Search Results

You are looking at 1 - 10 of 68 items for

  • Author or Editor: Thomas R. Karl x
  • All content x
Clear All Modify Search
Thomas R. Karl

Abstract

Statistical forecast equations have been developed for the Greater St. Louis, Missouri, area using Model Output Statistics (MOS) derived from the National Meteorological Center's Limited-Area Fine Mesh (LFM) model. They are used to forecast both the probability of ozone concentrations exceeding the 1971 National Ambient Air Quality Standard and the daily 1 h maximum. Predictions extend out to two days (48 h).

The application of MOS to forecasts of maximum O3 concentrations lead to skillful [better than chance, persistence or climatology (seasonality)] 24 and 48 h objective predictions. The application of MOS to probability statements about O3 concentrations also resulted in reasonable success for 24 h and 48 h probability forecasts. These forecasts appear sufficiently successful to warrant consideration of the development of other equations for large metropolitan area where O3 concentrations commonly exceed the standard.

Full access
Thomas R. Karl
Full access
Thomas R. Karl

Abstract

The Palmer Drought Severity Index (PDSI) is routinely made available by NOAA for operational use, and it has also been calculated across the United States on a historical basis back to 1895 (Karl et al., 1983). Traditionally, the coefficients used in the calculation of the PDSI have been based on an anomalously hot and dry period across much of the United States (1931–60). By changing the base period used to calibrate the coefficients, the magnitude and the sign of the PDSI change significantly in many areas of the United States. Often the changes are larger than those that occur when the potential evapotranspiration is forced to a constant equal to the long-term monthly mean potential evapotranspiration. This sensitivity to base period calibration has important implications in the interpretation of operational or hindcast values of the PDSI for forest fire danger and other applications. The less frequently used Palmer moisture anomaly index (Z-index) is much less sensitive to changes in the calibration periods, and also has some desirable characteristics which may make it preferable to the PDSI for some agricultural and forest fire applications, i.e., it is more responsive to short-term moisture anomalies.

Full access
Thomas R. Karl

Abstract

The water content parameter and the moisture anomaly index, both derived from the Palmer Drought Severity model, were correlated against subsequent mean monthly and seasonal (three-month means) temperatures for 344 climatic divisions in the United States (1931–83). During spring and summer, Monte Carlo field significance tests demonstrate that the correlation fields produced from the soil moisture parameters are significantly larger than those derived using persistence of monthly and seasonal temperature anomalies. The areas of the United States with enhanced soil moisture parameterization–temperature correlations tend to be confined to the interior. The average reduction of the standard error of estimate (root-mean-square error) is 0.15°C for seasonal forecasts made at the end of April and May over inland, nonarid climate divisions. The reduction of the standard error is less for monthly forecasts (∼0.05°C) during the April through July forecast time periods. The areas with the greatest reduction in error favor the southern portions of the United States during early spring, but in late spring and summer they are also found in the central and northern states.

The empirical relationships found between soil moisture parameters and subsequent monthly and seasonal temperature suggest that soil moisture is important on a local scale, not only in a diagnostic mode, but also in a prognostic sense. Since the spring and early summer are times when the persistence of temperature anomalies is not very effective as a prediction tool, the existence of a demonstrable increase in predictability of monthly and seasonal temperatures using soil moisture indices, easily calculated on an operational basis, suggests that another objective long-range forecast aid is available for immediate use.

Further increases of the correlations of soil moisture with subsequent monthly and seasonal mean temperature may come from improvements in the water balance computations which are part of the Palmer model, i.e., estimation of evapotranspiration, the treatment of runoff, the inclusion of snow cover, the inclusion of irrigation estimates, or from other models, or preferably from a network of soil moisture measurements.

With respect to the Palmer water budget evapotranspiration calculations, sensitivity studies of the soil moisture parameterizations were performed using a fixed annual cycle of evapotranspiration—no year-to-year variations of the annual cycle. Statistically significant, but not drastic, degradations of the correlations of parameterized soil moisture with subsequent seasonal and monthly mean temperature were noted. The largest increase of forecast error occurred during the seasonal forecast period. This suggests that improved estimates of evapotranspiration in the water balance equation may be especially important for long (seasonal) forecasting periods during late spring and summer, but dramatic improvements in forecast skill may be difficult to achieve.

