A synergistic use of spectral data with moderate to high spatial resolution from more than one source is getting momentum due to successful launches under the ESA Sentinel program, the continuing NASA-USGS Landsat program and the increased use of commercial high-resolution data. Landsat and Sentinel-2 optical data are now used synergistically by many researchers, often combined with Sentinel-1 radar data. Efficient and synergistic use of these sensor data increases the number of observations available for studies. This presentation will provide an overview of the Multi-Source Land Imaging (MuSLI) projects – the component under the NASA Land-Cover/Land-Use Change program that started 5 years ago. The talk will include a digest of the results from the first MuSLI round directed at prototyping the synergistic use of Landsat and Sentinel data, discussion of the issues of three projects that aim at developing a system for routine continental-scale products based on the first-round prototyping study, and preliminary results from the second MuSLI round prototyping projects. The products include circumpolar albedos, burnt areas in Africa, vegetation phenology of North America and many more that will be showcased in the presentation. Prospects on the near-term future use of available constellations of sensors providing high-resolution observations will also be briefly discussed.

To meet the nutritional needs of the world’s growing population, global food production will need to roughly double by 2050. This increased production must be accomplished within the constraints of a non-uniform distribution of freshwater resources and an amplifying climate cycle. To make significant strides toward these goals, we must first better understand the variability in water usage, crop yield, and water productivity embedded in landscapes under current conditions as it relates to variations in land and crop management strategies. This paper describes methods for routine production of high spatiotemporal resolution evapotranspiration (ET), crop stress and yield products using a multi-sensor data fusion approach. ET is estimated using a well-established surface energy balance algorithm, which uses thermal infrared (TIR) retrievals of land-surface temperature along with vegetation cover and albedo information from the surface reflectance bands. This approach fuses moderate resolution, near-daily retrievals of ET from sensors like MODIS and VIIRS with periodic finer scale data from Landsat, Sentinel-2, ECOSTRESS and other Landsat-like sensors to generate multi-year timeseries of gridded products at daily timesteps and 30m spatial resolution. Collectively, these high spatiotemporal resolution ET “datacubes” provide valuable field-scale diagnostics of water use and moisture stress required for monitoring efficiency of agricultural production systems and forecasting yield. Improvement in moisture stress and yield mapping capabilities using multi-sensor data will be demonstrated over highly managed agricultural systems in the U.S., Brazil, Czech Republic and Lebanon – where performance of coarser resolution datasets is known to be degraded due to mixed pixels and shifting phenology. ET retrieval performance is assessed in comparison with flux tower observations where available, while utility for yield prediction is evaluated using field-to-district level reports for various crop types. In combination, the multi-sensor ET and yield datasets can be used to map water productivity (“crop-per-drop”) at field scale over agricultural landscapes to quantify differences in efficiencies between regions, climates and cropping systems.

Global multiannual fire products derived systematically from coarse spatial resolution satellite data, such as the 500 m NASA Moderate Resolution Imaging Spectroradiometer (MODIS) burned area product, provide an important source of information for the multiple user communities involved in fire science and applications. There is, however, an outstanding demand for systematically generated higher spatial resolution burned area products. This paper presents a combined Landsat-8 Sentinel-2 burned area mapping algorithm. The algorithm is automated as even modest levels of user intervention are not scalable to the considerable data volumes acquired by the Landsat-8 and Sentinel-2 sensors. The data handling integrates recent research on Landsat-8 and Sentinel-2 pre-processing to generate registered, surface nadir BRDF-adjusted reflectance (NBAR) sensor time series that are used as an input. The different sensor data are combined through a random forest change regression, trained with synthetic data built from laboratory and field spectra of unburned and burned vegetation and using a spectral model of fire effects on reflectance. The random forest regression is applied independently at each gridded 30 m pixel location on a temporally rolling basis considering three months of sensor data to map the central month. Temporal consistency checks are used to reduce commission errors due to non-fire related spectral changes, and a region growing algorithm is used to reduce omission errors due to temporally sparse observations. In the resulting product, each 30 m pixel is labeled as burned, unburned or unmapped; at burned pixels the estimated day of burning is also defined, together with an estimate of the product of the subpixel fraction burned (f) and the combustion completeness (cc). The algorithm is demonstrated using six months of every available Landsat-8 and Sentinel-2A acquisition over 10° × 10° of Southern Africa. Experiments comparing the mapped burned areas considering Sentinel-2A only and both Landsat-8 and Sentinel-2A data, indicate a greater area burned and a smaller number of unmapped pixels when both sensors are used. The results are compared with contemporaneous NASA Collection 6 MODIS fire products. Temporally, the Sentinel-2A and Landsat-8 30 m product report the day of burning on average three days later than the MODIS 500 m burned area product, because of the lower revisit frequency of the Sentinel-2A and Landsat-8 observations. Spatially, the Sentinel-2A and Landsat-8 30 m product capture more detail than the MODIS 500 m burned area product, with systematically higher burned area estimates. Despite the areal differences, the spatial pattern of the two products is similar, as reflected by the correlation (r2 ~ 0.7) and slope (> 0.8) of monthly regressions of the proportions of area burned defined in coarse resolution cells between the two products, taking into account the 3 day temporal bias. Finally, the Sentinel-2A and Landsat-8 30 m product is validated by comparison with select multi-date 3 m PLANET imagery, with high visual agreement in the location and approximate day of burning and higher f.cc values in the interiors of distinct burns than at the edges, and lower values in areas containing mixes of burned and unburned 3 m pixels. New NASA finding for this project has been accrued and plans for future Africa-wide processing are discussed.

