Plentiful documentary and pre-twentieth century instrumental data from Louisiana, U.S.A., provide a record of continuous tropical cyclones, with daily resolution dating back to the late eighteenth century. The reconstruction provided new specific information for 83 storms prior to 1872. Parts of the early and mid-nineteenth century exhibit greater tropical cyclone and hurricane activity than at any time within the last few hundred years. A major hurricane that impacted southeast Louisiana in August 1812 is very likely the closest landfalling hurricane known to impact New Orleans. The longer temporal perspective provides insight on historical hurricane impacts and information on assessing future hurricane mitigation strategies concerning potential worst-case scenarios.
 The modern instrumental record of Atlantic Basin tropical cyclones, extending back to the late nineteenth century, reveals prominent variations at interannual and decadal timescales [Holland and Webster, 2007; Landsea, 2007; Mann et al., 2007]. These variations have profound impacts on society and terrestrial ecosystems, affecting insured property up to tens of billions of U.S. dollars [Clark, 1997]. However, the current Atlantic Basin record is too short to encompass the full range of temporal variability needed to calculate accurate probabilities and recurrence intervals essential for long-range hurricane prediction and hazard assessment. A longer temporal perspective of hurricane activity would be quite reassuring, particularly since the characteristics of climatic forcing mechanisms of the previous centuries, as well as the last few decades, are different, and because increased coastal development and population is likely to continue in conjunction with anticipated future climate change [Emanuel, 2005; Landsea, 2007]. Documentary records from historical archives provide precisely dated daily resolution of tropical cyclones [e.g., García-Herrera et al., 2005, 2007], including a recent compilation of all known early nineteenth century U.S. storms [Bossak and Elsner, 2004]. This paper presents a new unique documentary reconstruction of tropical cyclones for Louisiana, U.S.A. that extends continuously back to 1799 for tropical cyclones, and to 1779 for hurricanes. This is clearly the longest continuous tropical cyclone reconstruction conducted to date for the United States Gulf Coast, with the reconstruction focusing on the addition of newly documented storms from 1799 to 1871 and on new specific knowledge of the intensity of many other storms, augmenting this information with the North Atlantic hurricane database (HURDAT) as it pertains to all Louisiana tropical cyclones up through 2007.
2. Data and Methods
 Detailed weather descriptions of Louisiana tropical cyclones date back to the late eighteenth century. New Orleans served as the state's main center of maritime, urban, and economic activity, enabling the buildup and archival preservation of colonial documentary data (Figure 1). About two dozen different newspapers provide some detailed information of tropical cyclones, as at times they record the exact hourly timing of storm impact, wind direction, wind intensity, rainfall, storm surge, damage to buildings and trees, specifics on geographic extent of damage, ship disasters, and deaths [Ludlum, 1963; Mock, 2004; Chenoweth, 2006]. Generally, newspapers contained more information concerning stronger storms and their impacts. Newspapers at times contained “contributed” letters, appearing in print up to several weeks after the occurrence of a storm from localities spread throughout Louisiana. Most newspapers prior to the mid-1830s appeared in print on a weekly basis, and newspapers thereafter were published on a daily basis, with some even providing two newspaper editions per day. Private diaries, journals, and letters provide the majority of land-based documentary manuscripts from the early 1820s up through the American Civil War. Most of these manuscripts were in the form of plantation diaries, as plantations were a prominent commercial enterprise that required meticulous record-keeping. Also of valuable use, a few hundred ship protest records from the New Orleans Notarial Archives provide unique abundant documentary data on storm impacts in the Mississippi Delta and along the Louisiana coast, dating back to the late eighteenth century. Many contained valuable information on weak tropical cyclones not documented in newspapers and diaries. Occasional instrumental weather records in Louisiana and nearby Mississippi and Alabama, mostly dating after the late 1830s, provided important quantitative weather data that occasionally included barometric pressure. Other meteorological variables observed include temperature, precipitation, wind speed, wind strength, cloudiness, and written remarks were often provided as well. These instrumental data were recorded by weather enthusiasts from personal diaries, the U.S. Army Surgeon General, and the Smithsonian Institution, U.S. Navy and British logbooks, and the U.S. Coastal Survey [Fleming, 1990; Chenoweth, 2006].
