TY - BOOK

T1 - Improving Usability of Weather Radar Data in Environmental Sciences

T2 - Potentials, Challenges, Uncertainties and Applications

AU - Kreklow, Jennifer

N1 - Doctoral thesis

PY - 2020

Y1 - 2020

N2 - Precipitation is a crucial driver for many environmental processes and exhibits a high spatiotemporal variability. The traditional, widely-used point-scale measurements by rain gauges are not able to detect the spatial rainfall distribution in a comprehensive way. Throughout the last decades, weather radars have emerged as a new measurement technique that is capable of providing areal precipitation information with high spatial and temporal resolution and put precipitation monitoring on a new level. However, radar is an indirect remote sensing technique. Rain rates and distributions are inferred from measured reflectivities, which are subject to a series of potential error sources. In the last years, several operational national radar data archives exceeded a time series length of ten years and several new radar climatology datasets have been derived, which provide largely consistent, well-documented radar quantitative precipitation estimate (QPE) products and open up new climatological application fields for radar data. However, beside uncertainties regarding data quality and precipitation quantification, several technical barriers exist that can prevent potential users from working with radar data. Challenges include for instance different proprietary data formats, the processing of large data volumes and a scarcity of easy-to-use and free-of-charge software, additional effort for data quality evaluation and difficulties concerning data georeferencing. This thesis provides a contribution to improve the usability of radar-based QPE products, to raise awareness on their potentials and uncertainties and to bridge the gap between the radar community and other scientific disciplines which are still rather reluctant to use these highly resolved data. First, a GIS-compatible Python package was developed to facilitate weather radar data processing. The package uses an efficient workflow based on widely used tools and data structures to automate raw data processing and data clipping for the operational German radar-based and gauge-adjusted QPE called RADOLAN (“RADar OnLine Aneichung”) and the reanalysed radar climatology dataset named RADKLIM. Moreover, the package provides functions for temporal aggregation, heavy rainfall detection and data exchange with ArcGIS. The Python package was published as an Open Source Software called radproc. It was used as a basis for all subsequent analyses conducted in this study and has already been applied successfully by several scientific working groups and students conducting heavy rainfall analysis and data aggregation tasks. Second, this study explored the development, uncertainties and potentials of the hourly RADOLAN and RADKLIM QPE products in comparison to ground-truth rain gauge data. Results revealed that both QPE products tend to underestimate total precipitation sums and particularly high intensity rainfall. However, the analyses also showed significant improvements throughout the RADOLAN time series as well as major advances through the climatologic reanalysis regarding the correction of typical radar artefacts, orographic and winter precipitation and range-dependent attenuation. The applicability of the evaluation results was underpinned by the publication of a rainfall inter-comparison geodataset for the RADOLAN, RADKLIM and rain gauge datasets. The intercomparison dataset is a collection of precipitation statistics and several parameters that can potentially influence radar data quality. It allows for a straightforward comparison and analysis of the different precipitation datasets and can support a user’s decision on which dataset is best suited for the given application and study area. The data processing workflow for the derivation of the intercomparison dataset is described in detail and can serve as a guideline for individual data processing tasks and as a case study for the application of the radproc library. Finally, in a case study on radar composite data application for rainfall erosivity estimation, RADKLIM data with a 5-minute temporal resolution were used alongside rain gauge data to compare different erosivity estimation methods used in erosion control practice. The aim was to assess the impacts of methodology, climate change and input data resolution, quality and spatial extent on the R-factor of the Universal Soil Loss Equation (USLE). Moreover, correction factors proposed in other studies were tested with regard to their ability to compensate for different temporal resolutions of rainfall input data and the underestimation of precipitation by radar data. The results clearly showed that R-factors have increased significantly due to climate change and that current R-factor maps need to be updated by using more recent and spatially distributed rainfall data. The radar climatology data showed a high potential to improve rainfall erosivity estimations, but also a certain bias in the spatial distribution of the R-factor due to the rather short time series and a few radar artefacts. The application of correction factors to compensate for the underestimation of the radar led to an improvement of the results, but a possible overcorrection could not be excluded, which indicated a need for further research on data correction approaches. This thesis concludes with a discussion of the role of open source software, open data and of the implementation of the FAIR (Findable, Accessible, Interoperable, Re-usable) principles for the German radar QPE products in order to improve data usability. Finally, practical recommendations on how to approach the assessment of QPE quality in a specific study area are provided and potential future research developments are pointed out.

