Verification of the OCSI Atlantic Hurricane Predictions Since 1985

Verification of the OCSI Atlantic Hurricane Predictions Since 1985

By:

Jill F. Hasling, CCM and Dr. John C. Freeman, CCM

Weather Research Center – Houston, Texas

www.wxresearch.com

INTRODUCTION

The Orbital Cyclone Strike Index [OCSI®] is based on the premise that the orbit of the Sun about the center of gravity of the solar system has an influence on the general circulation of the earth’s atmosphere. This influence is reflected in the paths of Atlantic tropical cyclones. Freeman and Hasling (2002) have recently pointed out that this orbit of the sun is shared by the earth. Jose (1969) and Landscheidt (1976) found that the orbit of the Sun is related to solar activity including the sun spot cycle. This seems to indicate that there are orbital influences reflected in the solar activity.

Forecasts of the sections of the United States Coast with the highest risk of experiencing a landfall of a tropical storm or hurricane have been made since 1985. These forecasts have been verified and the results of this verification is given in the paper below.

Secondary forecasts are made each year of the number of tropical cyclones expected each year as well as the number that would intensify into hurricanes. The number of storm days and hurricane days were also included in the outlook. The verification of these forecasts is also given with a comparison to Bill Gray’s seasonal outlooks.

OCSI

Recent research indicates that the premise that there could be orbital influences on the large scale circulation patterns in the earth’s atmosphere is plausible. Freeman and Hasling (2002) believe that the earth revolves around the center of the sun as the sun goes through its orbit about the center of gravity of the solar system. This orbit is believed to be one of the influences that drives the solar cycle [sun spot cycle] as well as cycles in the earth’s atmosphere.

The OCSI starts with the years when the sunspot minimum occurs. These years are Phase 1 of the index. The year after the sunspot minimum would be Phase 2 and so on. A cycle could have as few as 9 phases or as many as 14 phases. This process creates the OCSI index in Table 1 which is made up of 10 cycles that start years 1878, 1889, 1901, 1913, 1923, 1933, 1944, 1954, 1964 and 1976.

The Atlantic tropical cyclone data is then sorted into the index to determine which sections of the US coast has the most years with landfall of tropical storms or hurricanes in each Phase. Table 2 gives the probabilities for each section of the US coast by phase. The probability was computed by taking the number of years out of 10 that a storm made landfall on a particular section of the coast. This information is used to make a prediction of which section had the highest risk of experiencing a tropical cyclone landfall each year as well as which sections had the second highest risk each year.

Forecasts have been made each year since 1985. The forecasts from 1985 which was phase 10 of Cycle 10 through 2003 were used for this verification. The cycle which started with 1851 and the cycle which started with 1867 were also used for verification.

Figure 1 gives the forecast for each section of the United States Coast for Phase 9 [which will be the outlook for the 2004 Atlantic Hurricane Season]. As you can see from Table 1 the years which make up Phase 9 are 1886, 1897, 1909, 1921, 1931, 1941, 1952, 1962, 1972, and 1984. Figure 1 indicates how the highest and second highest probability was computed. Figure 1 indicates that the West Coast of Florida had 7 out of the 10 years [1886, 1897, 1909,1921,1931, 1941, 1952, 1962, 1972, 1984] which experienced landfall storms along the west coast of Florida. This gives the West Coast of Florida a 7 out of 10 chance or 70% chance of experiencing a tropical storm or hurricane. For Texas there 6 out of the ten years when a tropical storm or hurricane made landfall on the Texas coast.

The years that had Texas landfalls were 1886, 1897, 1909, 1921, 1931 and 1941. This gives Texas the second highest probability in the 2004 season out look or a 60% chance.

