Report of a Workshop on Predictability and Limits-To-Prediction in Hydrologic Systems (2002) / Chapter Skim
Currently Skimming:
Appendix B: Summary of Selected Papers Presented at Workshop
Appendix B: Summary of Selected Papers Presented at Workshop
Pages 41-110
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 41... ...
Appendix B Selected Papers Presented at Workshop on Predictability and Limits-to-Prediction In Hydrologic Systems
|
From page 43... ...
Thus, the memory associated with land surface soil moisture may turn out to be the chief source of midIatitude forecast skill. The accurate initialization and modeling of soil moisture can contribute to a seasonal forecast only if two conditions are met: (1)
|
From page 44... ...
. Soil Moisture Memory We recently manipulated the water balance equation at the soil surface into a relationship between the autocorrelation of soil moisture and the statistics of the atmospheric forcing, the variance of soil moisture at the beginning of the time period in question, and the structure of the land surface scheme used (Koster and Suarez, in preps.
|
From page 46... ...
Ensemble ~ consisted of 16 45-year simulations with interannually varying sea surface temperatures (SSTs) and interactive land surface processes.
|
From page 47... ...
Ha/ 1 It ~ . ~ IIlIllss~s~: ~ : ~ ~ .
|
From page 48... ...
Thus, although the members of the ensemble differed because of their different atmospheric initial conditions, each was forced to maintain the same time series of (geographically varying) land surface prognostic variables.
|
From page 50... ...
Only the land surface feels this more realistic precipitation; the GCM's water vapor fields and the latent heating of the atmosphere are not replaced. The land surface presumably develops more realistic soil moisture states in response to the more realistic precipitation forc~ng.
|
From page 51... ...
How relevant is each factor in the real world, and do GCMs simulate their relative importance correctly? Can we specify regions and seasons for which useful soil moisture memory is essentially unattainable?
|
From page 52... ...
in this phase, the coupled land-atmosphere model is forced to maintain realistic soil moistures at all times, so that the focus is mainly on atmospheric response to land conditions (i.e., on the strength of the coupling)
|
From page 53... ...
The first phase should address the second question above regarding the degree to which the atmosphere is coupled to the land surface.. An improvement in simulated precipitation due to more realistic soil moistures, if it occurs, would serve both to demonstrate the existence of coupling in the real world and to validate the coupling strength in the modeling system.
|
From page 54... ...
In addition, maintenance and continuation of observational programs, particularly for soil moisture and meteorological data, are critical for providing a basis for the evaluation of prediction systems.
|
From page 55... ...
, with the mode! hierarchy on the outside, covering the various spatial scales, and the data assimilation system in the inside, for determining the proper initial conditions for the models.
|
From page 56... ...
Uncertainty in the Initialization of the Models State variables in many of the models have to be initialized. This has led to the development of sophisticated variational methods in atmospheric and ocean sciences, while for operational purposes, methods to initialize land surface models are still in the early stages of development.
|
From page 57... ...
Finally, the "DA" for "data assimilation" in LDAS denotes a later "Phase TI" thrust that will include the assimilation of satellite-derived land-surface fields, such as skin temperature, soil moisture, snowpack, and vegetation density and greenness. Another future but central EDAS phase will be a forecast component, wherein LDAS will be integrated days, weeks, and months into the future using ensemble surface forcing (including ensemble quantitative precipitation forecasting (QPF)
|
From page 58... ...
How do we estimate mode] parameters in hydrologic models?
|
From page 59... ...
The soil moisture data set, with daily updates through yesterday 12Z, thanks to daily precipitation analysis by Higgins and Shi, has been widely applied for drought and wetness monitoring, the launching of monthly/seasonal prediction tools, and many research projects. Other outputs of the mode!
|
From page 60... ...
The soil plays an important role in these shorter-range forecasts, one main problem being drift of the data assimilation system due to biased precipitation input. We will only mention below the EDAS activity, which is a modern data assimilation, but observed precipitation is used to avoid such drifts.
|
From page 61... ...
The global data sets are mainly used for two kinds of applications: one where soil moisture is a prescribed lower boundary condition, and the other for various geodetic research activities such as investigation of the feasibility of detecting soil moisture anomalies by measuring gravity from satellite and determinations of the annual cycle of the geoid etc. Near Future (2-3 years)
|
From page 62... ...
consistent prescribed lower boundary conditions in AMIP-like experiments to determine predictability due to soil moisture anomalies. Priority science questions in predictability are first and foremost the design of experiments that could give us pertinent information about hydrology and atmospheric predictability.
|
From page 63... ...
Key Issues Driving the WEB Imperative sing: Key issues driving the WEB imperative include the follow
|
From page 64... ...
· Regional interests, agency missions, disciplinary training, and the institutionalization of field experiments have intellectually partitioned water studies into domains. · Scientists working in each domain increasingly recognize that their findings depend highly on boundary conditions at the interfaces with the other domains.
|
From page 65... ...
