Periodic Reporting for period 1 - IMPASTO (Inverse Modeling of PArameterized physics STochastic uncertainty using process-level Observations)
Reporting period: 2020-05-01 to 2022-04-30
Seasonal forecast and different climate services that can be delivered using the forecast results at regional and local scales are important for management and planning in many industries and services such as agriculture, water resources management, energy production and distribution, transport, health, disaster risk reduction, etc. On of the key element for successful application of the prediction data for these purposes relies on understanding and quantification of prediction uncertainty.
Within the IMPASTO project stochastic representations of modeling uncertainties associated with physical parameterizations in a regional seasonal forecast system and their impacts on seasonal climate ensemble prediction at local to regional scales were addressed. IMPASTO was organized around two main objectives
1. Implementation of stochastic parameterization schemes in a regional seasonal prediction system, and examination of their impacts on the seasonal climate prediction uncertainty at local to regional scales. The stochastic parameterization schemes considered are the Stochastically Perturbed Parameterization Tendency (SPPT) and Stochastically Perturbed Parameters (SPP).
2. Investigation of a new approach to representing the stochastic uncertainty of physical parameterizations at process-level based on inverse modeling with process-sensitive observations. The approach is based on applying the estimates of statistical distributions of physical parameters within a physical parameterization that are constrained by process-sensitive observations into the SPP scheme.
The first objective was addressed by implementing the SPPT scheme in the Nonhydrostatic Multiscale Model on B-grid (NMMB model) provided by the Faculty of Physics University in Belgrade (FPUB) and performing the ensemble sensitivity experiments at the seasonal climate time scales. This model was used for the dynamical downscaling of the ECMWF's SEAS5 seasonal forecast. The second objective was addressed by investigating sensitivity of cloud and precipitation microphysics parameterization to different configurations of the SPP scheme informed by results of nonlinear estimation of parameter distributions using microphysics-sensitive simulated satellite observations.
Within the IMPASTO project stochastic representations of modeling uncertainties associated with physical parameterizations in a regional seasonal forecast system and their impacts on seasonal climate ensemble prediction at local to regional scales were addressed. IMPASTO was organized around two main objectives
1. Implementation of stochastic parameterization schemes in a regional seasonal prediction system, and examination of their impacts on the seasonal climate prediction uncertainty at local to regional scales. The stochastic parameterization schemes considered are the Stochastically Perturbed Parameterization Tendency (SPPT) and Stochastically Perturbed Parameters (SPP).
2. Investigation of a new approach to representing the stochastic uncertainty of physical parameterizations at process-level based on inverse modeling with process-sensitive observations. The approach is based on applying the estimates of statistical distributions of physical parameters within a physical parameterization that are constrained by process-sensitive observations into the SPP scheme.
The first objective was addressed by implementing the SPPT scheme in the Nonhydrostatic Multiscale Model on B-grid (NMMB model) provided by the Faculty of Physics University in Belgrade (FPUB) and performing the ensemble sensitivity experiments at the seasonal climate time scales. This model was used for the dynamical downscaling of the ECMWF's SEAS5 seasonal forecast. The second objective was addressed by investigating sensitivity of cloud and precipitation microphysics parameterization to different configurations of the SPP scheme informed by results of nonlinear estimation of parameter distributions using microphysics-sensitive simulated satellite observations.
During the IMPASTO project Faculty of Physics University in Belgrade (FPUB) provided to the researcher, scientific and technical training on the seasonal climate modeling and the numerical modeling system based on the NMMB model. Faculty of Physics also provided support to setup the necessary infrastructure on computer which was purchased from the project funds including: operating system setup, compiler and other specific software setup, permanent internet access, model installation and model setup, download of necessary data for model run. Through this training the researcher acquired skills to independently use and modify the NMMB regional seasonal prediction modeling software and to perform numerical experiments with that software.
Faculty of Physics also. due to large volume of output data, provided necessary infrastructure for project results dissemination (specifically result from numerical experiments with NMMB model) by setting up data repository web page http://haos.ff.bg.ac.rs/IMPASTO/(opens in new window) that enables browsing and download of the project results without any restriction.
During the project the researcher applied her expertise into implementing the SPPT scheme in the NMMB modeling system by which the research capability using this model at the FPUB was enhanced. After the successful implementation of the schemes in to the NMMB model, four different experiments were performed, using NNMB model for a seasonal forecast downscaling and ECMWF's SEAS5 for initial and boundary conditions. These four experiments includes one control and three different sensitivity experiments. Results from these experiments are available on project data repository.
