Sensitivity Analysis of the Spatial Structure of Forecasts in Mesoscale Models: Continuous Model Parameters (original) (raw)
Related papers
Monthly Weather Review, 2020
In a recent work, a sensitivity analysis methodology was described that allows for a visual display of forecast sensitivity, with respect to model parameters, across a gridded forecast field. In that approach, sensitivity was assessed with respect to model parameters that are continuous in nature. Here, the analogous methodology is developed for situations involving noncontinuous (discrete or categorical) model parameters. The method is variance based, and the variances are estimated via a random-effects model based on 2 k2p fractional factorial designs and Graeco-Latin square designs. The development is guided by its application to model parameters in the stochastic kinetic energy backscatter scheme (SKEBS), which control perturbations at unresolved, subgrid scales. In addition to the SKEBS parameters, the effect of daily variability and replication (both, discrete factors) are also examined. The forecasts examined are for precipitation, temperature, and wind speed. In this particular application, it is found that the model parameters have a much weaker effect on the forecasts as compared to the effect of daily variability and replication, and that sensitivities, weak or strong, often have a distinctive spatial structure that reflects underlying topography and/or weather patterns. These findings caution against fine-tuning methods that disregard 1) sources of variability other than those due to model parameters, and 2) spatial structure in the forecasts.
A Sensitivity Analysis of Two Mesoscale Models: COAMPS and WRF
Monthly Weather Review, 2020
A sensitivity analysis methodology recently developed by the authors is applied to COAMPS and WRF. The method involves varying model parameters according to Latin Hypercube Sampling, and developing multivariate multiple regression models that map the model parameters to forecasts over a spatial domain. The regression coefficients and p values testing whether the coefficients are zero serve as measures of sensitivity of forecasts with respect to model parameters. Nine model parameters are selected from COAMPS and WRF, and their impact is examined on nine forecast quantities (water vapor, convective and gridscale precipitation, and air temperature and wind speed at three altitudes). Although the conclusions depend on the model parameters and specific forecast quantities, it is shown that sensitivity to model parameters is often accompanied by nontrivial spatial structure, which itself depends on the underlying forecast model (i.e., COAMPS vs WRF). One specific difference between these models is in their sensitivity with respect to a parameter that controls temperature increments in the Kain-Fritsch trigger function; whereas this parameter has a distinct spatial structure in COAMPS, that structure is completely absent in WRF. The differences between COAMPS and WRF also extend to the quality of the statistical models used to assess sensitivity; specifically, the differences are largest over the waters off the southeastern coast of the United States. The implication of these findings is twofold: not only is the spatial structure of sensitivities different between COAMPS and WRF, the underlying relationship between the model parameters and the forecasts is also different between the two models.
Meteorological Applications, 2019
Stochastic Kinetic Energy Backscatter Schemes (SKEBS) are introduced in numerical weather forecast models to represent uncertainties related to unresolved subgrid-scale processes. These schemes are formulated using a set of parameters that must be determined using physical knowledge and/or to obtain a desired outcome. Here, a methodology is developed for assessing the effect of four factors on spatial features of forecasts simulated by the SKEBS-enabled Weather Research and Forecasting (WRF) model. The four factors include two physically motivated SKEBS parameters (determining amplitude of perturbations applied to streamfunction and potential temperature tendencies), a purely stochastic element (a seed used in generating random perturbations), and a factor reflecting daily variability. A simple threshold-based approach for identifying coherent objects within forecast fields is employed, and the effect of the four factors on object features (e.g., number, size, and intensity) is assessed. Four object types are examined: upper-air jet streaks, low-level jets, precipitation areas, and frontal boundaries. The proposed method consists of a set of standard techniques in experimental design, based on the analysis of variance, tailored to sensitivity analysis. More specifically, a Latin Square Design is employed to reduce the number of model simulations necessary for performing the sensitivity analysis. Fixed effects and random effects models are employed to assess the main effects and the percentage of the total variability explained by the four factors. It is found that the two SKEBS parameters do not have an appreciable and/or statistically significant effect on any of the examined object features.
Natural Hazards and Earth System Sciences, 2011
The Valencia region, on the Mediterranean coast of the Iberian Peninsula, is propitious to heavy precipitation, especially the area encompassing the South of Valencia province and the North of the Alicante province. In October 2007 a torrential rain affected the aforementioned area, producing accumulated rainfall values greater than 400 mm in less than 24 h and flash-floods that caused extensive economic losses and human casualties. This rain event has been studied in numerical experiments using the Regional Atmospheric Modeling System. The present paper deals with the effect of using the different convective parameterizations (CP) currently implemented in the Regional Atmospheric Modeling System (Kuo and Kain-Fritsch) in the forecast results, in particular on precipitation forecast. Sensitivity tests have been run with and without these parameterizations activated in a series of combinations of the different grids. Results are very different depending on the model convective parameterization setting. A statistical verification has also been undertaken by calculating different skill scores for each simulation in the experiment.
