Periodic Reporting for period 3 - MULTIPLY (MULTIscale SENTINEL land surface information retrieval PLatform)
Reporting period: 2018-07-01 to 2019-12-31
With the start of the SENTINEL era, an unprecedented amount of Earth Observation (EO) data has become available. Currently there is no consistent but extendible and adaptable framework to integrate observations from different sensors in order to obtain the best possible estimate of the land surface state. MULTIPY proposes a solution to this challenge. The project will develop an efficient and fully traceable platform that uses state-of-the-art physical radiative transfer models, within advanced data assimilation (DA) concepts, to consistently acquire, interpret and produce a continuous stream of high spatial and temporal resolution estimates of land surface parameters, fully characterized. These inferences on the state of the land surface will be the result from the coherent joint interpretation of the observations from the different Sentinels, as well as other 3rd party missions (e.g. ProbaV, Landsat, MODIS). This implies that optical data are linked to passive microwave data to help better constrain land surface variables that these domains have in become. Moreover, coarse resolution data (e.g. from MODIS) help to constrain high resolution SENTINEL data, achieving information-rich retrieval that is consistent with all data sources. To achieve consistently, also a generic atmospheric correction scheme is developed to retain full consistently from data access to land surface products. Finally, the MULTIPLY platform also allows users to exchange components as plug-ins according to their needs. Altogether, the MULTIPLY platform will pave the way towards services, such as the Copernicus services, based on the best possible estimates of the land surface state.
Now that we reached the end of this project, we have achieved most if not all objectives anticipated at the start of the project. Based on the requirements of key users, we designed a beta version of the MULTIPLY platform that allowed retrieval of land surface variables from satellite data from the optical domain. Based on the user requirements, major investments were made to achieve a near real-time retrieval by the use of emulators and an improved optimisation routine (to achieve the best solution given the data and their uncertainties). The platform was made modular and open access to comply with the user needs. To better constrain the solutions, a global database has been set up with vegetation characteristics to provide prior information. The beta version was demonstrated and tested during a well-visited workshop with key users. In the remaining period of the project, the consortium worked on further improving the performance of the platform. The implied: a. employing a large validation campaign as time series of land surface variables were missing in other sources, b. developing a generic atmospheric correction scheme that also allowed linking coarse and high resolution data, c. linking optical and short wave data, d. creating posterior routines on fire and functional diversity from land surface variables, e. developing two inference engines, one that works well if you have time series that are processed in combination and one that allows retrieval of individual observations (in combination with prior data), f. providing a graphical user interface to allow those who are not fluent with working in Python to use the platform, and g. developing e-learning materials for those that work in Python. In combination, this provided a multi-purpose generic retrieval platform that is open-access and available on DIAS to enhance the wider use of SENTINELS data.
We have provided a fully generic flexible data retrieval platform for Copernicus services that provides integrated and consistent data products in an easily accessible virtual machine with advanced visualisation tools. The platform is scientifically highly innovative in both in its individual components (such as the generic atmospheric correction routine, the inference engine and the posterior routine for deriving functional diversity from satellite information) as well as the combinations achieved such linking optical data with passive microwave data to connect retrieval of land surface variables, joint retrieval of coarse and high resolution data. Also the architecture of the platform, with its fully flexible modular structure and containing everything from data access till visualisation of land surface variable retrieval is highly innovative.
Our impacts occurred at multiple levels. First of all, we provided retrieval products that already has helped the scientific community and the companies active in space products. For instance, the emulators to speed up calculations have already been used in the FLEX mission, the atmospheric correction routine has been picked up by NASA and so on. Second, by involving key users from the start till the end of the project (from the moment of obtaining their key requirements, to demonstrating the beta-version till training them with e-learning materials in using the MULTIPLY platform), we have raised awareness in the community in the enhanced possibilities of retrieval of land surface features when using our state-of-the-art tools. Moreover, we have run various user demonstrator projects, e.g. we have shown how remote sensing information can be used to better understand the temporal dynamics in crop development (and thus to optimise crop yields). We have provided other satellite operators with an opportunity to cross-calibrate their data to the science-grade Sentinel standards. We have incorporated and validated multiple radiative transfer models using the platform and we have shown how land surface variables derived with the MULTIPLY platform can be used to better constrain global climate-vegetation models. Third, through our philosophy of open-access development, we have created software that can be used by the entire community as it is available on Github, including an e-learning environment for users to be trained in using and adapting the software. Moreover, the MULTIPLY platform is available as virtual machine on DIAS as one of the first validated platforms for retrieval of SENTINELS data. This has allowed for and will continue to stimulate the wider use of SENTINELS data.
Our impacts occurred at multiple levels. First of all, we provided retrieval products that already has helped the scientific community and the companies active in space products. For instance, the emulators to speed up calculations have already been used in the FLEX mission, the atmospheric correction routine has been picked up by NASA and so on. Second, by involving key users from the start till the end of the project (from the moment of obtaining their key requirements, to demonstrating the beta-version till training them with e-learning materials in using the MULTIPLY platform), we have raised awareness in the community in the enhanced possibilities of retrieval of land surface features when using our state-of-the-art tools. Moreover, we have run various user demonstrator projects, e.g. we have shown how remote sensing information can be used to better understand the temporal dynamics in crop development (and thus to optimise crop yields). We have provided other satellite operators with an opportunity to cross-calibrate their data to the science-grade Sentinel standards. We have incorporated and validated multiple radiative transfer models using the platform and we have shown how land surface variables derived with the MULTIPLY platform can be used to better constrain global climate-vegetation models. Third, through our philosophy of open-access development, we have created software that can be used by the entire community as it is available on Github, including an e-learning environment for users to be trained in using and adapting the software. Moreover, the MULTIPLY platform is available as virtual machine on DIAS as one of the first validated platforms for retrieval of SENTINELS data. This has allowed for and will continue to stimulate the wider use of SENTINELS data.