Multi-Temporal Abrupt Change Estimation on Landsat Time Series Imagery: Determining Burned Areas in La Primavera, Mexico

• Activity Number: 274 - Statistical Analysis of Large Time Series of Remotely Sensed Environmental Measurements: Tales of the Collision Between Statistics, Remote Sensing and Geo-Computation
• Type: Topic Contributed
• Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Inder Tecuapetla* and Gabriela Villamil-Cortez and María Isabel Cruz-López
• Companies: CONABIO and National Commission for the Knowledge and Use of Biodiversity and National Commission for the Knowledge and Use of Biodiversity
• Keywords: abrupt change estimation; Landsat-7 images; mapping burned area; missing values; spectral indices; time series

Abstract

La Primavera is a Flora and Fauna protected area and in the last 20 years has undergone to a series of wildfires. Due to geographic and economic reasons a complete estimation of the areas affected by these wildfires is lacking. It is possible to overcome this problem by producing reliable burned area maps based on satellite products. More precisely, we estimate statistically abrupt changes in time series of spectral indices (vegetation, burn ratio) derived from satellite images. We analyze images from the open access Landsat-7 collection which provides both the best spatial (30m) and temporal (2003-2016) resolutions for monitoring La Primavera. Because of a sensor failure these images, however, present at least 20% missing values. Through simulations we calibrate (against missing values) the parameters of the applied abrupt change estimation method. In addition of detecting burned areas from the massive fires of 2005 and 2012 our approach is able to determine burned areas that might have not been officially recorded (e.g. in 2008). The overall accuracy of our 2012 burned area map reaches up to 83% in pixels with moderate data quality (up to 47% of missing values).