VOLUME 4 JOURNAL OF HYDROMETEOROLOGY DECEMBER 2003

Hydrometeorology and Variability of Water Discharge and Sediment Load in the Inner Gulf of , Western Caribbean

DEEPTHA THATTAI Department of Geological Sciences, University of South Carolina, Columbia, South Carolina

BJOÈ RN KJERFVE Department of Geological Sciences and Marine Science Program, University of South Carolina, Columbia, South Carolina

W. D. H EYMAN The Nature Conservancy, Punta Gorda,

(Manuscript received 21 December 2001, in ®nal form 19 February 2003)

ABSTRACT The hydrological and meteorological characteristics of the watersheds of the inner in the western Caribbean, including runoff, sediment load and yield, and the effects of the El NinÄo±La NinÄa cycle, are examined using available data. The inner Gulf of Honduras, bordered by the second-longest complex in the world, the MesoAmerican Barrier Reef, receives runoff from the watersheds of 12 rivers with a total simulated annual discharge of 1232 m3 sϪ1. Expanding agricultural and industrial activities contribute to the in¯ux of sediments, nutrients, and pollutants from these rivers, leading to increased threats to the health of the reef ecosystem. The watersheds of the Moho, SarstuÂn, and Polochic-Dulce Rivers receive more than 4000 mm of rainfall annually and are major sources of discharge and sediment load, along with the Motagua and Ulua, farther to the east. The drainage basins are characterized by runoff ratios of 0.30±0.55 and simulated sediment yields as high as 869 t kmϪ2 yrϪ1. The results from two different sediment load/yield models agree to within Ϯ2.3% at the 95% con®dence level. Sediment load estimates increase by as much as 5 times on model comparisons of present land use to increased land use. Time series of precipitation for the inner Gulf of Honduras exhibit bimodal distribution with maxima in May±June and in September±October. Analysis of long-term climatic data reveals only a weak but measurable correlation with El NinÄo±La NinÄa. The Southern Oscillation index explains on average 7%±15% of the precipitation and temperature variability for the inner Gulf of Honduras.

1. Introduction longer, whereas human in¯uences on decadal scales or Large-scale atmospheric circulation and local pro- shorter have signi®cantly accelerated changes in the lo- cesses in¯uence climate, which is the state of weather cal hydrology during the last ®ve centuries (Kaczmarek averaged over months or years (Rasmusson et al. 1992). et al. 1996). Local controls on climate include topography, vegeta- The objective of this paper is to assess the hydrolog- tion, land±ocean interaction, land use, and catchment ical and meteorological characteristics of the watersheds geology. Climatic processes are characterized by tem- of the inner Gulf of Honduras (iGoH) in the western perature and rainfall variability and affect humans di- Caribbean. Runoff mean and variability, sediment load rectly through in¯uence on soil, agriculture, hydrolog- and yield, and the effect of the El NinÄo-La NinÄa cycle ical processes, and water availability. At least several are examined using available data. The iGoH is of par- cycles of extreme weather phenomena, such as tropical ticular interest, because it borders the most extensive storms, droughts, ¯oods, and El NinÄo±La NinÄa, must coral reef complex in the Atlantic Province, the be averaged to establish the long-term climate (Kund- MesoAmerican Barrier Reef System (MBRS; Heyman zewicz et al. 1993; Kousky et al. 1984; Ropelewski and and Kjerfve 2001). The iGoH has more than a dozen Halpert 1987; Rogers 1988; Glantz 1997). Regional and designated, protected, coastal marine conservation ar- local climates vary with time, on scales of centuries and eas. These are among the premier ecotourism destina- tions in the Western Hemisphere (Perkins and Carr 1985). Corresponding author address: Dr. BjoÈrn Kjerfve, Marine Science Program, University of South Carolina, Columbia, SC 29208. Watershed boundaries and regional weather systems E-mail: [email protected] do not respect man-made sociopolitical boundaries.

᭧ 2003 American Meteorological Society 985

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC 986 JOURNAL OF HYDROMETEOROLOGY VOLUME 4

FIG. 1. Map showing the watershed (light gray) of the inner Gulf of Honduras. Locations with measurements are Corozal (precipitation), Belize Airport (precipitation, temperature), Central Farms (precipitation), Punta Gorda (pre- cipitation), La Ceiba (precipitation), City (precipitation, temperature), and Kingston (precipitation, tem- perature).