Atmospheric water vapor and the conflicting effects of clouds on radiation (Webster, 1994) are linked to the hydrologic cycle via cloud formation, precipitation, and evapotranspiration.
|
From page 66... ...
WEB in River Basins as an Illustrative Theme River basins integrate physical, chemical, and biological processes in the spatial and temporal organizations of fluxes and structures. Science issues that require focused data collection, hypothesis development, and tests and that have the potential for significant cross-disciplinary theoretical advances are identified in a river-basin context as an illustrative example.
|
From page 67... ...
· Human modification to a river basin alters local structure and fluxes in ways that have significant impacts on both upscaling and downscaling. An Interdisciplinary fr~itiative for Education and Public Outreach An interdisciplinary initiative for education and public outreach will enable students and the general public to understand the unique role of the planetary water cycle in the co-evolution of life and of a habitable climate on Earth for 3.8 billion years and also to understand and solve the water problems facing society.
|
From page 68... ...
68 Appendix B science theme at the river-basin scale. A WEB Office for Science Support (WOSS)
|
From page 69... ...
. What is commonly referred to as the "initial value problem" is one aspect of predictability.
|
From page 70... ...
These coupled variations result from the fact that the aforementioned terrestrial water storages have a persistence timescale that is comparable to or longer than the characteristic timescales of the interacting atmospheric processes. Therefore, it follows that regions of strong persistence of "active" water storage (i.e., terrestrial water storage available to exchange processes between the land and the atmosphere)
|
From page 71... ...
Even still, we must verify that the resulting runoff rates provided to the hydrologic prediction system are predictable. Is the Predictability for Real?
|
From page 72... ...
through linear approximations of evaporation and runoff controls as a function of soil moisture. The results underscore the potential to effectively summarize and verify the degree and controls of soil-water persistence for the variety of models used to represent the continental surfaces in weather and climate models.
|
From page 73... ...
Ensemble prediction is now an everyday part of operational short- and medium-range weather forecasting at European Center for Medium Range Weather Forecasts (ECMWF) and National Centers for Environmental Prediction (NCEP)
|
From page 74... ...
formulation and low-level moisture, uncertainty in these aspects will likely require a different approach than the one taken on the synoptic scale. At long range, the initial conditions of an ensemble are not treated as elaborately as in the short and medium range, since at these time scales initial conditions are nearly completely forgotten, and anomalous boundary forcing is the sole residual memory driving predictability.
|
From page 75... ...
With respect to precipitation, global prediction and climate models have significant mean climate errors in both the location and magnitude of precipitation. Because the targeted forecast problem is the response to boundary anomalies, (frequently tropical ocean sea surface temperature anomalies)
|
From page 76... ...
76 Appendix B Concluding Remarks Because the current generation of global models and mesoscale models have strong systematic biases in the monthly to seasonal timeframe for global models and 3- to 12-hour range for mesoscale models, the optimal ensemble strategies for precipitation prediction in these time and space scales will include a multimode! ensemble.
|
From page 77... ...
Each parameter and rainfall input has its own spatial variability, which can be characterized by informational entropy (Vieux, 1993~. Incorporating spatial variability of a particular parameter related to vegetative cover, soil infiltration rates, rainfall, or topographic characteristics may or may not have important conse-quences on the hydrograph response of the model.
|
From page 78... ...
Unlike empirically based models, differential equations are used to describe the flow of water over the land surface or through porous media to describe, or energy balance in the exchange of water vapor through evapotranspiration. If the physical character of the hydrologic process is not supported by a particular analogy, then errors result in the physical representation.
|
From page 79... ...
Research Questions following: Research questions related to parameter modeling are the . Parameterizing a distributed hydrologic model requires adequate spatial variability.
|
From page 80... ...
Having access to real-time fusion of rain gauge, satellite, and radar estimates of precipitation would advance hydrologic prediction. Algorithms and systems that can ingest multiple-sensor platforms and produce high-resolution rainfall estimates in space and time are needed to make flood forecasts.
|
From page 81... ...
Research on the relationship between storm and basin scale and the time necessary to issue warnings that are reliable and of value is needed so that action can be taken to mitigate flood damages. Specific Research Questions: Specific research questions related to flood forecasts in small basins are the following: · Computational time may be limited when forecasting small basins.
|
From page 82... ...
At soil moisture contents of greater than 50 percent, a greatly amplified response to soil moisture conditions results. Research Questions Research questions related to the interdependence of components include the following: · Can adjoins methods used in atmospheric models be exploited in physics-based hydrologic models to automate calibration or make real-time adjustments to flood forecasts?
|
From page 83... ...
However, there are numerous other problems with rain gauge measurements. All gauge designs require frequent maintenance and are subject to
|
From page 84... ...
