H14D-01 INVITED
The US CLIVAR Working Group on Drought: A Multi-Model Assessment of the Impact of SST Anomalies on Regional Drought
The USCLIVAR working group on drought recently initiated a series of global climate model simulations forced with idealized SST anomaly patterns, designed to address a number of uncertainties regarding the impact of SST forcing and the role of land-atmosphere feedbacks on regional drought. Specific questions that the runs are designed to address include: What are mechanisms that maintain drought across the seasonal cycle and from one year to the next. What is the role of the land? What is the role of the different ocean basins, including the impact of El Nino/Southern Oscillation (ENSO), the Pacific Decadal Oscillation (PDO), the Atlantic Multi-decadal Oscillation (AMO), and warming trends in the global oceans? The runs were done with several global atmospheric models including NASA/NSIPP-1, NCEP/GFS, GFDL/AM2, and NCAR CCM3 and CAM3. In addition, runs were done with the NCEP CFS (coupled atmosphere-ocean) model by employing a novel adjustment technique to nudge the coupled model towards the imposed SST forcing patterns. This talk provides an overview of the experiments and some initial results.
H14D-02 INVITED
Scientific Foundations for Ensemble Precipitation Forecasts for Hydrologic Application in the U.S. National Weather Service
This presentation will begin with a review the functional requirements for ensemble precipitation forecasts and the sources of "raw" atmospheric forecast information that are being used experimentally as input to an Ensemble Pre-Processing (EPP) system that produces the required ensemble precipitation forecasts. The EPP can support use of single-value and/or ensemble "raw" atmospheric forecasts from multiple sources and for different forecast periods ranging from less than one-day up to one-year. These "raw" forecasts often have bias and spread problems that the EPP is designed to fix. The current EPP generates ensemble members in a two-step process. The first step is to specify a set of "events" to be predicted. An event is the amount of precipitation that will occur over a specified time period beginning with a prescribed lead time to the event. The available "raw" atmospheric forecast information is used to estimate a probability distribution for each of these events taking into account the climatology of the events and the forecasts as well as their past joint distribution. The second step is to create a set of sample functions (i.e. ensemble members) whose values preserve this set of forecast distributions. Test results suggest that this approach can preserve temporal scale-dependency in the variability of the ensemble members and in the forecast uncertainty associated with the members. Some of the many science issues that need to be explored, not only to test and improve the EPP algorithms but to make the best possible use of them with the information that is currently available, will be identified.
H14D-03
Consistency in Global Climate Change Model Forecasts of Regional Precipitation Trends
Predictions of the impacts of anthropogenic climate change arising from the emission of radiatively-active chemical constituents typically involves multiple integrations of numerous numerical models to arrive at multi- model ensembles, from which mean/median values and probabilities can be inferred about the response of various components of the climate system. Some responses are considered reliable in as much as the simulated responses show consistency within ensembles and across models. Other responses - particularly at regional levels and for certain parameters such as precipitation - show little inter-model consistency, even in the sign of the projected climate changes. Here we analyze the consistency in global climate change model predictions of regional precipitation responses to anthropogenic-induced climate change. Results indicate that while certain regions show high inter- and intra-model consistency, in other regions the large spread of individual model realizations - each of which is considered an equally-plausible evolution of the actual climate system - precludes using the multi-model ensemble means to make predictions about the response of regional precipitation to anthropogenic forcing. However, we find that in these regions, short-term trends of individual model realizations provide improved skill in predicting their own state by the end of the simulation period, as compared with multi-model ensemble-mean values. In addition, we find that the climate forcing for which this forecast skill becomes relatively large (e.g. correct in 75% of the individual model runs) is equivalent to the anthropogenic climate forcing imposed over the past century. This result suggests that the historical evolution of the observed climate system, vis a vis precipitation changes, may inherently contain information about its own future evolution that can be used to augment model-based climate change projections and provide guidance about the direction of future precipitation changes over the course of the next century.
H14D-04 INVITED
User-Relevant Spatial Methods for Evaluating Gridded Precipitation Forecasts
In recent years, as atmospheric forecast models have moved toward higher resolution, and the need for metrics that correspond to operational needs have increased, a number of new approaches have been developed for evaluation of spatial forecasts (e.g., for forecasts of precipitation). In general, these development efforts are geared toward development of verification approaches that provide more meaningful information about forecast performance than can typically be obtained using traditional verification measures (e.g., RMSE, Brier Skill Score). These approaches are particularly relevant for gridded forecasts of precipitation intensity or amount, which represent contiguous and organized fields. In particular, the new approaches provide diagnostic information that is not provided by the more traditional evaluation methods; this information can feed back into the forecast development process and provide specific information for application in decision-making or hydrologic models. Several different types of spatial and user-relevant verification approaches and methods have been developed or are currently under development. These methods can be broadly categorized as "neighborhood" methods; scale separation methods; feature-based methods; and field methods. Each of these approaches will be described, and examples of their application will be provided. Each method provides different types of information about forecast performance, and many of the approaches can be used together to provide complementary information about forecast quality. Results will be provided from an intercomparison project in which the various methods have been compared.