Full access
Thomas R. Karl

Abstract

By using the Palmer Drought Severity Index (PDSI) as calculated from state averages of temperature and precipitation and from numerous single station analyses, it has been demonstrated that droughts (as defined by the PDSI) persist longer in the interior of the United States than in areas farther east or west. The question arises whether this is merely an artifact of the PDSI calculations, or whether there is actually more persistence of abnormally dry (wet) weather in the interior.

The sensitivity of the PDSI was tested in relation to changes of derived and prescribed parameters included in the PDSI calculations in order to determine their effect on the spatial characteristics of drought duration. The sensitivity tests indicated a negligible effect. Contingency tables which use the PDSI as the predictor for the following one-month, six-month and 12-month precipitation anomalies (and also anomalies of precipitation minus potential evapotranspiration) however, are generally characterized by significantly greater skill in the interior portions of the United States, confirming the nation that spells of abnormally dry or wet weather do have more persistence in the Rocky Mountain and High Plains states than states farther east or west. Unfortunately, the forecasts derived from the “operational” PDSIs were not a significant improvement from what would have been obtained by using precipitation persistence forecasts.

Full access
Thomas R. Karl

Abstract

The ratio of the probability of at least one extremely cold or warm month (standardized departures ≥ 1.282 or ≤ −1.282) in a season with near-normal mean temperatures (standardized departures ≤ 0.524 but ≥−0.524) to the probability of such an event in abnormal seasons has been calculated using statewide average monthly temperatures (1895–1983) across the United States. Values of this ratio, termed the “ratio of variability” (RV), near one reflect nearly an equal probability of one or more extreme months in both near-normal and abnormal seasons, while values near zero indicate little chance of an extreme month in a near-normal season. The values of RV vary with geographic location and the time of year in a systematic predictable manner. The magnitude of RV is greater during the transition seasons than during either summer or winter. The gradients of RV are relatively flat in the autumn, but comparatively sharp in spring with a maximum in the east and central United States and a minimum in the west. In the summer the largest values of RV are found along the northern tier of states and along the mid-Atlantic coastal states, but during winter the central portions of the United States and portions of the northeast have the largest values of RV.

The two factors responsible for the seasonal changes and spatial gradients of RV are the spatial and temporal changes of the month-to-month persistence (or lack of persistence) of unusually cold or warm weather and the unequal contribution of variances within a season by each of the months in the season. This is demonstrated using Monte Carlo simulations of an autoregressive model. Users of seasonal average temperatures, whether in a forecast or hindcast mode, whose operations are sensitive to persistent unusually cold or warm temperatures within a season with near-normal temperatures, should be cognizant of the spatial and temporal changes of RV.

Full access
David R. Easterling, Thomas C. Peterson, and Thomas R. Karl

Abstract

At the National Climatic Data Center, two basic approaches to making homogeneity adjustments to climate data have been developed. The first is based on the use of metadata (station history files) and is used in the adjustments made to the U.S. Historical Climatology Network monthly dataset. The second approach is non-metadata based and was developed for use with the Global Historical Climatology Network dataset, since there are not extensive station history files for most stations in the dataset. In this paper the two methodologies are reviewed and the adjustments made using each are compared, then the results are discussed. Last, some brief guidelines on the limitations and uses of these data are provided.

Full access
Mark D. Schwartz and Thomas R. Karl

Abstract

In spring many plants break dormancy and begin foliage production. The appearance of leaves (the “green-up” period) triggers a rapid increase in transpiration at the surface as well as changes in albedo. Subsequently, these processes alter the thermodynamic properties of the surface layer. Normally, seasonal variations in tropospheric thickness, and year-to-year variability in the green-up date, mask the impact of these effects on surface maximum temperatures. In this study we used first leaf phenological data (from the clone lilac Syringa chinensis in the central and eastern United States) as the indicator of transpiration onset, in order to reveal the effects of this change from a dormant vegetative surface, perhaps comparable to the worst summertime droughts, to an active foliage-producing and transpiring vegetative surface.