Remote sensing sensors onboard operational satellites cannot have high spectral, spatial and temporal resolutions simultaneously. In addition, clouds and aerosols can adversely affect the signal contaminating the land surface observation. Common solutions to tackle these problems rely on fusing images of different sensors with different characteristics to overcome individual limitations by combining their specifications. Here, we present a general synergistic approach to combine multispectral images of different sensors to obtain gap free and reduced-noise Sentinel 2 observations. Moreover, the proposed Bayesian blending method provides information about the uncertainty associated with the final reflectance estimates, which is necessary to conduct error propagation analyses to higher level remote sensing products. Our approach uses imagery from the Landsat mission (30 m spatial resolution and 16 day revisit cycle), MODIS mission from Terra and Aqua platforms (500 m spatial resolution and daily revisit cycle), and Sentinel 2 (10/20 m spatial resolution and 5 days revisit cycle). We have implemented a bias aware Kalman filter method in the Google Earth Engine (GEE) platform to obtain the fused images using Landsat, MODIS, and Sentinel time series. A bias correction was added in the Kalman filter estimates to account for the fact that both model and observation errors may have a non-zero mean and are temporally autocorrelated. These error characteristics are common in remote sensing observations and are mostly related with cloud and aerosols contamination. The followed approach takes full advantage of the computational power of GEE to process large amounts of data (Landsat, MODIS and Sentinel), and uses simple and scalable operations that are able to run optimally in a cloud computing environment. The proposed method shows positive results producing gap free Sentinel 2 images along with their associated uncertainties.

The Harmonized Landsat/Sentinel-2 (HLS) project is providing a harmonized Level-2 surface reflectance product from USGS/NASA’s Landsat-8 and ESA’s Sentinel-2 missions, which, combined, provide a high revisit frequency (3-5 days globally). These satellites’ sampling characteristics provide nearly constant observation geometry and low illumination variation within the scene. However, they are influenced by the seasonal illumination variations which impact the surface reflectance increasing it for low solar zenith angles and decreasing it for large zenith angles. In addition, Sentinel-2 and Landsat 8 images differ in view geometry – in extreme cases the difference in view angle for ground target can be 20 degrees. This results in a significant reflectance variability when data from the two sensors are combined together in a single time series. In this work, we present an approach to derive the BRDF parameters at the HLS product spatial resolution (30 m). It is based on Franch et al. (2014) method to derive Landsat surface albedo. It is based on the BRDF parameters from the MODerate Resolution Imaging Spectroradiometer (MODIS) surface reflectance product (M{O,Y}D09) resampled at 1km spatial resolution using the VJB method (Vermote et al., 2009). HLS unsupervised classification images are used to disaggregate the BRDF parameters to the HLS spatial resolution and to build a BRDF parameters database at HLS scale. Franch et al. (2014) just analyzed the accuracy of the model for surface albedo estimation. However, the algorithm also calculates the BRDF parameters for each pixel, allowing the correction of directional effects through the computation of the normalized surface reflectance. In this work, we apply the Franch et al. (2014) method to normalize the surface reflectance for BRDF effects using the NASA’s Harmonized Landsat Sentinel-2 (HLS) product. We test the results over five different sites of the US SURFRAD network, five sites of the Australian OzFlux network and a site in the Peruvian Amazon using field surface albedo measurements as a reference to validate the performance of the BRDF model. Preliminary results of the albedo validation show a good performance with Root Mean Square Error (RMSE) of 0.015 and a coefficient of determination higher than 0.8.

In the last decade the offer of Earth Observation (EO) data has considerably increased, in particular in case of multi spectral high resolution data. Whilst the interest of merging these various data streams is now prevalent in the community, there is not a common approach to achieve this objective. Also, several major scientific and technical questions are still pending.
In this context, ESA has proposed the development of an innovative prototype (pre-operational demonstration) service providing Analysis Ready Harmonized Landsat 8 (LS8) and Sentinel2 (S2) data/products to the user. This harmonisation process allows to increase the theoretical number of acquisitions of this virtual constellation (95 products / year) by 30 % with respect to Sentinel-2 (S2A & S2B) only acquisitions (73 products / year). The final objective is to promote the pixel-based analysis with the extraction of fit-for-purpose dense time series, essential for bio-geophysical variables monitoring for instance.

For this purpose, the sen2like tool has been developed. Using the S2 tiling system based on the Military Grid Reference System (MGRS), the sen2like tool processes S2/MSI Level-1 products and LS8/OLI Level-1 products and create a harmonized surface reflectance data stack. Working on the same baseline principles as NASA HLS initiative, geometric, radiometric and image processing algorithms are applied. A significant part of the work has been focused on the validation activities of each algorithm involved and also on the final data stack products.

The scope of this paper is to present a detailed description of the functionalities available in the sen2like tool as well as the trade-off adopted to keep an efficient solution in term of processing time and usage. Furthermore, the results of the processing algorithm validation and the results of the time series quality control will be given. A discussion on the sen2like product format and associated uncertainties will be undertaken. Finally, the sen2like next evolutions and the technical/scientific efforts for integrating a new mission will be addressed.