H14D-05
Evaluation of NWP Precipitation Forecasts for Global Flood Warning
Precipitation forecasts from numerical weather prediction (NWP) models can potentially improve our ability for global flood and landslide warning. In this study, the skills and errors of three NWP precipitation forecast products were analyzed. These forecast products include GEOS5, GDAS and ECMWF, with lead time ranging from 12 hours to 5 days. They were evaluated against the satellite-based, gauge-corrected precipitation estimates, TMPA 3B42, over the land surface as well as the globe. To gain a better perspective, we also evaluated several other satellite-based precipitation products, including GPCP, TMPA 3B42RT, CMORPH and PERSIANN, against TMPA 3B42. Our analysis shows the three NWP forecasts tend to systematically over-estimate global precipitation by approximately 50%. This positive bias does not change much with lead time. In contrast, the satellite-based estimates (GPCP, TMPA, 3B42RT, CMORPH and PERSIANN) have biases mostly less than 20%. In addition, the RMS errors increase with the lead time in NWP forecasts, and in particular for GEOS5, the most increase in RMS errors takes place when the lead time goes from 1 day to 2 days. The RMS errors in the NWP products are also about twice as much as those of the satellite-based products. Further analysis indicates false alarms dominate the errors in the NWP forecasts. Among the NWP products, GEOS5 has slightly better performance than the other two. The implication of these error characteristics on global flood and landslide warning will be discussed.
H14D-06
On Development of a Performance Measure for Extreme Quantitative Precipitation Forecasts Using Data from HMT-2006 in California
Quantitative precipitation forecasts (QPF) are extremely important for many applications, but have remained one of the major challenges in meteorology. Improvements have traditionally been measured using a "threat score," and by this measure annually averaged performance nationally has improved very slowly from ~0.25 to ~0.30 over many years. In addition to the slow rate of improvement, the measure itself is not as effective as needed in cases of extreme precipitation. These, and other drivers, led to the creation of the NOAA Hydrometeorological Testbed (HMT), which was first implemented in California. Out of this effort has emerged a potential new performance measure for QPF in extreme precipitation events associated with land-falling Pacific winter storms. The new method is presented here, along with results from the very wet HMT-2006 field season. The analysis included sites that received up to 100 inches of rainfall that winter and a number of events that produced >5 inches of rain in 24 hours. Working closely with NOAA/NWS providers of the formal QPF for the area, i.e., NCEP's Hydrometeorological Prediction Center (HPC) and the California/Nevada River Forecast Center (CNRFC), a data set of forecasts and verification for 17 sites representative of coastal, inland valley, and mountain conditions, was collected. A methodology was then developed to determine the probability of detection (POD) and false alarm rates (FAR) for events characterized by 1-3 inches, 3-5 inches and >5 inches of rain in 24 hours at each site for forecast lead times of 1, 2, and 3 days. These results include 16 events that had >5 inches of rainfall, of which 2 were predicted to be extreme at 1 day lead time.
H14D-07
Generation of Ensemble Precipitation Forecasts From Single-Value QPF via Mixed-Type Meta-Gaussian Model
In this presentation, we describe generation of ensemble precipitation forecasts from single-value quantitative precipitation forecasts (QPF) via the mixed-type bivariate meta-Gaussian model (Herr and Krzysztofowicz 2005). Because of the intermittent nature of precipitation, it is necessary to model precipitation amount as a mixed variable. The joint distribution of single-value QPF and observed precipitation amounts may then be modeled by the mixed-type bivariate meta-Gaussian distribution. From the single-value QPF, one may generate ensemble precipitation forecasts by sampling from the conditional distribution of the mixed-type bivariate meta-Gaussian distribution. The marginal distributions of the meta-Gaussian distribution are estimated using the Gaussian kernel smoothing technique with a plug-in bandwidth selection procedure. This methodology attempts to capture the skill and uncertainty in the QPF. We present both dependent and independent validation results for selected river basins in the AB-, CN-, and MA-RFC areas.
H14D-08
Snowcover and North American Monsoon Precipitation: An Occasional Partnership
It has long been thought that winter snow cover may play a role in modulating subsequent summer monsoon precipitation, both for the Asian monsoon, and for the North American monsoon system (NAMS). Basic physical reasoning suggests that snow cover decreases the amount of energy available for surface heating. Consequently, when there is an above/below average amount of snow cover, the NAMS should be weaker/stronger. Previous studies have had difficulty in obtaining such a relationship from observational analyses. We present results that suggest the period from 1979-1991 was very different from the period 1992-2005. During the first period, a strong negative correlation is found between western US snow cover and NAMS precipitation the following summer; this is the 'expected' relationship. During the second period, the correlation becomes positive instead, and still fairly strong. Over the entire 1979-2005 time period, the correlation is therefore quite weak. We explore possible reasons for this apparent regime shift, in particular the role that the large-scale circulation may play. Snow cover in the western US tracks the Pacific Decadal Oscillation (PDO) fairly well, and is known to be influenced by ENSO, though no obvious changes occur in either during the early 1990's. One clue is that the phase relationship between the Arctic Oscillation (AO) and ENSO reverses in the early 1990's, which may in turn be related to a major phase change in the Atlantic Multi-decadal Oscillation (AMO).