When reviewing the years in each phase some the significant events are noted. For example in Phase 9 of the OCSI, there have been several very significant hurricanes. There was the 1875 Texas hurricane which almost destroyed the town of Indianola and in 1886 another hurricane did destroy the town of Indianola. Other strong Texas storms during this Phase of the OCSI were in 1909, 1941. There was a very strong Florida Keys hurricane in 1909 as well as a very strong Louisiana hurricane in 1909.

VERIFICATION OF THE OCSI

The verification of the forecast made each year from 1985 to 2003 is in Table 3. A forecast verifies with landfall in at least one the sections with the highest probability of landfall. In the past 19 years there have been only 2 years [1987 and 1992] when a storm did not make landfall in any of the sections with the highest probability. This gives you an accuracy of 89.5%.

Hindcasts were made for 1851 10 1877. These hindcasts were verified and the results appear in Table 4. There were 22 years that hindcasts were made. Out of those 22 years there were 5 years when a cyclone did not make landfall in the section of the coast with the highest risk, for an accuracy rate of 77%. Summaries of Table 3and 4are given below.

Years Highest Forecasts

1856	11 years	4 years missed	64%
1867	11 years	1 year missed	91%
1986	10 years	2 years missed	80%
1996	8 years	1 years missed	87.5%

An alternate verification process would be to consider that 30 forecasts of the highest probable section of the US coast could be made in the 19 years. There were 9 times when a storm did not make landfall for one of these forecasts. In other words, there were 9 missed forecasts out of the 30 forecasts with an accuracy rate of 70%. For the hindcasts made from 1856 to 1877, there were 34 hindcasts made in the 22 years of the sections of the United States coast with the highest risk of a tropical cyclone making landfall. There were 7 missed hindcasts for an accuracy rate of 79%.

Total Highest Forecasts

1856	17 forecasts	5 missed	71%
1867	17 forecasts	2 missed	88%
1986	16 forecast	7 missed	56%
1996	13 forecasts	1 missed	92%

Looking at Table 2, climatology would have forecast each year that the West Coast of Florida had the highest chance each year of experiencing a landfall from a tropical storm or hurricane. Based on climatology there is a 71 percent chance of a storm on that section of the coast line. Out of the 40 years there were 10 years when storms did not make landfall on the west coast of Florida for an accuracy rate of 75%.

Highest Climatology Forecasts

1856	11 years	5 years missed	55%
1867	11 years	1 year missed	91%
1986	10 years	4 years missed	60%
1996	8 years	0 years missed	100%

There have been 40 years that forecasts or hindcasts were made. Of those 40 years, there were only 7 years when a tropical storm or hurricane did not make landfall in one of the sections with the highest risk. This gives you an accuracy rate of 83%. There have been 62 forecast made in the 40 years with the highest risk and there have only been 14 times in the 40 years when a tropical storm or hurricane did not make landfall in one of 62 forecast areas. This gives you an accuracy rate of 77%.

WRC’s FORECAST COMPARED TO GRAY’s FORECAST

In order to compare the OCSI with other Atlantic Seasonal hurricane outlooks, secondary predictions where made of the number of tropical cyclones, number hurricanes, number of tropical storm days, and number of hurricane days each year. These forecasts were compared to climatology as well as the forecast made by Colorado State Professor Bill Gray. The years used for this comparison were for the twenty years from 1984 to 2003.

The secondary predictions were made by averaging the number of storms for the years in each phase as well as average the number of tropical storm days and hurricane days. These numbers were used to make the secondary outlook each year.

This verification of the twenty years of forecast demonstrates that WRC’s OCSI model is as accurate if not more accurate than Gray’s model. The advantage of the OCSI model is that WRC’s model can make a prediction years in advance. Table 5 gives a summary of the forecasts comparisons that are shown in Table 6 through 9. This table gives the number of years that each model was closest within the limits indicated.