On the other hand, SPP scheme was implemented into the 1D column cloud model followed by investigation of impacts of its different configurations for sampling the parameter perturbations informed by the inverse modeling results. The inverse modelling was performed with simulated satellite-based observations. The results from this experimets are available via Zenodo platform and are presented in the paper that is currently in review in scientific journal.
Researcher also, made several presentations for the modeling team at FPUB and their collaborators about the methodology of stochastic uncertainty representations in the ensemble modeling and the utility of inverse parameter estimation.
In addition, the researcher also, taught a 12-lecture seminar course on Data Assimilation and Inverse Modeling to interested senior year undergraduate students in the Fall semester of 2020. This course was an adopted short version of graduate level courses she has taught in the past at several universities in the USA and at CMCC in Italy. After the course was finished, she mentored one of the faculty students master thesis (the student was one of the attendants of the course) at the FPUB.
Overall, the results obtained in this project indicate that the regional seasonal ensemble prediction may benefit from using the SPPT representation of the modeling uncertainty at local scales for some prediction quantities and for some periods. The extent of the impacts may be limited by the one-way interactive boundary conditions which inhibit upscale error propagation. Further investigation with larger domains and for more seasons, as well as using different models would be needed to reach more conclusive results.
The research results were presented at three conferences and in one research paper currently in review in the Journal of Advances in Modeling Earth Systems. She also presented result on seminar in the Hurricane Research Division monthly modelling meeting at the Atlantic Oceanographic and Meteorological Laboratory NOAA in Miami, USA.
Faculty of Physics also. due to large volume of output data, provided necessary infrastructure for project results dissemination (specifically result from numerical experiments with NMMB model) by setting up data repository web page http://haos.ff.bg.ac.rs/IMPASTO/(opens in new window) that enables browsing and download of the project results without any restriction.
During the project the researcher applied her expertise into implementing the SPPT scheme in the NMMB modeling system by which the research capability using this model at the FPUB was enhanced. After the successful implementation of the schemes in to the NMMB model, four different experiments were performed, using NNMB model for a seasonal forecast downscaling and ECMWF's SEAS5 for initial and boundary conditions. These four experiments includes one control and three different sensitivity experiments. Results from these experiments are available on project data repository.
On the other hand, SPP scheme was implemented into the 1D column cloud model followed by investigation of impacts of its different configurations for sampling the parameter perturbations informed by the inverse modeling results. The inverse modelling was performed with simulated satellite-based observations. The results from this experimets are available via Zenodo platform and are presented in the paper that is currently in review in scientific journal.
Researcher also, made several presentations for the modeling team at FPUB and their collaborators about the methodology of stochastic uncertainty representations in the ensemble modeling and the utility of inverse parameter estimation.
In addition, the researcher also, taught a 12-lecture seminar course on Data Assimilation and Inverse Modeling to interested senior year undergraduate students in the Fall semester of 2020. This course was an adopted short version of graduate level courses she has taught in the past at several universities in the USA and at CMCC in Italy. After the course was finished, she mentored one of the faculty students master thesis (the student was one of the attendants of the course) at the FPUB.
Overall, the results obtained in this project indicate that the regional seasonal ensemble prediction may benefit from using the SPPT representation of the modeling uncertainty at local scales for some prediction quantities and for some periods. The extent of the impacts may be limited by the one-way interactive boundary conditions which inhibit upscale error propagation. Further investigation with larger domains and for more seasons, as well as using different models would be needed to reach more conclusive results.
The research results were presented at three conferences and in one research paper currently in review in the Journal of Advances in Modeling Earth Systems. She also presented result on seminar in the Hurricane Research Division monthly modelling meeting at the Atlantic Oceanographic and Meteorological Laboratory NOAA in Miami, USA.
This project investigated for the first time a potential for enhancing the quantification of the regional seasonal climate prediction uncertainty by means of ensemble prediction that includes stochastic representations of the modeling errors. The new knowledge gained in this project is expected to lead to further R&I in this research domain, toward improving the utility of the seasonal climate prediction at local to regional scales for societal benefits linked to management and planning in industry and service sectors such as transport, agriculture, energy, water resources, health and emergency response. The project also expanded technical research capabilities for continued research in this domain by developing new research tools at the FPUB and by demonstration of the novel methodology applicable to any regional modeling system.