South Florida experienced a significant change in land usage during the twentieth century, including the conversion of natural wetlands into agricultural land for the cultivation of winter vegetable, sugar cane and citrus crops. This movement of agriculture from more northerly areas was intended partly to escape the risk of damaging winter freezes. Here we present evidence from a case study using a coupled atmosphere and land-surface computer-modelling system that suggests that the draining of wetlands may have inadvertently increased the frequency and severity of agriculturally damaging freezes in the south of Florida.

On 19 January 1997, a rare freeze inflicted severe damage in agricultural areas of south Florida that were once natural wetlands, with below-freezing temperatures extending to the tip of the peninsula. This event, chosen here for our case study, resulted in losses in the fresh-vegetable and sugar-cane sectors that alone exceeded US$300 million¹. Furthermore, nearly 100,000 migrant farm workers were displaced or unemployed as a result of the freeze².

We used the Regional Atmospheric Modeling System (RAMS)³, a comprehensive meteorological modelling system that includes a sophisticated land-surface scheme to represent the effects of surface properties on atmospheric processes⁴, to investigate the impact of anthropogenic changes in land coverage on this freeze. A pair of simulations was undertaken in which the model configuration was identical, except that in one simulation the data represented pre-1900s (almost natural) land cover, whereas in the other they represented 1993 (near-present-day) land usage. These data sets (Fig. 1a, b) reflect the conversion of natural wetlands to agricultural land in the areas of south Florida that were affected by the freeze.

In key agricultural areas that were once natural wetlands, particularly the areas used for high-density cultivation of winter vegetables, sugar cane and citrus fruits to the south and southwest of Lake Okeechobee, the simulation incorporating current land coverage produced minimum temperatures that were generally colder (Fig. 1c) and were below freezing for a longer period (Fig. 1d) than that using natural land coverage. The results reveal that when land-surface properties were specified to represent natural land cover, a persistent heat flux from wetlands was sufficient to hold the simulated temperature above freezing throughout the night in many of these areas (results not shown).

Figure 1 Minimum temperatures, and the duration of periods of of sub-zero temperatures, in areas cultivated from drained wetlands in south Florida. a, b, Classes of land use specifying model properties for simulations incorporating land-surface conditions before the 1900s (a) and in 1993 (b). NL, needleleaf; BL, broadleaf. c, Difference between the model's simulated minimum temperatures (in 7C) near ground level on 19 January 1997; differences were determined as the values derived from 1993 usage minus those from pre-1900s usage. Locations inside the zero contour experienced colder minimum temperatures when 1993 land use was used in the model. d, Difference in the duration (in min) of subzero-temperature periods for the two different model simulations; differences were determined as in c. Areas inside the zero contour experienced freezing temperatures for longer when 1993 land use was used in the model. [larger image]

* Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado 80523, USA
e-mail: curtis@atmos.colostate.edu

† US Geological Survey, EROS Data Center and NASA/Goddard Space Flight Center, Greenbelt, Maryland 20771, USA