Hence, any impacts in the interior reaches of a watershed pollutants, and nutrients in the reef-associated waters; in one country are bound to have repercussions in neigh- and in predicting the impacts of climate change and boring countries and receiving water bodies. The global warming on the reef ecosystem. MBRS, which receives runoff from Belize, Guatemala, Honduras, and Mexico, is thus dependent on the wa- 2. Geographical setting tershed processes in these other countries. Runoff comes from regions with extensive banana plantations, aqua- The iGoH, measuring 10 000 km2, is situated in the culture ponds, new roads, and otherwise altered land northwestern (Fig. 1). It is bordered by use (Gibson et al. 1998; LoÂpez and Scoseria 1996), Belize, Guatemala, and Honduras, and extends from Dan- yielding high levels of sediment and pollutant discharg- griga (Stann Creek), Belize, to La Ceiba, Honduras. The es into the coastal fringes of the iGoH, which impact 42 408 km2 watersheds that border the iGoH are princi- the coral reef and its biota (Gibson et al. 1998; Katz pally drained by the Moho, SarstuÂn, Polochic-Dulce, Mo- 1989). Maintenance of healthy conditions along the tagua, and Ulua Rivers (Fig. 2). The iGoH includes the MBRS complex is important for biodiversity conser- southern portion of the MBRS complex. The principal vation and local economic development. Thus, the coastal cities are Belize City (population 56 000), Belize, MBRS is a focal area for local and international con- just north of the de®ned area; Livingston (population servation organizations (e.g., The Nature Conservancy 40 000) and (population 100 000), Gua- and the World Wildlife Fund) and has recently attracted temala; and Puerto CorteÂs (population 100 000), Hon- several multimillion-dollar bank projects to the region. duras. Belize City, Big Creek (Belize), Puerto Barrios, Knowledge of the underlying hydrometeorological var- and Puerto CorteÂs are the major commercial ports. iability is an important component in assessing reef vi- The watersheds of the iGoH extend from high moun- ability and impacts on ¯ora and fauna; and also in un- tains to the coast. The smaller rivers are con®ned to an derstanding oceanic circulation, transport of sediments, elevation no more than 100 m above sea level, whereas

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC DECEMBER 2003 THATTAI ET AL. 987

FIG. 2. Isohyets of mean annual precipitation (mm) in the watersheds of the inner Gulf of Honduras. the larger rivers originate at 2500±3000-m elevation and and Fernandez-Partagas, 1997; Neumann et al. 1993) have average elevations above 500 m. The rapid ex- reveals that this is less than 5% of the total number of pansion of agricultural operations in Belize is restruc- Atlantic hurricanes in the same period. Most storms pass turing the watershed dynamics (Stednick et al. 1995). well to the north of the iGoH. Still, the effects of hur- The expansion of citrus and shrimp aquaculture (in Be- ricanes are considerable. The most severe hurricane to lize) and banana plantations (in Guatemala and Hon- have impacted both the Caribbean and iGoH is Mitch duras), and the accompanied road building, have led to (28 October±5 November 1998, category 5), which increased erosion of land, and consequently, more sed- damaged parts of Honduras, Nicaragua, El Salvador, and iment is carried by rivers to the coast. Other sources for Guatemala, and resulted in more than 9000 fatalities. sediment accumulation are the destruction of man- The second-worst hurricane in terms of damage was Fi® groves, mining, and dredging activities (Gibson et al. (14±22 September 1974, category 2). The main damage 1998). The rivers also transport chemicals, herbicides/ from Mitch and Fi® was due to land slumping caused pesticides, and other pollutants as they ¯ow through by heavy rains. Fi® traveled west across the Caribbean agricultural lands, industrial zones (sugar re®neries, cit- rus processing plants, and metal industries), and zones and made landfall near Placencia, Belize, after passing of high population. These pollutants eventually enter between the Bay Islands and the northern coast of Hon- the MBRS, and threaten the and reef systems duras. (26±31 October 1961, category and impact ®shing resources (YaÂnÄez-Arancibia et al. 5) destroyed Belize City (Heyman and Kjerfve 2001), 1999). An excess of nutrients leads to increased algal and Hurricane Iris (5±9 October 2001, category 4) dev- growth, which eventually kills corals. Rapid sedimen- astated several coastal towns in central and southern tation likewise leads to coral die-off (McField et al. Belize. Hurricane-caused damages to the coast, reefs, 1996). and cays usually persist for more than a de- Hurricanes are a constant threat in the Caribbean, cade. Many shallow reefs were severely degraded as a although only 12 hurricanes have made landfall in the result of Mitch (1998), whereas the battering waves and iGoH during the past 50 yr. Analysis of data (Rappaport storm surge accompanying Iris (2001) resulted in severe

