GEE‐PICX: generating cloud‐free Sentinel‐2 and Landsat image composites and spectral indices for custom areas and time frames – a Google Earth Engine web application

Abstract

Earth observation satellites are collecting vast amounts of free and openly accessible data with immense potential to support environmental, economic, and social fields. As the availability of remotely sensed data increases, so do the methods for accessing and processing it. Many solutions exist for creating cloud‐free image composites from often cloudy satellite data, but these typically require coding skills or in‐depth training in remote‐sensing techniques. This technical barrier prevents many researchers and practitioners from utilising available satellite data. The few user‐friendly solutions that exist often have limitations in terms of data export size and quality assessment capabilities. We developed GEE‐PICX, a web application with an intuitive graphical user interface on the cloud computing platform Google Earth Engine. This tool addresses the aforementioned challenges by creating cloud‐free, analysis‐ready image composites for user‐defined areas and time periods. It utilises Sentinel‐2 and Landsat 5, 7, 8, and 9 images and offers global coverage. Users can aggregate image composites annually or seasonally, with data availability starting from 1984 (the launch of Landsat 5). The workflow automatically filters all available satellite data according to user input, removing clouds, cloud shadows, and snow. It provides spectral band information, calculates various thematic spectral indices (including vegetation, burn, built‐up area, bare soil, snow, moisture, and water indices), and includes a quality assessment band indicating the number of valid scenes per pixel. GEE‐PICX offers a customizable tool for creating custom data products from freely accessible satellite data, catering to researchers with limited remote sensing experience. It provides extensive temporal and global spatial coverage, with server‐side processing eliminating hardware constraints. The tool facilitates easy export of time series as ready‐to‐use rasters with numerous spectral indices, supporting environmental programmes and biodiversity research across various disciplines.

Keywords: cloud masking, cloud‐free image mosaic, environmental monitoring, remote sensing, satellite imagery, time series

Introduction

Understanding environmental changes such as deforestation, desertification, urbanisation, or the expansion of croplands over time is of utmost importance for quantifying and managing the anthropogenic impacts on earth and to support sustainable development and environmental protection (Weng et al. 2008, Mallinis and Georgiadis 2019, Chaves et al. 2020). Satellite remote sensing is a widely used method for monitoring such environmental changes due to the multitude of available sensors and platforms providing continuous data of the Earth's surface (Cord et al. 2017). Remote sensing data is often freely available (e.g. Landsat since the 1980s), enabling scientists to monitor and quantify short- and long-term environmental changes.

Optical remote sensing imagery provides (multi-)spectral information, yet the presence of clouds, cloud shadows, and highly reflective surfaces such as snow can adversely affect sensor measurements, posing challenges in acquiring unbiased and gap-free information (Zhu et al. 2015). Opaque clouds cover approximately 31% of the Earth's surface on average at any time (Guzman et al. 2017), necessitating automatic detection and accurate removal from remote sensing data prior to analysis to prevent data errors at the respective positions. Cloud removal causes gaps in satellite images which can complicate analysis. This can be overcome by merging multiple images from different time points to create cloud-free and gap-free image products. These composite products can then be used for land cover classifications (Verhoeven and Dedoussi 2022), land monitoring applications (Parmes et al. 2017, Carrasco et al. 2019), time series analyses (Yin et al. 2020, Lasaponara et al. 2022), or spatial modelling (Guharajan et al. 2021).

After cloud-correction, multi-spectral information can either be used directly (Zhu et al. 2018) or via derived spectral indices, which are combinations of the spectral reflectance from two or more wavelengths (Chaves et al. 2020, Rudd et al. 2021). Spectral indices are often more suitable for specific analyses than raw spectral information due to more clearly defined and interpretable properties (Rudd et al. 2021). The most popular spectral indices are vegetation indices, but other indices (e.g. for burned areas, man-made (built-up) features, water or ice) are available, too (Petropulos and Kalaitzidis 2011, Chaves et al. 2020, Montero et al. 2023).