AB - Precipitation is a crucial driver for many environmental processes and exhibits a high spatiotemporal variability. The traditional, widely-used point-scale measurements by rain gauges are not able to detect the spatial rainfall distribution in a comprehensive way. Throughout the last decades, weather radars have emerged as a new measurement technique that is capable of providing areal precipitation information with high spatial and temporal resolution and put precipitation monitoring on a new level. However, radar is an indirect remote sensing technique. Rain rates and distributions are inferred from measured reflectivities, which are subject to a series of potential error sources. In the last years, several operational national radar data archives exceeded a time series length of ten years and several new radar climatology datasets have been derived, which provide largely consistent, well-documented radar quantitative precipitation estimate (QPE) products and open up new climatological application fields for radar data. However, beside uncertainties regarding data quality and precipitation quantification, several technical barriers exist that can prevent potential users from working with radar data. Challenges include for instance different proprietary data formats, the processing of large data volumes and a scarcity of easy-to-use and free-of-charge software, additional effort for data quality evaluation and difficulties concerning data georeferencing. This thesis provides a contribution to improve the usability of radar-based QPE products, to raise awareness on their potentials and uncertainties and to bridge the gap between the radar community and other scientific disciplines which are still rather reluctant to use these highly resolved data. First, a GIS-compatible Python package was developed to facilitate weather radar data processing. The package uses an efficient workflow based on widely used tools and data structures to automate raw data processing and data clipping for the operational German radar-based and gauge-adjusted QPE called RADOLAN (“RADar OnLine Aneichung”) and the reanalysed radar climatology dataset named RADKLIM. Moreover, the package provides functions for temporal aggregation, heavy rainfall detection and data exchange with ArcGIS. The Python package was published as an Open Source Software called radproc. It was used as a basis for all subsequent analyses conducted in this study and has already been applied successfully by several scientific working groups and students conducting heavy rainfall analysis and data aggregation tasks. Second, this study explored the development, uncertainties and potentials of the hourly RADOLAN and RADKLIM QPE products in comparison to ground-truth rain gauge data. Results revealed that both QPE products tend to underestimate total precipitation sums and particularly high intensity rainfall. However, the analyses also showed significant improvements throughout the RADOLAN time series as well as major advances through the climatologic reanalysis regarding the correction of typical radar artefacts, orographic and winter precipitation and range-dependent attenuation. The applicability of the evaluation results was underpinned by the publication of a rainfall inter-comparison geodataset for the RADOLAN, RADKLIM and rain gauge datasets. The intercomparison dataset is a collection of precipitation statistics and several parameters that can potentially influence radar data quality. It allows for a straightforward comparison and analysis of the different precipitation datasets and can support a user’s decision on which dataset is best suited for the given application and study area. The data processing workflow for the derivation of the intercomparison dataset is described in detail and can serve as a guideline for individual data processing tasks and as a case study for the application of the radproc library. Finally, in a case study on radar composite data application for rainfall erosivity estimation, RADKLIM data with a 5-minute temporal resolution were used alongside rain gauge data to compare different erosivity estimation methods used in erosion control practice. The aim was to assess the impacts of methodology, climate change and input data resolution, quality and spatial extent on the R-factor of the Universal Soil Loss Equation (USLE). Moreover, correction factors proposed in other studies were tested with regard to their ability to compensate for different temporal resolutions of rainfall input data and the underestimation of precipitation by radar data. The results clearly showed that R-factors have increased significantly due to climate change and that current R-factor maps need to be updated by using more recent and spatially distributed rainfall data. The radar climatology data showed a high potential to improve rainfall erosivity estimations, but also a certain bias in the spatial distribution of the R-factor due to the rather short time series and a few radar artefacts. The application of correction factors to compensate for the underestimation of the radar led to an improvement of the results, but a possible overcorrection could not be excluded, which indicated a need for further research on data correction approaches. This thesis concludes with a discussion of the role of open source software, open data and of the implementation of the FAIR (Findable, Accessible, Interoperable, Re-usable) principles for the German radar QPE products in order to improve data usability. Finally, practical recommendations on how to approach the assessment of QPE quality in a specific study area are provided and potential future research developments are pointed out.

U2 - 10.15488/10144

DO - 10.15488/10144

M3 - Doctoral thesis

CY - Hannover

ER -