WRC forecast the number of storms each year and this forecast was verified for the past 20 years. There were eight years out of the twenty years when WRC’s forecast was within plus or minus one storm. Gray’s forecast was only within one storm five of the twenty years and climatology was only within one storm two of the twenty years. This is shown in Table 6. Figure 2 shows a graph comparing the plus or minus error in the forecast versus observed. This graph also shows the forecast error for the forecast made by climatology.

The graph seems to indicate that climatology and WRC’s method has under forecast the number of storms in the Atlantic since 2001. In 1997 WRC forecast the correct number of storms. Other years when WRC forecast the exact number of storms were 1994, 1992, and 1987

WRC forecast the number of hurricanes each year. The forecast verified within one hurricane ten out of the twenty years and Gray’s method forecast within one hurricane eight of the twenty years. WRC forecast the number of hurricane days each year. This verified within five days eight out of the twenty years. Table 7 shows the comparison of the forecast erros.

Gray’s method forecast the number of hurricane days within five days six out of the twenty years and climatology was five out of the twenty years. WRC’s model then forecast the number of storm days, this verified within ten days for ten of the twenty years. Gray’s model forecast the number of days within ten days for seven of the twenty years and climatology five out of the twenty years. Table 8 gives the detail of these comparisons.

Table 9 gives the forecast for both WRC’s and Gray’s model for the number of storm days in each year. WRC’s model forecast the number of days within ten days for ten of the twenty years. Gray’s model forecast the number of days within ten days for seven of the twenty years.

This verification of the twenty years of forecast demonstrates that the WRC’ OCSI model is as accurate if not more accurate than Gray’s model. The advantage of the OCSI model is that WRC’s model can make a prediction years in advance. .

REFERENCES:

Gray, William M., Philip J. Klotzbach, and Chrisopher W. Landsea, 2003, Updated Forecast of Atlantic Seasonal Hurricane Activity and US Landfall Strike Probability for 2003, Dept. of Atmopsheric Science, Colorado State University/

Gray, William M, Christopher W. Landsea, Paul W. Mielke, Jr., Kenneth J. Berry, and Erick Blake, 2000; Summary of 2000 Atlantic Tropical Cyclone Activity and Verification of Aurthor’s Seasonal Activity Forecasts – Appendix A: Verification of Past Seasonal Forecasts, Colorado State University.

Freeman, John C. and Jill F. Hasling, 2003, The Effect of the Orbit of the Sun on the Earth’s Atmosphere, AMS Annual Meeting

Jose, P.D., 1965, Sun’s Motion and Sunspots, Astronomical Journal 10,(1),193-200.

Table 1: Orbital Cyclone Strike Index

Phase
1	1878	1889	1901	1913	1923	1933	1944	1954	1964	1976
2	1879	1890	1902	1914	1924	1934	1945	1955	1965	1977
3	1880	1891	1903	1915	1925	1935	1946	1956	1966	1978
4	1881	1892	1904	1916	1926	1936	1947	1957	1967	1979
5	1882	1893	1905	1917	1927	1937	1948	1958	1968	1980
6	1883	1894	1906	1918	1928	1938	1949	1959	1969	1981
7	1884	1895	1907	1919	1929	1939	1950	1960	1970	1982
8	1885	1896	1908	1920	1930	1940	1951	1961	1971	1983
9	1886	1897	1909	1921	1931	1941	1952	1962	1972	1984
10	1887	1898	1910	1922	1932	1942	1953	1963	1973	1985
11	1888	1899	1911			1943			1974
12		1900	1912						1975

Table 2: OCSI FORECASTS

Phase	Mexico	Texas	Louisiana – Alabama	West Florida	East Florida	Georgia-N. Carolina	East US Coast	GOM Oil Block
1	40	40	60	70	50	90	60	90
2	40	40	70	60	40	60	40	80
3	70	60	40	90	70	20	20	80
4	40	70	90	90	60	70	40	100
5	20	30	60	60	30	60	20	90
6	40	40	70	70	30	60	50	70
7	20	60	60	80	40	30	10	90
8	30	40	50	50	30	70	70	80
9	50	60	40	70	30	40	30	70
10	50	70	50	70	30	60	20	100