The increasing availability of remote-sensing data is accompanied by advancements in software for managing large data sets and processing chains. In the geospatial community, Google Earth Engine (GEE), a cloud computing platform powered by Google Cloud infrastructure, has gained popularity for generating analysis-ready image products for various applications across different spatial and temporal scales (Piao et al. 2019, Wu 2020, Yang et al. 2021, Lasaponara et al. 2022). In recent years, developers have created various tools and applications designed to streamline access and pre-processing of satellite imagery, for example ClimateEngine (Huntington et al. 2017), Awesome Spectral Indices (ASI, Montero et al. 2023), geemap (Wu 2020), and rgee (Aybar et al. 2020). Furthermore, platforms like SentinelHub (Sinergise 2023a) or EarthExplorer (United States Geological Survey 2023) provide user-friendly access to large databases of remote-sensing data. While some of these tools require programming skills (ASI, geemap, rgee), others offer more user-friendly interfaces, but have limitations such as restricted downloads for larger areas (ClimateEngine, SentinelHub) or they do not provide options for image aggregation or further data processing like spectral index calculation (EarthExplorer). Thus, we developed GEE-PICX, a Google Earth Engine web application providing advanced satellite data products for non-experts, which addresses these limitations by offering cloud-masking, data aggregation, and spectral index calculation, similar to ClimateEngine or SentinelHub. It also provides a novel quality assessment band, specifying valid scenes per pixel in image aggregates. GEE-PICX allows for significantly larger data downloads compared to other platforms, though visualisation of large areas within the web application may be more limited than for smaller areas. Unlike specialised tools for specific applications such as land-cover classification (REMAP, Murray et al. 2018), or crop-climate-suitability mapping (Peter et al. 2020), GEE-PICX focuses on providing flexible access to satellite data with optional spectral index and data quality information. This approach ensures broad applicability across various research domains and analysis types (see the Supporting information for an overview and comparison of available applications).

GEE-PICX generates, visualises, and exports cloud-free, analysis-ready composites of satellite images for user-defined areas and time steps, with global data coverage. We followed five design principles in developing GEE-PICX:

• 1. Flexibility of user input.Users can select the satellite platform (Landsat or Sentinel-2), study area boundaries, time range, maximum cloud cover (for single images), aggregation mode, and image bands. Relevant scenes are automatically selected from the data catalogue according to user input. Moreover, the modular design allows users to easily add custom indices.

• 2. Ease of use.The application features a self-explanatory graphical user interface. It only requires a Google account, web browser, and internet connection, with no additional hardware or software requirements due to server-side processing.

• 3. Export of large data sets.Export size is limited only by Google drive storage capacity.

• 4. Generation of analysis-ready data.Produces cloud-free image composites with spectral bands, spectral indices, and a quality assessment band (valid scenes per pixel). The export image resolution and coordinate reference system are customizable.

• 5. Data visualisation. Data sets can be visualised in the browser prior to export.

This paper presents GEE-PICX, a web-based application designed to simplify access to satellite imagery analysis. Our objectives are to describe the technical workflow and features of GEE-PICX, highlighting how it addresses common challenges in satellite data processing, and to demonstrate the tool's versatility through two diverse use cases in ecological research. We then present two examples that illustrate the application of GEE-PICX in different ecological contexts and discuss the potential impact of GEE-PICXon broadening the use of remote-sensing data across various disciplines.

Workflow description

Overview

Users can access the web application via the provided application link (Data availability). The script is written in JavaScript and commented to facilitate orientation. No manual code adjustments are necessary. With the application running, users can define parameters according to their requirements in the application interface next to the map. The application then internally processes user inputs, executing functions for satellite image (pre-)processing, visualisation, and export preparation. Data visualisations are available directly in the application. The products can be exported at user-defined spatial resolutions and coordinate systems, and are ready to use for subsequent analyses.

User input

Below we provide a detailed overview of the choices users can make for creating customised data exports (see the Supporting information for advanced information on data processing).