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC 988 JOURNAL OF HYDROMETEOROLOGY VOLUME 4

TABLE 1. Calculated correlation coef®cients between precipitation and the Southern Oscillation Index (SOI) and temperature and SOI based on mean monthly data. Punta Gorda, La Ceiba, and Guatemala City lie within the drainage basin. Low rainfall is associated with the warm phase (El NinÄo) and high rainfall is associated with the cold phase (La NinÄa). Correlation coef®cient* Elevation Years Precipitation Precipitation vs SOI (m) of data (mm) JFM AMJ JAS OND Belize International Airport 5 51 1871 0.06 Ϫ0.01 Ϫ0.17 0.24 Central Farms 90 32 1598 0.20 Ϫ0.41 0.16 Ϫ0.05 Punta Gorda 6 39 3710 0.14 Ϫ0.28 Ϫ0.31 Ϫ0.33 La Ceiba 26 36 2975 Ϫ0.28 Ϫ0.05 Ϫ0.10 Ϫ0.37 Kingston 14 48 814 Ϫ0.44 Ϫ0.11 0.47 0.29 Guatemala City 1496 35 1178 Ϫ0.39 0.32 0.23 0.23 Temperature Temperature vs SOI (ЊC) Belize International Airport 5 56 26.4 Ϫ0.02 Ϫ0.28 0.05 Ϫ0.05 Central Farms 90 11 24.6 Ϫ0.26 Ϫ0.33 0.07 Ϫ0.28 Punta Gorda 6 6 24.9 Ϫ0.64 0.01 Ϫ0.35 0.02 La Ceiba 26 6 28.9 Ϫ0.93 Ϫ0.54 Ϫ0.11 0.11 Kingston 14 48 27.6 Ϫ0.47 Ϫ0.47 Ϫ0.48 Ϫ0.57 Guatemala City 1496 35 18.4 Ϫ0.34 Ϫ0.23 Ϫ0.52 Ϫ0.28

* JFM: Jan±Feb±March; AMJ: Apr±May±Jun; JAS: Jul±Aug±Sep; OND: Oct±Nov±Dec. damage to coastal infrastructure and vegetation (Bood were used. The precipitation and temperature anomalies 2001). were calculated as the difference between the precipi- tation/temperature of each month and the monthly mean over the length of each dataset. The time series of anom- 3. Data and analysis alies were then ®ltered (low pass) using a Butterworth a. Meteorological data ®lter (Middleton 2000) to remove seasonal trends. The Our analysis of the hydrometeorology of the iGoH is same procedure was followed for SOI and correlation based on the analysis and synthesis of available tem- coef®cients were calculated between precipitation and perature and precipitation data. The temperature and SOI, and temperature and SOI. The results are sum- precipitation data were analyzed for seven stations in/ marized in Table 1. near the iGoH: Corozal, Belize International Airport, Central Farms, Punta Gorda (all in Belize), La Ceiba b. Discharge modeling (Honduras), Guatemala City (Guatemala), and Kingston (Jamaica) (Fig. 1). Long-term (30 yr or more) data were There is no consistent gauging of the rivers discharg- obtained from 1) the World WeatherDisc CD (Spangler ing into the iGoH. Also, lack of long-term temperature and Jenne 1990), which is part of the World Monthly and precipitation data from stations within each of the Surface Station Climatology (WMSSC), for data until many watersheds makes it impossible to calculate the 1990; and 2) the University Corporation for Atmo- temporal variability of discharge and sediment load spheric Research (http://dss.ucar.edu/datasets/ds570.0/) within a year and between years. Hence the calculation for data from 1991 to 1998, also part of the WMSSC. of discharges for the regional rivers was limited to an The data for the seven stations span different periods; empirical water balance model (Schreiber 1904; Kjerfve the precipitation data overlap for at least 28 yr (1952± 1990), using annual precipitation and temperature data 79), but the temperature data is sparser with La Ceiba to estimate a mean or annual runoff ratio: and Punta Gorda having only 6 yr of data and Central Farms 11 yr. The other stations which overlap for at ⌬ fe least 31 yr (1954±85) lie outside the iGoH watershed. ϭ exp0 , (1) The Southern Oscillation index (SOI) data were ob- rr΂΃ tained from the U.S. National Weather Service's Climate where ⌬f is runoff (mm yrϪ1), e is potential evapo- Prediction Center Web site (http://www.cpc.ncep.noaa. 0 transpiration (mm yrϪ1), and r is precipitation (mm gov/data/indices/). The land cover data was obtained yrϪ1). The local annual potential evapotranspiration is from the Digital Atlas of Central America CDs prepared by the Center for the Integration of Natural Disaster a function of the mean temperature, and thus the latitude Information (CINDI 1999). In exploring the regional and elevation; and is expressed empirically as relationship between precipitation/temperature and the 4.62 ϫ 103 SOI in the iGoH, the monthly precipitation and tem- e ϭ 1.0 ϫ 109 exp Ϫ (2) perature data from the seven stations mentioned earlier 0 ΂΃T