 Initially, all Louisiana tropical cyclone frequencies were compiled for each year from the HURDAT database from 1851 to 2007, counting each tropical storm when it was centered within 160 km (100 miles) of the Louisiana coastline and each hurricane within 97 km (60 miles) [Chenoweth, 2003; Mock, 2004]. All available sources of documentary data were used to reconstruct potential tropical cyclones, significantly adding to existing data, especially for the period prior to 1872. This study also utilized some data for seven storms from a previous study which did not focus on Louisiana [Chenoweth, 2006]; yet this information enabled conclusive analyses specific for Louisiana when combined with new data retrieved in this study. In order to conclusively add a storm in the reconstruction, the data must show the classic signs of tropical systems which include sustained strong winds for quite longer than several hours, specific wind directions that indicate the motion of typical tropical systems, and descriptions of damage. Fixed hourly temperature and corrected pressure data to sea level provide valuable information to verify tropical characteristics and estimate intensity respectively. Tropical cyclones were classified into three intensity categories: tropical storm, category 1–2 hurricane, and category 3 or greater (major) hurricane. Building construction techniques vary continuously in time, which prevents direct comparisons of historical damages done to structures over time [Landsea et al., 2004]. Thus, a conservative content analysis approach was emphasized to estimate the intensity category of each storm, emphasizing historical characteristics such as tree damage and specific details of storm surge consistent with modern damage, described in the Saffir-Simpson scale as they relate to Louisiana coastal locations. Historical tidal tables were consulted in order to precisely calculate storm surge estimates. Generally, most storms after about 1800 were described in well over twelve different historical accounts. Prior to 1800, each storm was described by about a half dozen different sources. Any conflicting descriptions of the same event by reliable sources led to the exclusion of a candidate storm from the final list.
3. Results and Discussion
 The reconstruction provides new specific information for 81 tropical cyclones covering the period 1779–1871 that impacted Louisiana. Tropical storm frequencies found in this study from 1799 to 1871 indicate a significant new number of storms added to previous known storm counts, as well as a few storms that are clearly not tropical in character, especially for the pre-1851 (pre-HURDAT) period (Table 1). Comparisons of tropical cyclone reconstructions from Louisiana documentary data with an independent newspaper data set from the New York Times from HURDAT in the 1851–1871 period shows complete agreement for 14 of the 15 “known” storms (the single disagreement for a storm in 1870 clearly had no Louisiana impact), implying that the documentary reconstructions are very reliable. Results illustrate a seasonal maximum during September when considering all of the reconstructed storms, but the historical data also successfully reconstructed numerous early June and late October storms, and an early November storm.
Table 1. Comparisons of Louisiana Tropical Storm Frequencies as Grouped in Three Different Time Periods From 1799 to 1871
 The continuous Louisiana tropical cyclone record from 1799 to 2007 consist of 192 tropical cyclones, with 75 of these storms being at hurricane strength. Dividing the number of years by these storm totals for both all tropical cyclones and for hurricanes, reveal that Louisiana generally experiences a tropical cyclone impact every 1.09 years and a hurricane impact every 2.79 years (Figure 2). The most frequent annual tropical cyclone impacts occurred in 1834 and 2002, with four tropical cyclones affecting Louisiana for both years. The most frequent hurricane impacts occurred in 1860, numbering three storms. Ten-year running sums of Louisiana tropical cyclones reveal some cyclical behavior, with active periods centered around the period since 2000, the 1930s, the 1890s, and the late 1820s/early 1830s. The 10-year running sums for Louisiana hurricanes show similar active periods, with the exception of the 1840s and 1930s that look relatively quiet. The 1820s/early 1830s and the early 1860s are the most active periods for the entire record.
 The only other complete and continuous tropical cyclone reconstructions for the Atlantic Basin available for comparison with the Louisiana record at subannual resolution are from South Carolina and Jamaica, though several shorter and incomplete documentary records are available from scattered areas elsewhere in the Atlantic Basin [Chenoweth, 2003; Mock, 2004; Mock et al., 2004]. Some sediment and tree ring records also provide valuable information on tropical cyclone activity for comparisons with historical records. [e.g.,Nyberg et al., 2007]. The South Carolina historical record reveals active periods that are generally not in phase for the most part with those in the Louisiana record, with the exception perhaps of the active 1890s. South Carolina experienced an active tropical cyclone period in the 1830s that is later than the active period in the Louisiana record, which is broadly consistent with other archival evidence of higher tropical cyclone activity for the Bahamas and several other areas of the Caribbean [Walsh and Reading, 1991]. The Jamaican record of both tropical cyclone and hurricane counts, do not reveal specific similar results of active decades as either the Louisiana or the South Carolina record. The relatively inactive period in the Louisiana hurricane record from 1920 to 1940 also appears in the South Carolina record but not for Jamaica. A lagoonal sediment record from Puerto Rico of major hurricanes reveals an active period in the last 250 years [Donnelly and Woodruff, 2007], but the higher activity may date prior to 1800. A marine sediment record, representative of major hurricanes in the tropical Atlantic, illustrates that the 1810s were active that is consistent with the Louisiana record [Nyberg et al., 2007]. However, their results of less-active periods around the 1820s and 1830s, as well as of the active period in the 1840s, is not clearly in phase. A tree ring isotope proxy from southern Georgia also reveals that the 1840s was quite active, suggesting that regional variations of tropical cyclone activity occur over time [Miller et al., 2006].