Satellite data:The application can provide image composites based on either the Landsat or the Sentinel-2 mission. Both Landsat and Sentinel-2 data sets consist of atmospherically and topographically corrected Level-2A products that show surface reflectance values with atmospheric correction applied. The data set choice can be based on either the required spatial resolution or length of the time series. The earliest Level-2A products from Landsat date back to 1984 (at 30 m resolution), whereas Sentinel-2 Level-2A products have been available since 2017 (at 10 m resolution in the visible and NIR spectrum). The availability of Landsat data from the late 1980s and early 1990s is much lower than in recent years, when more Landsat missions are simultaneously acquiring imagery at a higher temporal frequency. When selecting Landsat missions for analysis it is important to consider the potential impact of the Landsat 7 scan line corrector (SLC) failure. The SLC, which normally ensures continuous image capture, failed in 2003. This failure resulted in data gaps affecting about 22% of each scene. While the overall image quality remains intact, these gaps can limit the usability of Landsat 7 images for applications requiring seamless coverage (United States Geological Survey 2022). However, researchers developed gap-filling techniques to mitigate this issue (Storey et al. 2005). In GEE-PICX, when selecting Landsat as the platform, users have the option to include imagery from all Landsat missions (5, 7, 8, 9) to create image composites, or they can choose to include only Landsat-8 and Landsat-9 data, which avoids including erroneous Landsat-7 images. However, Landsat-8 data are only available from 2013 and Landsat-9 data from 2022 (see the Supporting information for more specific information on the satellite missions).

Area of interest:The boundary of the study area can be defined either by uploading a shapefile as an Earth Engine asset (Google Earth Engine 2021), or by manually drawing a polygon on the Google Earth Engine map. Data coverage is global.

Time period: The time frame can be specified by year and month range. By default, scenes are aggregated for one year (months 1–12). Users can create seasonal image aggregates by narrowing the selection to specific consecutive months (also crossing the year boundary). Users can request export of imagery from multiple years at once.

Cloud cover filter:Optical satellite images may exhibit partial or complete cloud coverage. The pixel-level cloud masks included in scenes cannot perfectly detect and filter out all clouds and cloud shadows (Sanchez et al. 2020). Therefore, the cloud cover percentage per scene is utilised to enhance the quality of image composites by removing scenes exceeding a cloud cover threshold. By default, images with cloud cover exceeding 65% are excluded prior to aggregation. Opting for a 100% threshold includes all images captured within the specified study area and time frame. Cloud masking leads to data gaps in all images affected by cloud cover. If all scenes have data gaps at the same pixels, the image composite will also have data gaps at this location.

Image bands:Users can select single or multiple spectral bands, as well as spectral indices and a valid pixel band, by activating the corresponding checkboxes. Spectral bands convey surface reflectance data and are correlated with chlorophyll and other pigments, vegetation structure, and water content. Key correlations include Green, Red, and Red-edge bands with chlorophyll and pigments, NIR bands with leaf structure, and SWIR with vegetation structure and water content (Chaves et al. 2020). Spectral indices result from mathematical combinations of the spectral bands (see the Supporting information for details on all available indices). The valid pixel band is a quality assessment layer specifying the number of valid scene values that are aggregated at each pixel.

Aggregation mode: The aggregation mode determines which summary statistic is applied to the pixel values of all selected images. Available choices are mean, median, and standard deviation.

Coordinate system:The application offers the choice to export rasters in UTM and WGS 84 (EPSG 4326) formats. If UTM is selected, the application automatically identifies the appropriate zone. If the study area spans multiple UTM zones, images can only be exported in WGS 84.

Spatial pixel resolution: The application provides the options to export images at four different spatial resolutions ranging from 10 to 100 m. Opting for high resolutions in extensive study areas could yield products exceeding several gigabytes, potentially posing challenges for subsequent analyses. Users should choose a resolution that matches their research or monitoring objectives.

Image export

After initiating the export in the application user interface, users can inspect and execute the actual image export(s) within the upper-right window via the Console and Tasks tabs (see Data availability). Two image collections will be automatically added to the Console. The first contains all individual satellite images after filtering, the second contains the image aggregates available for export. Each annual/seasonal image that appears in the Tasks manager needs to be exported individually. When clicking ‘Run', a pop-up window will appear in which the user can optionally modify export names, coordinate reference system, scale, export destination, and file format.

