Resumen de Potencial del acervo de imágenes Landsat disponible en Google Earth Engine para el estudio del territorio mexicano

Jonathan V. Solórzano, J. Alberto Gallardo Cruz, Candelario Peralta Carreta

• español

  Actualmente, las imágenes Landsat corresponden a uno de los acervos de mediana resolución de mayor uso e impacto. Por otro lado, la aparición de herramientas como Google Earth Engine ha permitido potenciar enormemente el procesamiento y los resultados obtenidos a partir de estas imágenes. En este estudio se consultó el acervo de imágenes Landsat disponibles en Google Earth Engine para México en el periodo 1972 – 2017 con el fin de analizar el potencial de dicho acervo para estudiar el territorio nacional. En esta consulta se incluyó la información del sensor que lo registró, la clave de la imagen (path / row) y las condiciones de nubosidad sobre la superficie terrestre. En total se encontraron 89,649 imágenes disponibles que cubren todo el país en el periodo mencionado. No obstante, los resultados mostraron que para el período 1972 – 1992 (Landsat 1 – 4) prácticamente no existen imágenes disponibles para México. Por el contrario, a partir de 1993 se encontró un acervo de imágenes numeroso y continuo. Por ello, se sugiere utilizar como periodo inicial 1993 para cualquier estudio futuros interesado en caracterizar la dinámica de la superficie en territorio mexicano a mediano plazo (p. ej., cambio uso de suelo). El mayor número de imágenes fue registrado por el sensor Landsat 5 TM (38,897 imágenes), seguido del 7 ETM+ SLC-off (31,254 imágenes). Por último, para fomentar el uso de Google Earth Engine para el procesamiento de imágenes se programó una rutina para construir mosaicos anuales disponible para cualquier usuario.

• English

  Landsat imagery is one of the world’s longest-running and most widely used high-resolution collections. To make extensive use of this vast archive, platforms such as Google Earth Engine are necessary to reduce processing time and facilitate analyses. This study aimed to identify the Landsat scenes acquired between 1972 and 2017 covering the Mexican territory that are available through Google Earth Engine. The query was conducted on the Tier 1 raw scenes imagery collection (as these are the images with the lowest geospatial error between scenes) using a Javascript program in the Google Earth Engine platform. For each scene the query obtained information on the sensor, acquisition date, Landsat key (path/row), and cloudiness percentage over the Earth's surface. The information obtained was processed in R 3.5.1.Data acquisition took approximately 10 seconds, which shows the enormous processing power of Google Earth Engine. A total of 146 Landsat keys are necessary to encompass the entire Mexican territory with scenes acquired by the sensors Landsat 1-MSS through Landsat 3-MSS, and 134 keys are necessary for scenes recorded by sensors Landsat 4 MSS, TM and Landsat 8 OLI. We gathered a total of 89,649 scenes acquired between 1972 and 2017 covering the country. However, the number of scenes available for a given year varied widely; only 9 scenes were found for 1972 (0.06 images per path/row, on average), and 5403 for 2017 (40.32 images per path/row, on average). Over this period, the number of scenes available increased in those years when a new sensor started operations (e.g., 1984, 1999, and 2013) and when the Landsat archive was centralized by the USGS (in 1993). By contrast, the number of scenes available decreased in those years when a satellite ceased operations (e.g., 2012).The sensors that acquired the greatest number of scenes were Landsat 5 TM (38,897 scenes), followed by Landsat 7 ETM+ SLC-off (31,254) and Landsat 8, although with a significantly smaller number (12,796). In addition, almost half of scenes had less than 50% cloudiness over the Earth's surface, although when examined by path/row, it is clear that the driest areas of the country had a lower cloudiness vs. more humid zones. By design, Landsat images should cover the entire Earth's surface since 1972; however, for many areas of the world, including Mexico, the oldest scenes are usually very scarce for the following reasons: lack of reception and/or storage facilities in those regions; lack of proper storage practices; the scenes did not meet the quality standards for inclusion in the Tier-1 collection. Therefore, characterizing the collection of Landsat scenes actually available for the country contributes to understand the true possibilities that this archive provides for studying the Mexican territory.The results of this study are expected to guide future investigations using Landsat imagery to explore the Mexican territory, by providing information on the number of scenes actually available per year and their cloudiness condition. Finally, in order to foster the Google Earth Engine for image processing, a JavaScript routine to build annual mosaics of the highest quality Tier-1 surface reflectance data available was written and made available to any user. This routine is supplemented by a brief tutorial in Spanish that aims to provide an introduction to Google Earth Engine and promote its use in Spanish-speaking countries.