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC DECEMBER 2003 THATTAI ET AL. 989

TABLE 2. Explanation of the variables used in the RUSLE equation and their units. The values are assigned based on an assessment of the individual polygons. The slope-length and slope steepness factors vary for each polygon and are speci®ed in the GIS grid. Variable De®nition Units Values used A Annual average soil erosion expected on ®eld thaϪ1 yrϪ1 Ð slopes R Rainfall-runoff erosivity factor, function of MJ mm haϪ1 hϪ1 yrϪ1 1649, 1872, 2221, 3519 storm energy and intensity (based on intensities of 5.1, 5.6, 6.4, 8.9 mm hrϪ1) K Soil erodibility factor, a measure of the soil thahhaϪ1 MJ Ϫ1 mmϪ1 0.14, 0.24, 0.25 for sand, properties sandy loam, and clay loam L Slope-length factor Nondimensional Ð S Slope steepness factor Nondimensional Ð C Cover-management factor, a measure of the land Nondimensional 0.003, 0.1, 0.13, 0.3 use P Support practice factor, a measure of best man- Nondimensional 1 agement practice

(Holland 1978), where T is surface temperature (K). The to estimate the K values were obtained from the CINDI discharge q (m3 sϪ1) of each river was then calculated as dataset (CINDI 1999). The slope-length factor is given by L ϭ (␭/22.1)m, where ␭ is the horizontal projection n ⌬ fri of the slope length (m). The value m is a variable slope- q ϭ a , (3) ͸ i 9 length exponent. The value of m depends on the slope iϭ1 ΂΃r i 2.592 ϫ 10 angle. The slope steepness factor S is a measure of the 2 where a (m ) is the subarea of a basin represented by effect of the slope on erosion. The values ␭,m,and S a meteorological station and n is the total number of for each grid cell of the watershed were calculated from polygons for each basin (Kjerfve et al. 1997). Similar the Digital Elevation Model (DEM; GLOBE Task Team estimates have previously been simulated for southern et al. 1999), using ArcView routines. Belize by Heyman (1996) and Heyman and Kjerfve The cover-management factor, C, re¯ects the effects (1999) and the model has been veri®ed with good suc- of cropping and management practices on erosion rates. cess for Rio Grande, Belize, based on a comparison to Simulations have been run with C values of 0.003, 0.10, limited gauging data. 0.13, and 0.30 corresponding to forested land, grazing land, irrigated land, and plantations, respectively (Shen c. Sediment load modeling I and Julien 1992). The support practice factor, P, is the ratio of soil loss The annual average erosion was ®rst calculated using with implemented support practice, such as contouring, the Revised Universal Soil Loss Equation (RUSLE) strip cropping, and terracing, to straight-row farming (Renard et al. 1997, chapter 1, p. 15). The RUSLE model along a slope. A value of 1 for P has been used in cases has had a number of successful validations and predic- where there is no information available on conservation tions within the United States, and is increasingly being practices, or when the model is used on larger watershed applied to tropical regions outside the United States as basins rather than smaller agricultural plots similar to more and more data become available in these regions the standard RUSLE plot. The value P is also assumed (El-Swaify and Dangler 1982; Millward and Mersey to be one when the model is applied to nonagricultural 1999; Kinnell 2000). RUSLE calculates the erosion A areas, since it is a ratio of soil loss before and after Ϫ1 Ϫ1 (ton ha yr ) (ha: hectare) expected on ®eld slopes as conservation practices on agricultural lands (Boggs et A ϭ R ϫ K ϫ L ϫ S ϫ C ϫ P, (4) al. 2001). The main agricultural industries in the iGoH watershed (citrus and banana farming) do not apply