The easiest way to save the files on a local computer is to export them to a Google drive folder which is connected to the users' Google account, and then download the data from there. Multiple image exports run in parallel, and depending on study area size each export can take from minutes to hours (or even days for study regions measuring hundreds of thousands of square kilometres). When exporting large data sets, GEE splits each image into smaller tiles. After downloading them from Google Drive, they can either be merged to a large contiguous mosaic, or be used as a virtual raster.

Except for the ‘valid-pixel' band, all band values of the export images are multiplied by 10 000. This allows the raster values to be stored as integer values (signed 16-bit) instead of floating point values, thus reducing the file size of exports.

GEE assigns a value of zero to data gaps in image composites during export to Google Drive, potentially biassing subsequent analyses. We provide an R script for converting zero values back to NA with the help of the valid pixel layer prior to further analyses (Data availability).

Data visualisation

Users can visualise their export data on the map by selecting either a spectral index or various band combinations. Band combinations can highlight certain features (e.g. vegetation types, water bodies, and urban areas) due to correlations between measurable biophysical properties on the Earth's surface and remotely sensed surface reflectance (Price et al. 2002). After choosing the visualisation parameter, all aggregated images will be added to the map with default visualisation settings. Adjustments to visualisation parameters can be made individually within the map's layer panel box (follow instructions on Zenodo link, see Data availability). All indices have a valid value range from −1 to 1 in the web application. GEE may encounter computational problems for visualisation if the data are too large due to the size of the study area and/or the length of the time period. This may lead to scaling error messages and some objects will not be displayed on the map (or also the GEE Console). Visualisation problems, however, do not affect image exports, which are always possible and only limited by the storage capacity of the user's Google drive.

In addition to the visualisation options in GEE, we provide an interactive R Shiny application for visualising image time series (Data availability).

Case examples

Example A shows deforestation in Brazil using historical Landsat images, while Example B focusses on seasonal land cover changes in the city of Würzburg (Germany), emphasising the enhanced level of detail provided by Sentinel-2 imagery. In both examples a combination of three spectral bands (SWIR1-NIR-R; NIR-R-G) and a spectral index (NDVI) are shown together with the number of valid pixels (Fig. 1). The data contain more spectral bands and indices not shown here.

[IMAGE OMITTED. SEE PDF]

In the Amazon rainforest, deforestation has become a pressing environmental concern over the past several decades. Soy farms, along with other agricultural expansion, have played a significant role in driving deforestation in the Amazon (Nepstad et al. 2006). We used GEE-PICX to generate and export annual image aggregates for an area in Ariquemes, Rondônia, Brazil, for 1991 and 2021, illustrating the magnitude of change over three decades. Such annual aggregates (or composites) are suitable for inferences on broad trends, but average seasonal dynamics or land cover changes within a year, making them unsuitable for mapping floods, for example.

The second example shows the seasonal changes in land cover/land use in the city centre of Würzburg, Germany, and highlights the surrounding ring-shaped park. The region's transition between summer and winter was captured in seasonal satellite image composites and showcases the distinct phenological variations. The higher spatial resolution of Sentinel-2 imagery allows better discrimination of small-scale features and proves particularly valuable in the context of land cover and land use monitoring. Seasonal variation in cloud cover can lead to seasonal bias in the available data. Snow cover can also affect the quality of seasonal image compositions because the applied cloud mask algorithm (Supporting information) does not perfectly mask highly reflective surfaces such as clouds or snow in individual scenes.

Conclusion

Satellite imagery is essential for many environmental and conservation studies. However, utilising freely available satellite products often requires expertise in data selection and pre-processing, and substantial computational resources. Many environmental and conservation studies therefore primarily rely on pre-packaged thematic products (Wong et al. 2022), which may lack the detail necessary to address specific research questions. GEE-PICX addresses these challenges by simplifying access to cloud-free satellite image composites. It effectively addresses cloud cover issues, which are particularly problematic in tropical or mountainous regions (Hribljan et al. 2017, Sanchez et al. 2020). Through its intuitive interface, users can easily generate and export (multi-)temporal cloud-free satellite images for any region, with data availability starting from 1984 (varying by region).