Calibrated Probabilistic Mesoscale Weather Field Forecasting: The Geostatistical Output Perturbation
Probabilistic weather forecasting consists of finding a joint probability distribution for future weather quantities or events. It is typically done by using a numerical weather prediction model, perturbing the inputs to the model in various ways, often depending on data assimilation, and running the model for each perturbed set of inputs. The result is then viewed as an ensemble of forecasts, taken to be a sample from the joint probability distribution of the future weather quantities of interest. This is typically not feasible for mesoscale weather prediction carried out locally by organizations without the vast data and computing resources of national weather centers. Instead, we propose a simpler method which breaks with much previous practice by perturbing the outputs, or deterministic forecasts, from the model. Forecast errors are modeled using a geostatistical model, and ensemble members are generated by simulating realizations of the geostatistical model. The method is applied to 48-hour mesoscale forecasts of temperature in the US Pacific Northwest in 2000 and 2002. The resulting forecast intervals turn out to be well calibrated for individual meteorological quantities, to be sharper than those obtained from approximate climatology, and to be consistent with aspects of the spatial correlation structure of the observations.
Journal of the American Statistical Association, 2004
Probabilistic weather forecasting consists of finding a joint probability distribution for future weather quantities or events. It is typically done by using a numerical weather prediction model, perturbing the inputs to the model in various ways, often depending on data assimilation, and running the model for each perturbed set of inputs. The result is then viewed as an ensemble of forecasts, taken to be a sample from the joint probability distribution of the future weather quantities of interest. This is typically not feasible for mesoscale weather prediction carried out locally by organizations without the vast data and computing resources of national weather centers. Instead, we propose a simpler method which breaks with much previous practice by perturbing the outputs, or deterministic forecasts, from the model. Forecast errors are modeled using a geostatistical model, and ensemble members are generated by simulating realizations of the geostatistical model. The method is applied to 48-hour mesoscale forecasts of temperature in the US Pacific Northwest in 2000 and 2002. The resulting forecast intervals turn out to be well calibrated for individual meteorological quantities, to be sharper than those obtained from approximate climatology, and to be consistent with aspects of the spatial correlation structure of the observations.
Journal of Aerospace Technology and Management, 2015
The sensitivity of the planetary boundary layer (PBL) parameterizations is investigated using the hybrid Fifth Generation Penn State University/National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5) and its results were compared with observations from two field campaigns held during the dry and wet seasons at the Centro de Lançamento de Alcântara (CLA). The comparisons were made using the integrated zonal and meridional components of observed and forecasted winds. Initially, three boundary layer parameterizations, in addition to the current parameterization, were selected for evaluation: Blackadar (BLK), Medium Range Forecast (MRF), Janjic (ETA) and Burk-Thompson (BT). The MRF and BLK schemes produced better results than the ETA and BT schemes. Nevertheless, MRF and BLK underestimate the zonal and meridional wind components by around 16% in the rainy season and overestimate them by on average 18% in the dry season.
Ciencia e investigación agraria, 2009
D.O. Silva, F.J. Meza, and E. Varas. 2009. Use of mesoscale model MM5 forecasts as proxies for surface meteorological and agroclimatic variables. Cien. Inv. Agr. 36(3):369-380. There is increasing interest in meteorological information and its application to strategic planning at the farm as well as regional level. Although we have recently seen signifi cant improvements to strengthen and enlarge networks of weather observations, their density is still insuffi cient to cover large extents at the desired spatial and temporal resolution. Climate scientists have developed and used mesoscale models to understand and predict future atmospheric conditions. These models represent a major contribution to objective weather forecasts throughout numerical simulations. They use global circulation outputs as boundary conditions and can be run in a nested manner so as to increase their spatial resolution. Because of this, we can obtain information about weather variables in grid cells spaced 15 km apart covering important areas and providing information in places where analog or automatic stations are not available. The objective of this work is to evaluate the use of raw data from the MM5 mesoscale model as well as MOS-corrected information (a statistical post-processing of MM5 outputs) as a proxy for surface meteorological data. Temperature, wind speed, relative humidity, and daily solar radiation forecasts were evaluated for eleven stations in the Maipo river basin. In all cases, the MOS forecast produced better results than the raw MM5 data. Determination coeffi cients reached values near 0.9, and the RMSE was usually smaller for MOS-corrected data. The small variability of the MOS parameters allows their use as regional values to estimate meteorological data for the whole region, particularly at a weekly time step.