any where R is the rainfall-runoff erosivity factor (MJ mm conservation strategy. Thus, in this study, P has been haϪ1 hϪ1 yrϪ1), K is the soil erodibility factor (t ha h Ϫ1 Ϫ1 Ϫ1 assumed to be one throughout (Millward and Mersey ha MJ mm ; Table 2). Only A, R, and K in the 1999). equation have units. The rainfall erosivity factor R is The average annual sediment yield (t kmϪ2 yrϪ1)is computed as a product of the total storm energy and the obtained by multiplying the annual average erosion val- maximum 30-min intensity, summed over the storms ues by the sediment delivery ratio (SDR) since only a occurring through the year. The rainfall intensity data portion of the eroded material ultimately reaches the for the Belize watersheds comes from daily measure- river mouth. The SDR is computed from the equation ments provided by the Belize Meteorological Service. Average intensity values for the Guatemalan and Hon- Ϫ0.3 SDR ϭ 0.41 ϫ A t , (5) duran watersheds were used to calculate R (Mikhailova 2 et al. 1997; Portig 1976). The soil erodibility factor, K, where At is the drainage area of the basin (km ) (Shen is a measure of the soil properties. The soil data needed and Julien 1992). The annual average sediment load

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC 990 JOURNAL OF HYDROMETEOROLOGY VOLUME 4

FIG. 3. Isotherms of mean annual temperature (ЊC) in the watersheds of the inner Gulf of Honduras.

(t yrϪ1) is obtained by multiplying the sediment yield tain ridge and the intertropical convergence zone (ITCZ; with the drainage basin area. Nieuwolt 1977; Martyn 1992). Along the coast of the iGoH, the climate is tropical with an average June± d. Sediment load modeling II August air temperature of 27ЊC, and an average Janu- Since there are no actual sediment load measurements ary±March air temperature of 24ЊC. The annual average available, we used a second model to determine sedi- temperatures in the iGoH basins are shown in Fig. 3. ment loads for comparative purposes. This model is The precipitation follows a north±south gradient, in- based on calculating the sediment load using the basin creasing towards the south. Easterly waves bring intense area, relief, and mean annual temperature (Morehead et rains to the coast of southern Belize, where the annual rainfall exceeds 4000 mm. However, rainfall is reduced al. 2003). The long-term mean sediment load Qs (kg sϪ1) is given as along the Honduran coast of the iGoH, because it trends Ϫ5 3/2 1/2 0.0578 ϫ T east±west, almost parallel to the prevailing wind direc- Qs ϭ 2 ϫ 10 ϫ R ϫ A ϫ (0.2 ϫ 10 ), (6) tion. The isohyets of mean annual precipitation (Fig. 2) over the watershed of the iGoH indicate regions of pro- 2 where R is basin relief (m), A is basin area (m ), and nounced maxima along the coast in southern Belize and T is mean annual basin temperature (ЊC). This model is Guatemala (adapted from Portig 1976). On the other much easier to apply as compared to the RUSLE model hand, the interior of the valley is semi- and does not require the detailed polygon-based data of arid and lies in a rain shadow. the RUSLE approach. The results from the two models were compared. The monthly precipitation and temperature data for the regional meteorological stations (Figs. 4a,b) are mostly similar. The monthly temperature variability is 4. Results and discussion small for all six stations with the temperature in Gua- a. Temperature and rainfall temala City being signi®cantly lower because of the The climate of the iGoH is controlled by the easterly higher elevation (1500 m). Precipitation exhibits max- trade winds and their interactions with the central moun- ima in May±June and in September±October, with a

