余所有扰动均为皆伐作业，原因是林业为该区域的主导生产活动。该假设通过在流域内随机布设的35个手动绘制多边形开展数据验证得到确认。随后研究完成了精度评估，得到的混淆矩阵平衡精度为0.86，F1值为0.69。 每个TIFF文件均为二值栅格，其中"0"代表无扰动发生，"1"代表该文件名称所对应季节内出现的植被扰动。数据集同时附带一个GIF文件，用于可视化展示扰动时间序列，便于直观呈现相关结果。 参考文献： 1. J. Cabezas、F. E. Fassnacht. 利用Landsat、BFAST与Landtrendr重建智利中部生物多样性热点地区的植被扰动历史[C]//2018 IEEE国际地球科学与遥感研讨会（IGARSS 2018）论文集, 第7636–7639页. IEEE, 2018. ISBN 978-1-5386-7150-4. DOI: 10.1109/IGARSS.2018.8518863. 2. S. Foga, P. L. Scaramuzza, S. Guo, Z. Zhu, R. D. Dilley, T. Beckmann, G. L. Schmidt, J. L. Dwyer, M. Joseph Hughes, B. Laue. 业务化Landsat数据产品的云检测算法对比与验证[J]. 环境遥感, 2017, 194: 379–390. ISSN 00344257. DOI: 10.1016/j.rse.2017.03.026. 3. J. Verbesselt, R. Hyndman, G. Newnham, D. Culvenor. 卫星影像时间序列的趋势与季节变化检测[J]. 环境遥感, 2010, 114(1): 106–115. ISSN 00344257. DOI: 10.1016/j.rse.2009.08.014. 链接：http://linkinghub.elsevier.com/retrieve/pii/S003442570900265X.