A Linear Model for Orographic Precipitation in meteorological and climatological downscaling
Not peer reviewed
MetadataShow full item record
Orographic precipitation has always been a major field of study in atmospheric sciences, because of its major role in the water budget and its influence on environmental hazards like floods and droughts - and thus on human activity. Understanding its features has become even more important with the rise of the challenges posed by the changing climate. In the last two decades, several models have been developed to advance orographic precipitation science.
This thesis used mainly one of these models, Smith and Barstad’s Linear Model for orographic precipitation, in high space-resolution studies for downscaling and model validation, both on the meteorological and on the climatological time scales. The problem of model validation and precipitation downscaling was introduced in the first paper, while the Linear Model itself was used in downscaling in three other papers: in one instance to downscale 3- and 6-hour meteorological forecasts and reanalysis down to 250 meters of space resolution; in one case, to downscale daily climate projections of 30 years of data down to about 1 km space resolution in western Norway; and lastly, to provide orographic correction at 1 km resolution to a spatially homogeneous statistical downscaling model.
The first paper dealt with ENSEMBLES model validation at the 25 km scale. The validation was performed as part of CLIMB, an EU FP-7 project that studied climate changes at selected Mediterranean hydrological basins. The control time was 1951-2010. Model validation allowed to select four models in the ENSEMBLES dataset for use in later stages of the project; however, it also showed that 25 km resolution was too coarse a resolution to properly resolve physical phenomena that lead to orographic precipitation, and pointed at the need for proper downscaling of precipitation and temperature data.
Starting from this, this thesis introduced the Linear Model as a tool for climatological downscaling. For climate projections, the Linear Model’s low CPU demand allowed the production of many simulations to span the uncertainties over the whole model range of the projects taken into account. The four ENSEMBLES models selected in the first paper were used for the downscaling performed in Sardinia for the 1981-1990 time period in the fourth paper. In the second paper, fourteen IPCC AR4 models were used in Norway, for the 1971-2000 control period, and the 2046-2065 and 2071-2100 future assessements.
The results showed that when large-scale background precipitation is taken into account, the Linear Model was able to compare well with simulated and observational data both for Norway and for Sardinia, with promising results in i) reducing errors of past reconstruction and ii) producing reasonable climate assessments - in both instances at a high spatial resolution. The results at past control times showed that LM was able to provide a significant spatial separation of real data stations. In IPCC AR-4 downscaling in Norway, where large-scale precipitation was negligible compared to orographic precipitation, the Linear Model was able to reproduce past climatology and provide a future assessment as a stand-alone tool. In Sardinia, the Linear Model was used in conjunction with a statistically homogeneous Multifractal model.
The use of a two-models downscaling method for Sardinia allowed to circumvent one of the Linear Model’s main limitations, i.e. the fact that it is unfit for use on its own in warmer climates, where orographic forcing is just one of the components that lead to precipitation and thermal convection can play an important role. In regard to reproducing observed features of local precipitation, the Linear Model orographic correction used with the Multifractal model compared very well with a locally-calibrated orographic modulation of the same statistical model. This happened at mean areal precipitation level, but also at individual stations, for yearly means, and for monthly and daily distributions of precipitation events.
The third paper dealt with meteorological time scales, comparing three days of observations with WRF simulations, and LM downscaling of WRF data was used at different time and space resolutions over the Stord island in western Norway. In this study, the Linear Model compared well with a more complex model like WRF, in reproducing high resolution precipitation events. This was especially true at the higher spatial (250 m) and temporal (3 hours) resolutions used.
To sum up the findings of this thesis, we tested the boundaries of the Linear Model, already proven in literature to be a powerful tool for describing orographic precipitation. We showed that, with some care, it could be used both as a stand-alone model or an accompanying support for climatological downscaling studies, and to refine or better understand investigations on precipitation carried out through meteorological methods. Good results in the use of the Linear Model for climatological studies in a high-latitude area like Norway were somewhat expected, given the intrinsic strong points of the model. However, the robustness of the Linear Model’s underlying physics was put succesfully under stress in the third paper, where the model performed well also at hourly time resolution at meteorological time scales. Last but not least, the fourth paper showed another way to exploit the Linear Model’s features for climatological downscaling in warmer, mid-latitude regions.