Aim of WP1 is todevelop and apply data-driven methods based on machine learning to study the impact of environmental factors governing soil moisture patterns over time at various spatial scales and to model long-term, spatially continuous maps of soil moisture and precipitation at kilometric resolution and sub-daily revisit time.
In WP1, we will build robust models by merging multi-source datasets from coarse resolution remotely sensed products (~40 km resolution, sub-daily revisit time) and high-resolution land surface parameters, which will be trained against in-situ observations (traditional measurements and crowdsourced data). With such a unique and unprecedented dataset, we will investigate dependencies and interactions among factors contributing to the variability of hydrological variables at different scales. The downscaled datasets will be validated against in-situ observations from the four WATERLINE test sites and other freely available datasets (e.g. ISMN).
The downscaled datasets produced within the WP will serve as a forcing/reference dataset for WP2 and WP3.
WP2 aims to define a reference architecture of Waterline platform that enables the partners to:
In WP2, we will use the nature, type and constraintsto define the subsequent calibrating and validating approach required for each case study as well as the way to approach the real-time constraint utilizing inputs from WP1,WP4 and WP5.We will usethe data assimilation frameworkto enable the various partners to leverage sensory data from existing state-of-the-art and newly developed hydrological measurement systemsat different temporal and spatial scales paving the way for cross-fertilization and transfer learning-based reasoning that compensates for discrimination in data resolution and inherent limitations of measurement system by input from another site studyor measurement modality.
This expects to generate new framework for quantifying and handling various facets of uncertainty pervading both individual measurement system and hydrological model. In short,WP2 will deliver a framework and a toolkit in the form of hybrid-cloud architecture that can run on organization’s application and workloads easingthe communication between the various partners and enablingindividual partner to reuse, rescale and simulate outcomes obtained in another sitefor the purpose of data assimilation and uncertainty handling.This activity will provide input to WP3 where individual hydrological models will be constantly re-shaped, while enabling stakeholder to efficiently interact with research community as part of WP4 and WP5 tasks.
WP3 aims to develop predictive modelsbased on procedures for data input, calculations and output.WP3 willgenerate the data required to train machine learning approaches which are subsequently used for near-real time predictions of the system. Numerical models of different structures and complexity will be developed and will be forced with a wide range of climatic forcing functions such as different precipitation or land-use scenarios. The simplest models employed will be conceptual hydrological models, followed by established land-surface and subsurface hydrological models (e.g. SWAT, HYDRUS, MODFLOW), and finally fully integrated surface water-groundwater models (HGS) that simulate all relevant hydrological processes at a catchment scale in a spatially distributed manner. While different in structure and complexity, the models will provide the same type of predictions, e.g. storage in the subsurface or outflow of a catchment. This unique and systematic combination of different model types with machine-learning will allow identifying the appropriate level of model complexity for project areas other than simulated within the context of this proposal.
To serve the near real time prediction requirements that complex models are difficult to accommodate, model generated datasets will be used to train machine-learning algorithms to predict the same hydrological variables. We will generate high-resolution and reliable soil moisture and precipitation data, which along with other data sources (additional satellite products, historical archives, in-situ monitoring, crowdsourced observations) will improve the robustness of hydrological predictions. Handling large datasets through cloud computing and advanced filtering techniques, determining uncertainty in various modelling processes and employing parallel computing, will offer an innovative and versatile platform for environmental management problems.
WP4 aims to develop solutions that are useful for key project stakeholders'groups. We will engage stakeholders based on their interest and willingness to participate to the project, ranging from farmers, companies developing of monitoring solutions todecision makers in land and water management. From past projects we have a good network of stakeholders which will be updated before the project starts.
Specific objectives are to:
The objective of WP5 is to develop middleware services for distributed execution of WATERLINE models, orchestration of application workflows, advanced model visualization, and to develop web-based front-ends for WATERLINE case studies. The services developed in WP5 will integrate the WATERLINE platform into a working system by orchestrating the WATERLINE components into workflows that realize specific application scenarios defined by the case studies. WP5 will also study the hybrid cloud deployment model with cloud bursting to support demanding real-time scenarios. In this model, resources from public clouds are temporarily used to handle computational workloads for which the local computing resources do not have the sufficient capacity.Finally, advanced AR/VR framework and applications, in an effort to provide an impressive but also highly effective way to present the importance of hydrological data to the stakeholders and to general public will be developed.
WP5 will interact with other WPs as follows:
The main aim of WP6 is tointegrate and disseminate project results from WP 1 – 5 to various target groups and increase its added value and impact.
Specific objectives include:
Moreover, our team will use communication experts (e.g. journalists, head of communication) present at different institutes to provide user friendly outputs, to become easily absorbed by the various groups of stakeholders.Additionally, in WP1-4, we will write at least 8 refereed papers (at least two for each RO as described in different work packages) within the scientific WPs).
Ensure high quality and efficient running of the project tasks, manage scientific and financial issues, data and intellectual property, GDPR and ensure gender balance and provision of equal opportunities to all.
Specific objectives include: