Furthermore, the relationship posed by this hypothesis seems like it might be feasible when comparing the the Tropical Storm Intensity Zones map and the Proportional Economic Loss Risk Deciles in Florida as a Result of Hurricanes (1980-2000) map. To test this hypothesis, I began by performing a spatial autocorrelation for whole state in order to confirm that there was indeed clustering of polygons with similar proportional economic loss risk.

Proportional Economic Loss Risk for whole State

        From the output of the analysis, it was found that proportional economic loss risk deciles were very highly patterned, and that this patterning was statistically significant. However, I felt it would also be wise to perform a spatial autocorrelation for each of the portions of proportional economic loss risk which fell within each of the five storm zones, to see if there was more or less cluster depending on which storm zone an area was subject to.

Proportional Economic Loss Risk forStorm Zone 1 of Florida

Proportional Economic Loss Risk forStorm Zone 2 of Florida

Proportional Economic Loss Risk forStorm Zone 3 of Florida

Proportional Economic Loss Risk forStorm Zone 4 of Florida

Proportional Economic Loss Risk forStorm Zone 5 of Florida

        The results of these spatial autocorrelations are telling. They indicate that every grouping of proportional economic risk loss by storm zone is highly significant. This is indicated by Moran’s I Index values which are near to zero.

        Therefore, I investigated the spatial autocorrelation analysis by creating histograms for each of the portions of proportional economic loss risk which fell within each of the five storm zones. If a relationship between storm zones and proportional economic loss existed, I felt it should be apparent. I began by producing a histogram of proportional economic loss risk, divided among gridcode deciles, for all of Florida, in order to have a basis for viewing how individual regions differed from the state norm.

         For the state, the most commonly occurring gridcode is 6, indicating above average economic loss risk as a proportion of GDP per analytical unit. The distribution appears like it might be bi-modal, which may indicate that there are geographical regions in Florida which are associated more dominantly with certain levels of risk.

         In the north central portion of Florida, the gridcode occurring with the highest frequency, the mode, is 7, indicating high proportional economic loss risk. The count of gridcodes increases sharply at 5, and then steadily increases to 6, reaching an apex at 7 and then steadily declining at 8 and 9.

         This histogram covers Florida’s northwest, as well as northeast and central southeast coasts. The gridcode occurring with the highest frequency is 6, indicating above average proportional economic loss risk although a secondary peak occurs at gridcode 3. Were it not for the relatively low count of gridcode 4, this histogram might have a close to normal distribution.

         The central portion of the state, and northwest coast, coincide with the third storm zone, where there is a 10% chance that a category 3 hurricane will annually occur. The gridcode occurring with the highest frequency is 5, indicating average proportional economic loss risk. The distribution is highly abnormal.

         Most of southern Florida is encompassed by the fourth storm zone. The gridcode occurring with the highest frequency is 9, massive proportional economic loss risk. This histogram seems to have three clusters: gridcodes 1 to 4 with low frequency, 5 to 7 with moderate frequency, and 8 to 10 occurring most often.

        The very tip of Florida encompassed by storm zone is dominated by the gridcode 10. This indicates areas in which proportional economic loss risk is the greatest relative to all other areas.

         The results are interesting. In agreement with my hypothesis, parts of Florida coinciding with the fifth zone of tropical storm intensity are primarily (over 30%) composed of areas with the highest proportional economic loss risk (gridcode 10). Parts of Florida coinciding with the fourth zone of tropical storm intensity follow suit by being dominated by a majority of gridcode 9. However, the dominant gridcodes in zones 1, 2 and 3 do not support my hypothesis. Rather than being dominated by a low gridcode, areas of Florida in zone 1 had the third highest dominant gridcode. Also, rather than a trend of low towards high gridcode dominance as one moves from north to south, instead, the dominant gridcode decreases, between zones 1, 2 and 3.

        My second hypothesis states that risk of human mortality and property damage is greater in urban areas, where there are greater concentrations of people and infrastructure, compared to less developed and less populous non-urban areas. To assess the validity of this statement, I compared histograms of mortality risk and proportional economic loss risk for urban areas of Florida to that of non-urban areas.

         Mortality risk in urban areas of Florida appears to be close to a normal distribution, but skewed to the left. The most frequently occurring gridcode is 7, indicating a high mortality risk.

         For non-urban areas of Florida, the distribution of gridcodes appears to be close to normal. In non-urban areas, the most common gridcode is 4, representing areas of below average mortality risk.

         In urban areas, the dominant gridcode is 5, indicating average proportional economic loss risk. This distribution is not normal.

        Non-urban areas of Florida tended to have greater proportional economic loss risk, with the majority of gridcodes being 6, indicating above average risk.

         In the first pair of histograms, it seems that the findings corroborate the first part of my second hypothesis, that mortality risk from hurricanes is greater in urban areas compared to non-urban ones. However, the second pair of histograms appears to disagree with the second part of the hypothesis that proportional economic loss risk is greater in urban areas rather than non-urban areas.

         The third hypothesis is that human mortality and property damage more readily occur in urban areas at lower elevations than at higher elevations. Coastal areas tend to be heavily populated and to support those populations, costly infrastructure and amenities are necessary. A visual inspection of the Urban Elevation maps appears to support this theory, so some analyses are in order. First, I produced a scatter plot of mortality risk by elevation for all urban areas in Florida.

         There are many ways one can interpret this scatter plot, and the results are not altogether clear. All gridcodes of mortality risk persist from zero to about 120 feet in elevation, and every gridcode except 10 are present even above 160 feet above sea level. However, gridcodes 4, 5, 6, and 7 occurred above 180 feet, and some over 260 feet. From this scatter plot, all urban elevations close to sea level, such as less than 120 feet in elevation, are subject to the full range of mortality risk. Urban areas above 180 feet in elevation are exclusively exposed to average and above average levels of mortality risk.

        The scatter plot of proportional economic loss risk by urban elevation reveals that below 180 feet above sea level, urban areas are subject to all gridcodes. However, above 210 feet in elevation, urban areas are exposed preferential to gridcodes 6 through 10.

        These two results signify that moderate mortality risk appeared exclusively above 180 feet, and high, very high, and extreme potential economic loss risk appeared solely above 210 feet above sea level. To further explore the mortality scatter plot, I created box plots of each mortality gridcode, in order to better map the vertical distribution of each level of mortality risk.

        Among the lowest levels of mortality risk, three-quarters of the urban areas lie below 90 feet in elevation, while the remaining one-quarter lies above.

         Three-quarters of the mortality risk lies at or below 90 feet, while the remaining one- quarter lies above.

         Three-quarters lies below 85 feet, while the remaining one-quarter lies above.

         Three-quarters lies below 90 feet, while the remaining one-quarter lies above.

         Three-quarters lies below 110 feet, while the remaining one-quarter lies above.

         Three-quarters lies below 110 feet, while the remaining one-quarter lies above.

         Three-quarters lies below 75 feet, while the remaining one-quarter lies above.

         Three-quarters lies at or below 65 feet, while the remaining one-quarter lies above.

         Three-quarters lies below 30 feet, while the remaining one-quarter lies above.

         Three-quarters lies below 65 feet, while the remaining one-quarter lies above.