>10	25	50	75	63	13	50	25	88
Avg	40	51	59	71	41	56	36	85

Figure 1: Past Storms in Phase 9 by Section of US Coast

The sections with the highest number of storms in past years were West Florida Coast with storms in 8 of the 11 years and Texas with storms in 7 of the 11 years

Table 3: OSCI Verification

	Year	Highest
1	1985	TX 70% W FL 70%
		Kate
2	1986	GA/NC 90%
		Charley
3	1987	LA/AL - 70%
		X
4	1988	W. FL 90%
		Keith
5	1989	LA-AL 90% WFL 90%
		Allison
6	1990	W FL 60% GA-NC 60% LA-AL 60%
		Marco
7	1991	W FL 70% LA-AL 70%
		Fabian
8	1992	W FL 80%

9	1993	NE US 70% GA-NC 70%
		Emily
10	1994	W FL 70%
		Alberto Beryl Gordon
11	1995	TX 70% W FL 70%
		Dean Allison Erin Opal
12	1996	GA-NC 90%
		Arthur – Bertha – Fran
13	1997	LA-AL 70%
		Danny
14	1998	W FL (90%)
		Earl Mitch
15	1999	LA-AL 90% WFL90%
		Harvey Irene
16	2000	W FL 60% LA-AL 60% GA-NC 60%
		Gordon/Helene Gordon Helene
17	2001	W FL 70% LA-AL 70%
		Gabrielle Allison
18	2002	W FL 80%
		Hanna
19	2003	NE US 70% GA-NC 70%
		Isabel Isabel

Table 4: OSCI Verification 1856-1877

	Year	Highest
1	1856	GA/NC 90%
		Strm#3
2	1857	LA/AL - 70%

3	1858	W. FL 90%
		Strm#3
4	1859	LA-AL 90% WFL 90%
		Strm#5 Strm#5
5	1860	W FL 60% GA-NC 60% LA-AL 60%
		Strm#1 Strm#1/4/6
6	1861	W FL 70% LA-AL 70%
		Strm#6 Strm#2
7	1862	W FL 80%

8	1863	NE US 70% GA-NC 70%
		Strm#6 Strm#6
9	1864	W FL 70%

10	1865	TX 70% W FL 70%
		Strm#2/4 Strm#7
>10	1866	LA-AL 75%

	Year	Highest
1	1867	GA/NC 90%
		Strm#1
2	1868	LA/AL - 70%
		Strm#1
3	1869	W. FL 90%

4	1870	LA-AL 90% WFL 90%
		Strm#1 Strm#6/9
5	1871	W FL 60% GA-NC 60% LA-AL 60%
		Strm#6/7 Strm#3 Strm#7
6	1872	W FL 70% LA-AL 70%
		Strm#5 Strm#1
7	1873	W FL 80%
		Strm#3/4/5
8	1874	NE US 70% GA-NC 70%
		Strm#6 Strm#6
9	1875	W FL 70%
		Strm#3
10	1876	TX 70% W FL 70%
		Strm#5
>10	1877	LA-AL 75%
		Strm#14

Table 5. Summary of Model Comparison

# of Storms in Atlantic within 1 storm	WRC OCSI 8 Years	CLIMATOLOGY 2 years	Gray’s Fcst 5 years
# hurricanes in Atlantic within 1 storm	10 years	5 years	8 years
# of hurricane days with 5 days	8 years	4 years	6 years
# of storm days within 10 days	10 years	5 years	7 years