GEE-PICX offers access to both Landsat and Sentinel-2 archives and provides multispectral information complemented by various spectral indices and a data quality metric. These are typically not or only partly present in the output of other platforms. GEE-PICX is further set apart by its capability to process extensive areas with very large download sizes. By making these rich satellite data archives accessible to non-remote-sensing scientists and practitioners, GEE-PICX supports the integration of satellite data into a wide range of environmental and conservation projects.

By lowering the barriers to satellite data analysis, GEE-PICX aims to bridge the gap between satellite data availability and its practical use in environmental research and conservation efforts.

To cite GEE-PICX or acknowledge its use, cite this Software note as follows, substituting the version of the application that you used for ‘version 1.0':

Pflumm, L., Kang, H., Wilting, A. and Niedballa, J. 2024. GEE-PICX: Generating cloud-free Sentinel-2 and Landsat image composites and spectral indices for custom areas and time frames – a Google Earth Engine web application. – Ecography 2024: e07358 (ver. 1.0).

Acknowledgements

– We would like to express our gratitude to several individuals and organizations who have played an important role in the successful completion of this project. First of all, we would like to thank Marius Philipp, who gave us great technical support as a tutor during workflow and app development in Google Earth Engine. Moreover, we thank Matthias Baumann and Julian Oeser for insightful discussions about processing and utilization of satellite data time series. Open Access funding enabled and organized by Projekt DEAL.

Funding

– This work was supported by funding from the following sources: Luisa Pflumm: United States Agency of International Development (Biodiversity Conservation project in Viet Nam, 7204402CA00001). Hyeonmin Kang: United States Agency of International Development (Biodiversity Conservation project in Viet Nam, 7204402CA00001). Andreas Wilting: Leibniz Institute for Zoo and Wildlife Research. Jürgen Niedballa: United States Agency of International Development (Biodiversity Conservation project in Viet Nam, 7204402CA00001), Bundesministerium für Bildung und Forschung (SmartPatrol project, 16LW0439). The publication of this work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation – Projektnummer 491292795).

Author contributions

Luisa Pflumm: Conceptualization (lead), Methodology (lead), Software (equal), Validation (equal), Visualization (lead), Writing - original draft (equal), Writing - review and editing (equal). Hyeonmin Kang: Conceptualization (equal), Methodology (equal), Software (equal), Validation (equal), Writing - original draft (supporting). Andreas Wilting: Conceptualization (equal), Funding acquisition (lead), Project administration (lead), Supervision (equal). Jürgen Niedballa: Conceptualization (equal), Methodology (equal), Supervision (equal), Validation (equal), Visualization (supporting), Writing - original draft (equal), Writing - review and editing (equal).

Data availability statement

Source code is available from the Zenodo Repository: (Pflumm et al. 2024).

References

Aybar, C., Wu, Q., Bautista, L., Yali, R. and Barja, A. 2020. rgee: an R package for interacting with Google Earth Engine. – J. Open Source Softw. 5: 2272.

Carrasco, L., O'Neil, A. W., Morton, R. D. and Rowland, C. S. 2019. Evaluating combinations of temporally aggregated Sentinel‐1, Sentinel‐2 and Landsat 8 for land cover mapping with Google Earth Engine. – Remote Sens. 11: 288.

Chaves, E. D. M., Picoli, C. A. M. and Sanches, D. I. 2020. Recent applications of Landsat 8/OLI and Sentinel‐2/MSI for land use and land cover mapping: a systematic review. – Remote Sens. 12: 3062.

Cord, A. F., Brauman, K. A., Chaplin‐Kramer, R., Huth, A., Ziv, G. and Seppelt, R. 2017. Priorities to advance monitoring of ecosystem services using earth observation. – Trends Ecol. Evol. 32: 416–428.

Google Earth Engine 2021. Managing assets. – https://developers.google.com/earth‐engine/guides/asset_manager.

