Precipitation phase discrimination: diagnosing and nowcasting
[eng] Precipitation phase discrimination at ground level constitutes a fundamental variable in many meteorological and hydrological applications, including avalanche hazards, winter road safety, and flooding from rain on snow events. Discrimination of the precipitation phase at surface level has bee...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2022 |
| País: | España |
| Institución: | Universidad de Barcelona |
| Repositorio: | Dipòsit Digital de la UB |
| OAI Identifier: | oai:diposit.ub.edu:2445/186853 |
| Acceso en línea: | https://hdl.handle.net/2445/186853 http://hdl.handle.net/10803/674609 |
| Access Level: | acceso abierto |
| Palabra clave: | Meteorologia Previsió del temps Radar Precipitacions (Meteorologia) Neu Interpolació (Matemàtica) Meteorology Weather forecasting Precipitations (Meteorology) Snow Interpolation |
| Sumario: | [eng] Precipitation phase discrimination at ground level constitutes a fundamental variable in many meteorological and hydrological applications, including avalanche hazards, winter road safety, and flooding from rain on snow events. Discrimination of the precipitation phase at surface level has been widely studied following different approaches ranging from decision tree algorithms based on vertical temperature profiles parameters, to machine learning algorithms through schemes relying on microphysical parameterisations. However, precipitation phase discrimination is still challenging, specially at temperature close to freezing point. Several studies pointed out research gaps regarding this topic and the present thesis aims to make its small contribution to some of them. In addition, this thesis comes from the need to provide the Meteorological Service of Catalonia with an adjusted and verified precipitation phase discrimination product for diagnosing and nowcasting purposes. In order to achieve both kind of requirements six specific objectives were set and upon which this thesis was structured. These are the following: • SO1. Obtention of a dynamic interpolation scheme suitable for complex terrain, and high spatial and temporal resolution. • SO2. Evaluation and adjustment of different schemes and meteorological variables to diagnose discrimination of the surface precipitation phase. • SO3. Assessment of citizen science and crowd sourced observations for monitoring snow events. • SO4. Development and evaluation of different schemes to nowcast discrimination of the precipitation phase. • SO5. Evaluation of ensemble techniques to nowcast discrimination of the precipitation phase. • SO6. Implementation of a precipitation phase product in an operational chain. The present thesis is based in a compendium of three scientific publications and three major blocks were defined following each publication. The storyline of the thesis is first based on obtaining spatial surface information from point meteorological observations. Then, the spatial information is used to estimate precipitation phase for diagnosing purposes. And finally, include extrapolation techniques and numerical weather prediction models to nowcast the precipitation phase with a forecast lead time of 180 minutes. The first block of the thesis presents a methodology to interpolate high spatially and temporally resolved meteorological observations. Interpolation techniques have been widely studied and verified for daily and monthly observations, but limited for hourly or sub-hourly time scales. At these scales, observations tend to be more irregular and present higher variability as they are influenced by weather conditions, such as the presence of fog banks or thermal inversions. For this reason, an adaptive interpolation system was proposed. It is based on the combination of three elements: clustering, multiple linear regression, and residual correction. Meteorological observations are first divided in several clusters of variable size to separate areas prone to be affected by different weather conditions. A multiple linear regression is calculated for each cluster and then compared against an MLR that considers all data. It is in this step where the proposed system plays its role. The system, based on interpolation errors, decides which MLR uses in each cluster: that calculated using the stations of the cluster only or that using all stations available. The adaptive character of the system lays on using different number of clusters and test all them every time an interpolation is conducted. The system was successfully applied in three European regions, and results indicate a reduction of RMSE when the proposed interpolation system is used compared to using a single MLR considering all stations. Once the step to interpolate point meteorological observations is achieved, the thesis focuses on discrimination of the precipitation phase in the following two blocks. The second block evaluates different precipitation phase discrimination schemes based on surface observations for diagnosing purposes. These schemes set thresholds on meteorological variables upon which precipitation is classified as rain, mixed or snow. In order to perform the evaluation of the schemes around 7700 quality-controlled observations of precipitation phase were gathered from different sources concerning Catalonia. According to the verification results, the schemes including air saturation conditions perform best, that is wet bulb temperature or combining air temperature with relative humidity. When analysing the schemes for specific snowfall events, a certain variability among the optimum thresholds was identified. This lead to suggest a range of thresholds when monitoring snowfall events. In addition, apart from the quality-controlled observations, citizen science and crowd sourced observations were also collected and evaluated showing both advantages and limitations. The third block of the thesis is also focused on precipitation phase determination, but in this case for nowcasting purposes. Apart from considering surface precipitation phase discrimination schemes, algorithms based on vertical temperature profiles, which play a key role on determining precipitation phase at ground level, were also considered. According to the threshold and performance variability observed when diagnosing precipitation phase and based on previous studies, combinations of algorithms were also taken into account in this block. The performance of the different algorithms and their combinations was assessed in eight low-altitude snowfall events reported in Catalonia between 2010 and 2021. Verification results showed that a combination of algorithms is preferable as it may provide a wide perspective to forecasters during precipitation phase transitions. In addition, this block included the implementation of a probabilistic methodology to nowcast the precipitation field. The results obtained in the present thesis allowed to adjust and improve the real- time precipitation phase discrimination at the Meteorological Service of Catalonia. In addition, a nowcasting of precipitation phase product was also developed and operationally implemented. The results may also contribute to add a new verification dataset for precipitation phase discrimination purposes, together with the evaluation of precipitation phase schemes with interpolated meteorological variables and the development of spatially resolved products. |
|---|