Table 6: Number of Named Storms in the Atlantic

Year	OBS	WRC FCST	WRC Error	Gray APR FCST	GRAY APR Error	CLIMATE Error
1984	12	7	-5	10	-2	-2
1985	11	10	-1	11	0	-1
1986	6	11	5	8	2	4
1987	7	7	0	8	1	3
1988	12	8	-4	11	-1	-2
1989	11	10	-1	7	-4	-1
1990	14	8	-6	11	-3	-4
1991	8	9	1	8	0	2
1992	6	6	0	8	2	4
1993	8	7	-1	11	3	2
1994	7	7	0	9	2	3
1995	19	10	-9	10	-9	-9
1996	13	11	-2	11	-2	-3
1997	7	7	0	11	4	3
1998	14	8	-6	10	-4	-4
1999	12	10	-2	14	2	-2
2000	14	8	-6	11	-3	-4
2001	15	9	-6	10	-5	-5
2002	12	6	-6	12	0	-2
2003	16	7	-9	12	-4	-6

Figure 2: Comparison of the Error on the number of forecast storm.

Table 7: Number of Hurricanes in the Atlantic Basin

Year	OBS	WRC FCST	WRC Error	Gray Apr/Jun	Gray Error	CLIMATE Error
1984	5	4	-1	7	2	1
1985	7	5	-2	8	1	-1
1986	4	5	1	4	0	2
1987	3	4	1	5	2	3
1988	5	5	0	7	2	1
1989	7	6	-1	4	-3	-1
1990	8	5	-3	7	-1	-2
1991	4	6	2	4	0	2
1992	4	3	-1	4	0	2
1993	4	5	1	7	3	2
1994	3	4	1	5	2	3
1995	11	5	-6	6	-5	-5
1996	9	5	-4	7	-2	-3
1997	3	4	1	7	4	3
1998	10	5	-5	6	-4	-4
1999	8	6	-2	9	1	-2
2000	8	5	-3	7	-1	-2
2001	9	6	-3	6	-3	-3
2002	4	3	-1	7	3	2
2003	7	5	-2	8	1	-1

Table 8: Number of Hurricane Days

Year	OBS	WRC FCST	WRC Error Plus/Minus Days	Gray Apr/Jun FCST	Gray Error Plus/Minus Days	Climate Error
1984	18	16	-2	30	12	7
1985	21	21	0	35	14	4
1986	11	24	13	15	4	14
1987	5	7	2	20	15	20
1988	21	25	4	30	9	4
1989	32	30	-2	15	17	-7
1990	27	20	-7	30	3	-2
1991	8	20	12	15	7	17
1992	16	15	-1	15	0	9
1993	10	21	11	25	15	15
1994	7	16	9	15	8	18
1995	62	21	41	25	37	-37
1996	45	24	-21	25	-20	-20
1997	10	7	-3	25	15	15
1998	49	25	-24	20	-29	-24
1999	43	30	-13	40	-3	-18
2000	32	20	-12	25	-7	-7
2001	27	20	-7	25	-2	-2
2002	11	15	4	30	19	14
2003	32	21	-11	35	-3	-4

Table 9: Number of Storm Days in the Atlantic

Year	OBS	WRC FCST	WRC Error Plus/Minus Days	Gray Apr/Jun FCST	Gray Error Plus/Minus Days	Climate Error
1984	51	55	-4	45	6	-2
1985	51	68	-17	55	-4	-2
1986	23	83	-60	35	-12	26
1987	37	47	-10	40	-3	12
1988	47	57	-10	50	-3	2
1989	66	69	-3	30	36	-17
1990	66	58	8	55	11	-17
1991	22	64	-42	35	13	27
1992	39	41	-2	35	-4	10
1993	30	50	-20	55	25	19
1994	28	55	-27	35	7	21
1995	121	68	53	50	71	-72
1996	78	83	-5	55	23	-29
1997	28	47	-19	55	-27	21
1998	80	57	23	50	30	-31
1999	77	69	8	65	12	-28
2000	66	58	8	55	11	-17
2001	63	64	-1	50	13	-14
2002	54	41	13	65	-11	-5
2003	71	50	-21	65	6	-26