Guharajan, R., Mohamed, A., Wong, S. T., Niedballa, J., Petrus, A., Jubili, J., Lietz, R., Clements, G. R., Wong, W. M., Kissing, J., Lagan, P. and Wilting, A. 2021. Sustainable forest management is vital for the persistence of sun bear Helarctos malayanus populations in Sabah, Malaysian Borneo. – Forest Ecol. Manage. 493: 119270.

Guzman, R., Chepfer, H., Noel, V., Vaillant de Guélis, T., Kay, J. E., Raberanto, P., Cesana, G., Vaughan, M. A. and Winker, D. M. 2017. Direct atmosphere opacity observations from CALIPSO provide new constraints on cloud‐radiation interactions. – JGR Atmospheres 122: 1066–1085.

Hribljan, J. A., Suarez, E., Bourgeau‐Chavez, L., Endres, S., Lilleskov, E. A., Chimbolema, S., Wayson, C., Serocki, E. and Chimner, R. A. 2017. Multidate, multisensor remote sensing reveals high density of carbon‐rich mountain peatlands in the páramo of Ecuador. – Global Change Biol. 23: 5412–5425.

Huntington, J. L., Hegewisch, K. C., Daudert, B., Morton, C. G., Abatzoglou, J. T., McEvoy, D. J. and Erickson, T. 2017. Climate engine: cloud computing and visualization of climate and remote sensing data for advanced natural resource monitoring and process understanding. – Bull. Am. Meteorol. Soc. 98: 2397–2410.

Jiang, Z., Huete, A. R., Didan, K. and Miura, T. 2008. Development of a two‐band enhanced vegetation index without a blue band. – Remote Sens. Environ. 112: 3833–3845.

Lasaponara, R., Abate, N., Fattore, C., Aromando, A., Cardettini, G. and Di Fonzo, M. 2022. On the use of Sentinel‐2 NDVI time series and Google Earth Engine to detect land‐use/land‐cover changes in fire‐affected areas. – Remote Sens. 14: 4723.

Mallinis, G. and Georgiadis, C. 2019. Editorial of special issue “remote sensing for land cover/land use mapping at local and regional scales”. – Remote Sens. 11: 2202.

Montero, D., Aybar, C., Mahecha, M. D., Martinuzzi, F., Söchting, M. and Wieneke, S. 2023. A standardized catalogue of spectral indices to advance the use of remote sensing in Earth system research. – Sci. Data 10: 1–20.

Murray, N. J., Keith, D. A., Simpson, D., Wilshire, J. H. and Lucas, R. M. 2018. Remap: an online remote sensing application for land cover classification and monitoring. – Methods Ecol. Evol. 9: 2019–2027.

Nepstad, D. C., Stickler, C. M. and Almeida, O. T. 2006. Globalization of the Amazon soy and beef industries: opportunities for conservation. – Conserv. Biol. 20: 1595–1603.

Parmes, E., Rauste, Y., Molinier, M., Andersson, K. and Seitsonen, L. 2017. Automatic cloud and shadow detection in optical satellite imagery without using thermal bands – application to Suomi NPP VIIRS images over Fennoscandia. Remote Sens. 9: 806.

Peter, B. G., Messina, J. P., Lin, Z. and Snapp, S. S. 2020. Crop climate suitability mapping on the cloud: a geovisualization application for sustainable agriculture. – Sci. Rep. 10: 15487.

Pflumm, L., Kang, H., Wilting, A. and Niedballa, J. 2024. Codefrom: GEE‐PICX: Generating cloud‐free Sentinel‐2 and Landsat image composites and spectral indices for custom areas and time frames – a Google Earth Engine web application. – Zenodo Digital Repository, https://zenodo.org/records/14541814.

Piao, S., Liu, Q., Chen, A., Janssens, I. A., Fu, Y., Dai, J., Liu, L., Lian, X., Shen, M. and Zhu, X. 2019. Plant phenology and global climate change: current progresses and challenges. – Global Change Biol. 25: 1922–1940.