(2000) attempted such an evaluation of the hourly product used operationally by the National Weather Service as input to hydrologic models and concluded that they could not find indepen-dent information.
|
From page 85... ...
The mission continues stimulating much of rainfall research, including research on new rainfall estimation algorithms and their validation. Perhaps the most mature satellite-based rainfall estimation product is the ~ O year data set of monthly rainfall produced by the Global Precipitation Climatology Project (GPCP)
|
From page 86... ...
Also, as follows from the discussion above on the rainfall estimation methods, it is important to figure out a way to separate the forecasting error that is due to the method from that of the observation that serve as the forecast verification. The second group of methods is mathematical models of the physical processes involved in generation of precipitation.
|
From page 87... ...
· Develop methodologies of using rain gauge data based reference standards for evalua-tion of radar-rainfall estimates. · Develop new sensors and technologies that would provide a better reference standard than the gauges.
|
From page 88... ...
88 Appendix B Recommendations: QPF · Develop a set of benchmarks based on large samples of data. · Develop a framework for evaluating the quality of forecasts independent of the errors of the reference data.
|
From page 89... ...
Efforts aimed at quantitatively determining the range of predictability in the atmospherichydrologic system have been based either on idealized systems of dynamic equations or in model predictability studies employing more elaborate numerical models. in the first case, methods of nolinear dynamics (e.g., Lyapunov exponents)
|
From page 90... ...
has its own problems, which remain mostly unresearched to date. A main problem is the propagation of boundary conditions (e.g., see Chu, 1999)
|
From page 91... ...
A suite of multiscale methods for comparison of forecasted precipitation to observed precipitation has been developed and has shed some light into shortcomings of numerical weather prediction models in terms of capturing statisti
|
From page 92... ...
and has considerable scale dependency within the typical mesoscale range of 5 to 50 km (Tustison et al, 20001. In verification studies in which model performance is assessed as a function of mode} resolution, ignoring or mischaracterizing the scale dependent representativeness error can significantly affect inferences about the mode} performance as a function of scale.
|
From page 93... ...
Answer: Commonly used relationships must be studied to establish "scaling'' relationships, which depend on the form of the nonlinear relationship and the spatial variability of the involved variables. Supportine Evidence: Through the analysis of a commonly used relationship that parameterizes surface runoff as a function of soil moisture, it has been demonstrated how the parameters of the relationship must change with scale such that spatially averaged predicted fluxes can be preserved at any scale of interest (Nykanen and FoutoulaGeorgiou, 2000~.
|
From page 94... ...
94 Appendix B WHY AND HOW DOES PRECIPITATION CHANGE?
|
From page 95... ...
It Is also viable from many observations, from recording rain gauges and from radar (e.g., NEXRAD) and satellite (e.g., TRMM, GOES)
|
From page 96... ...
Held (1993) notes that extratropical storms are greatly influenced by moisture in the atmosphere and that one effect of increased moisture content in the atmosphere is to enhance the latent heating in
|
From page 97... ...
Rainfall amounts over the oceans are also important because of its contribution to the fresh water budget for the ocean—or equivalently salinity. It is a key in the potential changes in the thermohaline circulation as the climate changes (Gent 2000~.
|
From page 98... ...
Interactions within the climate system occur at a wide range of temporal and spatial scales, as discussed, for example, in Pielke (19981. In the context of prediction, the interactions that involve water with other components of the climate results in an initial value problem, if the feedbacks are sufficiently large and nonlinear, and occur within the time period which the forecasts are made for.
|
From page 99... ...
, there has been no assessment of whether water vapor, another radiatively active gas, would have increased input into the air as a result of this carbon mitigation approach. The net result of a carbon sequestration program that does not also consider the associated effect on hydrologic processes could be a greater net flux of radiatively active gases into the atmosphere (as well as a change of the net radiation at the surface by altering its albedo)
|
From page 100... ...
4. The success of seasonal weather forecasts has primarily been a result of the treatment of the sea surface temperatures as a static lower boundary condition (Landsea and Knaff 2000~.
|
From page 101... ...
1994. The global precipitation climatology project: First algorithm intercomparison project.
|
From page 102... ...
2000. Covariance propagation and updating in the context of real-time radar data assimilation by quantitative precipitation forecast models.
|
From page 103... ...
2000a. Rainfall forecasting us~ng variational assimilation of radar data in numerical cloud models.
|
From page 104... ...
1997. The interplay between transpiration and runoff formulations in land surface schemes used with atmospheric models.
|
From page 105... ...
1995. The status of satellite-based rainfall estimation over land.
|
From page 106... ...
Gash, and H Dolman, The role of the land surface in weather and climate: Does the land surface matter.
|
From page 107... ...
1999. Effect of bias adjustment and rain gauge data quality control on radar rainfall estimation.
|
From page 108... ...
2000. Distributed Hydrologic Modeling Using GIS.
|
From page 109... ...
~ 997. Atmospheric water vapor over China.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.