This page's content is no longer actively maintained, but the material has been kept on-line for historical purposes.
The page may contain broken links or outdated information, and parts may not function in current web browsers.

Research Results

Storms in the Future: Changes in Intensity, Cloudiness, Rainfall and Economic Costs

Title | Introduction | Methods | Results 1 | Results 2 | Results 3 | Discussion

Methods

Common Aspects

For the proxy method we had access to a variety of overlapping, though not coincident data sets. The National Centers for Environmental Prediction (NCEP) reanalysis provided data about storm dynamics obtained twice daily from weather stations on the ground over the period 1961 to 1998. This data includes the sea level atmospheric pressures we used to identify the storms, and the wind speeds, surface temperatures and the relative humidities in the regions included by the storms.

The International Satellite Cloud Climatology Project (ISCCP) provided the data about the cloud properties of the storms over the period July 1983 to December 1995. Satellite images taken every three hours were paired with the NCEP datasets to determine the cloud properties of the storms. These properties included cloud optical thickness, cloud top pressure (a measure of the height of the cloud top, the higher the cloud in the atmosphere, the lower the atmospheric pressure), the amount of cloud coverage (the percentage of the area covered by clouds), cloud top temperature (again, the higher the top of the cloud, the lower its temperature), and the amount of deep convective and nimbostratus cloud coverage (as these are the main rain producers in storms, the amount of these clouds gives an indication of the amount of rain associated with the storm).

A third set of data came from the Property Claim Service of the Insurance Association of the United States. This dataset included a variety of information including the dates, types, locations (state) and amounts of damages over $100,000 associated with each storm over the period January 1970 to December 2000. Unfortunately the dataset was incomplete and we were missing data for the years 1986, 1988, 1989, 1990, 1991 and 1992.

The last dataset came from two runs of the SI95 atmospheric model version of the GISS GCM, featuring the atmospheric model attached to a simple ocean model with fixed dynamical heat transports. Dr. Hansen gives the complete details and description of this version of the model in one of his papers.6 The model is run for fifty simulated years, once under normal conditions and then once under doubled CO2 conditions to replicate the effects of global warming conditions. The resulting final conditions after the two runs are saved and analyzed.

To isolate the midlatitude storms for the time period 1961 to 1998, the GISS storm-tracking program was utilized to examine the NCAR dataset. The storm-tracking program uses an algorithm to examine the sea level pressure field of this data set and then selects the low sea level pressures (SLP) that fit specific criteria as to direction of motion and speed between 12-hour observations. These associations of lows then become a record of the storm-tracks for each month and season. This information is saved and can be used to show not only the paths of the storms, but also the frequencies, average SLP and the maximum SLP for the various parts of the world on monthly and seasonal scales.


Team 1: Can warm and cold years of the past be used as proxies for future climate change?

In order to try to validate the use of the past data record as a proxy for future climate change, this team intended to compare the results of using the past data set with the results obtained from the GISS GCM to describe the changes in storm properties in the event of continued global warming. The storm tracks obtained from the program described above were analyzed for the months of January, April, July and October (chosen to be representative of the four seasons) to look for annual and seasonal trends over the 37-year period of available data. Frequency distributions of the sea-level pressures were analyzed and compared with the average temperatures of the northern midlatitudes for each season of each year. Comparisons were also made with two other accepted climate indicators, the North Atlantic Oscillation (NAO) and the Southern Oscillation Index (SOI, a measure of the El Niño phenomena).

To better isolate the changes in storm properties that may be expected to occur with continued global warming, the five warmest and five coldest years were selected for each season. This temperature ranking was based upon the average surface temperature of the northern midlatitudes for each month. The average frequency distributions of the sea level pressures for the five warmest years and the five coldest years for each season were then compared and contrasted. The same was done with the sea-level pressure frequency distributions for the single warmest and the single coldest years.

The average storm frequencies, the average storm sea-level pressures and the average maximum storm sea-level pressures were than determined across the globe for the five warmest and the five coldest winters (using the available January values to represent winter), as well as for the single warmest and the single coldest winter. The differences of these averages were found, and these differences were then compared and contrasted with the differences found between the CO2 doubled and control runs of the GISS GCM.


Team 2: How have the damages due to storms changed over time, and how are these changes related to the storms?