Price, K. P., Guo, X. and Stiles, J. M. 2002. Optimal Landsat TM band combinations and vegetation indices for discrimination of six grassland types in eastern Kansas. – Int. J. Remote Sens. 23: 5031–5042.

Rudd, D. A., Karami, M. and Fensholt, R. 2021. Towards high‐resolution land‐cover classification of Greenland: a case study covering Kobbefjord, Disko and Zackenberg. – Remote Sens. 13: 3559.

Sanchez, A. H., Picoli, M. C. A., Camara, G., Andrade, P. R., Chaves, M. E. D., Lechler, S., Soares, A. R., Marujo, R. F. B., Simões, R. E. O., Ferreira, K. R. and Queiroz, G. R. 2020. Comparison of cloud cover detection algorithms on Sentinel–2 images of the Amazon tropical forest. – Remote Sens. 12: 1284.

Sinergise 2023. SentinelHub. – https://www.sentinel‐hub.com/.

Sousa, D. and Small, C. 2023. Which vegetation index? Benchmarking multispectral metrics to hyperspectral mixture models in diverse cropland. – Remote Sens. 15: 971.

Storey, J., Scaramuzza, P., Schmidt, G. and Barsi, J. 2005. Landsat 7 scan line corrector‐off gap‐filled product development. – In: Proceedings of the Pecora 16 “Global Priorities in Land Remote Sensing” conference (Vol. 16, pp. 23–27). Sioux Falls, SD, USA.

United States Geological Survey 2022. Landsat – earth observation satellites. Fact Sheet 2015–3081, ver. 1.4. – https://pubs.usgs.gov/fs/2015/3081/fs20153081.pdf.

United States Geological Survey 2023. EarthExplorer. – https://earthexplorer.usgs.gov/.

Verhoeven, V. B. and Dedoussi, I. C. 2022. Annual satellite‐based NDVI‐derived land cover of Europe for 2001–2019. – J. Environ. Manage. 302: 113917.

Weng, Q., Quattrochi, D. and Xian, D. 2008. Special issue remote sensing of land surface properties, patterns and processes. – MDPI Remote Sens, https://www.mdpi.com/journal/sensors/special_issues/remote‐sensing‐land‐surface‐properties#published.

Wong, S. T. et al. 2022. How do terrestrial wildlife communities respond to small‐scale Acacia plantations embedded in harvested tropical forest? – Ecol. Evol. 12: [eLocator: e9337].

Wu, Q. 2020. geemap: a Python package for interactive mapping with Google Earth Engine. – J. Open Source Softw. 5: 2305.

Yang, W., Chen, X., Wang, C., Cao, R., Zhu, X. and Shen, B. 2021. Special issue time series analysis in remote sensing: algorithm development and applications. – MDPI Remote Sens., https://www.mdpi.com/journal/remotesensing/special_issues/time_series_algorithm_development.

Yin, H., Brandão Jr, A., Buchner, J., Helmers, D., Iuliano, B. G., Kimambo, N. E., Lewińska, K. E., Razenkova, E., Rizayeva, A., Rogova, N., Spawn, S. A., Xie, Y. and Radeloff, V. C. 2020. Monitoring cropland abandonment with Landsat time series. – Remote Sens. Environ. 246: 111873.

Zeng, Y., Hao, D., Huete, A., Dechant, B., Berry, J., Chen, J. M., Joiner, J., Frankenberg, C., Bond‐Lamberty, B., Ryu, Y., Xiao, J., Asrar, G. R. and Chen, M. 2022. Optical vegetation indices for monitoring terrestrial ecosystems globally. – Nat. Rev. Earth Environ. 3: 477–493.

Zhu, C., Zhang, X. and Huang, Q. 2018. Four decades of estuarine wetland changes in the Yellow River delta based on Landsat observations between 1973 and 2013. – Water 10: 933.

Zhu, Z., Wang, S. and Woodcock, C. E. 2015. Improvement and expansion of the Fmask algorithm: cloud, cloud shadow, and snow detection for Landsats 4–7, 8, and Sentinel 2 images. – Remote Sens. Environ. 159: 269–277.