 Differences in temporal periodicities of active and less-active periods between regions are likely indicative of different synoptic circulation controls that affect tropical cyclone activity and tracks. Louisiana storm formation is related somewhat with tropical origin-only storms that can form in the Gulf of Mexico and Caribbean, but likely includes more baroclinic-enhanced storms that originate in the Gulf of Mexico and also the western Atlantic. Mid-tropospheric flow related to the westerlies can also impact storm tracks in the Louisiana area. Louisiana storms are associated with a lesser number of storms related specifically to those that form in the central and eastern Atlantic off West Africa. Although these storms that form off West Africa are influenced by the strength of the Azores-Bermuda High system, their steering mechanisms near Louisiana are not usually associated directly with this high-pressure system.
 Early instrumental records, ship logs and protests, newspapers, and diaries that include wind direction and relative wind speed provided valuable information on reconstructing the tracks of prominent major hurricanes in 1812 and 1831 that made landfall within about 97 km of New Orleans (Figure 1 and Table 2). The density of data for both storms are concentrated along and offshore the Louisiana coast, and data are clearly copious enough to construct tracks that are as precise as for most storms in the current HURDAT nineteenth century database. For example, the data quantity and density for the 1812 and 1831 storms are clearly greater than those used to construct the track of the Last Island Hurricane of 1856 (consisting of nine locations off the coast) in Louisiana [Landsea et al., 2004].
Table 2. Number of Different Documentary Records That Contributed to the Analyses of the Major Hurricanes of 1812 and 1831 Within the Region Shown in Figure 1
Personal diaries and letters
 The 1831 storm has been termed as the Great Barbadoes hurricane, which impacted the Barbadoes, Haiti, Cuba, and eventually made landfall on 17 August 1831 near Grand Isle, Louisiana [Ludlum, 1963]. Marine reports and an instrumental record at Key West, Florida clearly make the connection of this storm's track from Cuba to the Louisiana coast. It followed a very similar track to the northwest after landfall as Hurricane Betsy in 1965. New Orleans received a storm surge of over ten feet resulting from an overflow of Lake Pontchartrain, but the city likely escaped direct category 3 impact. However, a major hurricane that made landfall near Breton Island, Louisiana on 19 August 1812 traversed westward toward New Orleans, and traversed to the east of Natchez, Mississippi, likely still at hurricane strength. It is likely the closest major hurricane to pass the city in the last few hundred years when considering all major hurricanes in the area from the historical record (Figure 1). This storm followed an earlier nearby landfalling hurricane in June that already weakened the levees, but descriptions left no doubt of major hurricane impact in the New Orleans area and to the south. Some descriptions to the south of New Orleans, as well as west of the track (thus not in the greatest area of storm surge) estimate that the Mississippi River must have rose as high as 18 feet.
 The Louisiana tropical cyclone and hurricane reconstructions presented here indicate that the modern records which cover just a little over a hundred years is too short to provide a full spectrum of tropical cyclone variability, both in terms of frequency and magnitude. Some active periods in the nineteenth century have not occurred during the recent rise of coastal population and development over the past several decades. Understanding the full range of variability is vital for long-range planning for hurricane preparedness and future expectations, as dramatic population and development has increased their hurricane vulnerability for much of the Gulf and Atlantic coasts. If a higher frequency of major hurricanes occurred in the near future in a similar manner as the early 1800s or in single years such as in 1812, 1831, and 1860, would have devastating consequences for New Orleans, perhaps equaling or exceeding the impacts such as in hurricane Katrina in 2005. Higher frequencies of hurricane and tropical cyclone for several decades would cause dramatic impacts from coastal erosion and more profound ecological changes such as those involving increased saltwater intrusions on plant communities. Long and continuous documentary reconstructions of tropical cyclones from other locations in North and Central America, dating back several hundred years with sources from various archives internationally [Chenoweth, 2006; García-Herrera et al., 2007], can reveal more potential worse-case scenarios and a network of such records will enable us to further understand the mechanisms behind these scenarios.
 I gratefully acknowledge Mike Chenoweth for his insight and sharing some data, Isabel Altamirano, Ann Wakefield, and Howard Margot for help in archival data, and Stephanie Dodds, David Glenn, Shelley Holmberg, Richard Murphy, James Tanis, Henry Young, and Chris Landsea for some help in various stages of the research. This research is supported by NSF grant ATM-0502105.