Team 2 began their task by analyzing the database they created from the data obtained from the Property Claim Service of the Insurance Association of the United States. As we are only interested in the damages caused by midlatitude storms, the damages due to hurricanes were removed from the cost of the damages used in this investigation. The team looked for trends in storm damages over the thirty-year period for which they had data. The magnitude of the damages in the later years of this period caused them to examine economic trends over the same period, and the amounts paid for damages were corrected for inflation and prosperity, using the 1970 United States dollar and the 1970 Gross Domestic Product (the GDP, as an indicator of prosperity) as baseline values. Trends were examined not only over the entire 30-year period, but also at a decadal level for the 1970s, the 1980s and the 1990s.

To determine changes in the storms over the United States for this time period, team 2 then looked at the average storm frequencies, the average storm intensities and the average maximum storm intensities, using the values obtained from the GISS storm-tracking program. Trends in these values were correlated with trends in the storm damages over the entire 30 years and for each decade.

The last analysis method used by this team was to examine the distributions of the average storm intensities and the average maximum storm intensities for the winters of each of the years for which they had data. Not only were trends looked for, but these distributions were also averaged for the five winters with the lowest storm damages and these were compared to the average distributions of the five winters with the highest damages. Comparisons were also made between the single winter with the highest damages and the single winter with the lowest damages.


Team 3: How are the properties of the storm clouds over the United States changing with time?

Team 3 took two approaches to examining the data for evidence of changes in the properties of the clouds of the storms across the United States. This team was the most restricted in their analysis since the ISCCP data was available for the shortest time period. One approach was to look at the averages, minimums ands maximums of the properties of the clouds over the United States by extracting those values from the ISCCP dataset. The cloud properties that were selected were cloud optical thickness, cloud top temperature (a measure of the height in the atmosphere of the top of a cloud; the higher the top, the lower the temperature), average amount of cloud cover (the percentage of the area that is covered by clouds), the daytime amount of deep convective clouds and the daytime amount of nimbostratus clouds. (These last two values are a measure of the amounts of the two major rainmaking clouds found in storms). The averages, minimums and maximums of these values were then examined for trends over the twelve-year period covered by the data set. And these trends were compared to those found in the storm damages.

The second approach utilized by team 3 was to examine the storm and cloud properties of specific storms in the winters of years with low storm damages and years with high storm damages. These high (1993) and low (1985) damage winters were selected by team 2 based upon their analysis of winter storm damages and the limitation of data available from the ISCCP. The storm-tracking program then provided the paths of the storms for these two winters, and two similar storms, one from each year, were selected for comparison. This initial study was limited to two storms due to the time constraints of the summer program, and will hopefully be expanded in the future.

Once the two storm tracks were selected, the data for the storm properties was extracted from the NCAR dataset and the data for the cloud properties was extracted from the ISCCP D1 data set. Since both datasets have the same resolution, a standard storm area was defined. This standard was determined by examining several days from each storm track. The team looked for the smallest area (in terms of the grid boxes of the datasets) that would include as many of the storm clouds as possible. A six by six grid box area around the position of the central low pressure was chosen. Using this standard storm area, the dynamic properties of the storms were pulled out of the NCAR dataset. Specific dynamic properties that were examined included: sea level pressure, air temperature at 850 mb, north/south and east/west wind velocities, vertical wind velocity, relative humidity, and geopotential height. The same storm area was used to extract the properties of the clouds from the ISCCP data set. The cloud properties that were extracted included: cloud optical thickness, cloud top temperature, amount of cloud coverage, and the amount of nimbostratus clouds.

Once the extractions were made, the properties of the storms and their clouds in the selected storm areas were analyzed. The averages, maximums and minimums of these properties were calculated. Theses values were obtained for each day of both storms, plotted against time and examined for relationships and trends over the lifetime of the storm. The expectation was that the storms of the high damage year would be stronger than the storms of the low damage year. Since the minimum sea level pressure of the storm and the wind speed of a storm usually determine its strength, a storm strength was defined as

Storm Strength = (Maximum Wind Speed) / (Minimum Sea Level Pressure)


References

6. Hansen, J., et al. 1997. Forcing and chaos in interannual to decadal climate change, J. Geophys. Res. 102, 25679-25720.

Title | Introduction | Methods | Results 1 | Results 2 | Results 3 | Discussion