Simple plots using average hypsometric layer-mean temperature (derived from geopotential thickness) and average surface daily maximum temperature were examined for variations in the thickness-maximum temperature relationship relative to first leaf date. We then employed a multiple regression model to test the significance and estimate the magnitude of changes in this relationship. The model confirmed that there are statistically and practically significant relationships between the timing of the green-up period and surface daily maximum temperature. For the same thickness value, one station type (“A”; generally in agricultural inland areas) showed at least a 3.5°C reduction in surface daily maximum temperatures over any two-week period subsequent to first leaf compared to a two-week period prior to first leaf. Station locations generally near major water bodies (type “B”) showed a smaller (1.5°C) reduction. From these results we infer that even without feed backs between the surface vegetation and the atmospheric circulation, the surface daily maximum temperature may be significantly altered. During extreme drought, which produces widespread plant wilt, similar effects may be expected. In addition, the results indicate that the daily weather forecasts from specification and model output statistics equations may be significantly improved during spring by inclusion of a green-up variable.

Full access
Thomas R. Karl and Claude N. Williams Jr.

Abstract

A method is described whereby climatological time series of temperature and precipitation can be adjusted for station inhomogeneities using station history information. The adjusted data retains its original scale and is not an anomaly series. The methodology uses the concepts of relative homogeneity and standard parametric (temperature) and nonparametric (precipitation) statistics. The technique has been tested in Monte Carlo simulations, and is shown to product climatological time series more consistent with the concept of a homogeneous climate record than would other be the case. Additionally, the technique provides an estimate of the confidence interval associated with each adjustment. It has been applied to over 1200 stations in the United States. In many instances the adjustments in temperature time series are substantial (as large as actual climate fluctuations during the twentieth century) often leading to a more consistent pattern of regional climate change than would otherwise be surmised from inspection of the unadjusted data.

Full access
Thomas R. Karl and William E. Riebsame

Abstract

A potentially fruitful approach to assessing society's sensitivity to climate change is to study the impacts, perceptions and adjustments of recent climate fluctuations. We set out to determine if the recent (1931–82) United States climate record exhibits fluctuations of sufficient scope and magnitude to be useful in a complement of retrospective, empirical studies of climate impacts. The search for fluctuations was designed specifically to identify areas and periods in which the climate within an epoch was terminated by a rather sharp transition to another epoch with a climate unlike the previous epoch. The largest 10- to 20-year temperature and precipitation climate fluctuations were identified across the contiguous United States, along with various scenarios of simultaneous change of temperature and precipitation for the four seasons and annually. All possible 10- to 20-year nonoverlapping “consecutive epochs” within 344 state climatic divisions (as defined by the National Climatic Data Center) were examined for the greatest temporal differences of temperature and precipitation. Additionally, large spatial gradients of climate fluctuations of opposite sign have been identified across climate divisions within short distances of each other (<750 km). The climate fluctuations were specifically identified for the purpose of studying and modeling climate impacts, but they are also of general interest to researchers investigating the physical behavior of climate across the United States.

On a seasonal basis our results indicate that over the past half century the most significant and widespread climate fluctuations for temperature of 10–20 years duration, in terms of standardized departures, have been associated with temperature changes of 2°C or more during the winter and summer seasons. Precipitation fluctuations of 25% or more have been detected for similar durations. A partial analog to the current prediction of climate change due to a doubling of CO2 concentration was also identified, namely, an increase of spring and summer temperature (approximately 1°C) and a decrease in precipitation (20–40%) in the central and northern Great Plains.

Relatively large spatial gradients of climate fluctuations were identified for nearby climate divisions with opposite climate fluctuations. These spatial “seesaws” of climate fluctuation may prove particularly useful as experimental controls in the study of climate impact, perception and adjustment. Furthermore, they point up the importance of a dense climatological monitoring network.

Full access