Search

Earth surfaces, such as deserts, salt lakes, and playas, have been widely used in the vicarious radiometric calibration of optical earth observation satellites. In 2009, the Infrared and Visible Optical Sensors (IVOS) sub-group of the Committee of Earth Observation Satellites (CEOS) Working Group on Calibration and Validation (WGCV) designated eight LANDNET reference sites to focus international efforts, facilitate traceability and enable the establishment of measurement "best practices." With support from the European Space Agency (ESA), one of the LANDNET sites, the Tuz Gölü salt lake located in central Turkey, was selected to host a cross-comparison of measurement instrumentation and methodologies conducted by 11 different ground teams across the globe. This paper provides an overview of the preliminary results of the cross-comparison of the ground-based spectral measurements made during the CEOS Land Comparison 13-27 August, 2010 with the simultaneous satellite image data acquisitions of the same site.

In recent years, climate change has greatly affected agricultural activity, sustainability and production, making it difficult to conduct crop management and food security assessment. As a consequence, significant changes in agricultural land and land cover (LC) have occurred, mostly due to the introduction of new agricultural practices, techniques and crops. Earth Observation (EO) data, cloud-computing platforms and powerful machine learning methods can certainly support analysis within the agricultural context. Therefore, accurate and updated agricultural land and LC maps can be useful to derive valuable information for land change monitoring, trend planning, decision-making and sustainable land management. In this context, this study aims at monitoring temporal and spatial changes between 2001 and 2021 (with a four 5-year periods) within the Chi River Basin (NE–Thailand). Specifically, all available Landsat archives and the random forest (RF) classifier were jointly involved within the Google Earth Engine (GEE) platform in order to: (i) generate five different crop type maps (focusing on rice, cassava, para rubber and sugarcane classes), and (ii) monitoring the agricultural land transitions over time. For each crop map, a confusion matrix and the correspondent accuracy were computed and tested according to a validation dataset. In particular, an overall accuracy > 88% was found in all of the resulting five crop maps (for the years 2001, 2006, 2011, 2016 and 2021). Subsequently the agricultural land transitions were analyzed, and a total of 18,957 km2 were found as changed (54.5% of the area) within the 20 years (2001–2021). In particular, an increase in cassava and para rubber areas were found at the disadvantage of rice fields, probably due to two different key drivers taken over time: the agricultural policy and staple price. Finally, it is worth highlighting that such results turn out to be decisive in a challenging agricultural environment such as the Thai one. In particular, the high accuracy of the five derived crop type maps can be useful to provide spatial consistency and reliable information to support local sustainable agriculture land management, decisions of policymakers and many stakeholders.