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC DECEMBER 2003 THATTAI ET AL. 991

itation data clearly indicate that most of the iGoH and the western Caribbean exhibit a bimodal precipitation pattern, for example, Corozal, Belize City, Central Farms, Guatemala City, and Kingston (Fig. 4a).

b. Estimated discharge and sediment load The major rivers Moho, SarstuÂn, Dulce, Motagua, and Ulua, in addition to many smaller streams, empty into the iGoH. The calculated average discharges for the regional rivers are presented in Table 3. The cumulative area of the drainage basins bordering the iGoH measures 42 408 km2 and yields a mean discharge of 1232 m3 sϪ1. The largest drainage basin is that of Ulua, which measures 16 880 km2 with a discharge of 370 m3 sϪ1. Even though SarstuÂn, Dulce and Moho have signi®- cantly smaller drainage basins compared to Motagua and Ulua, they have proportionally higher discharges because of greater rainfall. The isohyets of precipitation FIG. 4. (a) Comparison of mean monthly precipitation (mm) for (Fig. 2) over the watersheds indicate that the drainage Corozal (plus sign, 30 yr), Belize City (star, 51 yr), Central Farms basins of the larger rivers experience signi®cantly less (diamond, 32 yr), Punta Gorda (square, 39 yr), all in Belize; La Ceiba, Honduras (circle, 36 yr), Kingston, Jamaica (triangle, 48 yr), and precipitation. The watershed of Polochic-Dulce mea- Guatemala City, Guatemala (inverted triangle, 35 yr). (b) Comparison sures 5832 km2, exhibits 3150 mm annual rainfall, has of mean monthly temperatures (ЊC) for Belize City (star, 56 yr), a calculated average runoff of 313 m3 sϪ1, and transports Central Farms (diamond, 11 yr), Punta Gorda (square, 6 yr), La Ceiba, fertilizers, pollutants, and nutrients from the Verapaz (circle, 6 yr), Kingston, (triangle, 48 yr), and Guatemala City, (in- verted triangle, 35 yr). districts of Guatemala into the Caribbean Sea. The pro- portionally higher discharge rates for SarstuÂn, Dulce, and Moho are re¯ected in the high runoff ratios for these pronounced minimum in July±August. Tropical storms watersheds (Table 3). cause most of the rainfall in October and November. Wide ¯ood plains and meanders characterize the low- This bimodal distribution of precipitation, referred to as er ¯ood plain of the Motagua River (YaÂnÄez-Arancibia the midsummer drought (MaganÄa et al. 1999), is prev- et al. 1999). The northeast±southwest-trending Cayman alent in Mexico and along the Paci®c coast of Central Trench strikes through the Motagua basin as a depres- America. An absence of bimodal precipitation distri- sion. The Ulua is the largest river in Honduras, and its bution in Punta Gorda and La Ceiba is a result of the drainage area is characterized by a series of vegas, or local orography and is not representative of the entire discontinuous terraces, 10±15 m above sea level western Caribbean. Based on our analysis, the precip- (Schortman and Urban 1995). The watersheds of both

TABLE 3. Characteristics of the inner Gulf of Honduras watersheds as calculated from (i) the runoff-ratio-based discharge equation, (ii) the RUSLE-based sediment load equation (SLÐR), and (iii) the sediment load equation by Morehead et al. (SLÐM). Q SLÐR SLÐM A E, mean (max) R (mm (m3 (ϫ103 t (ϫ103 t Watershed (km2) (m) T(ЊC) yrϪ1) ⌬ f/r sϪ1) Soil type Land use yrϪ1) yrϪ1) Indian Hill 40 5 (5) 28.0 3000 0.42 2 Fe A 0.3 0.4 Monkey 1292 251 (816) 26.6 2800 0.42 49 L/Fe B, C 492 617 Ycacos 158 5 (5) 28.0 3400 0.47 8 Fe A 1 0.7 Deep 357 164 (421) 27.0 3100 0.45 16 Lu/Fe F/S, P 152 182 Golden Stream 207 50 (100) 28.0 2900 0.41 8 Lu/Fe F, A 39 27 Middle 52 27 (100) 28.0 3200 0.45 2 Lu/Fe A 3 5 Grande 722 80 (292) 27.6 3300 0.46 35 L/Fe/Lu M, C 59 95 Moho 1583 150 (350) 26.3 3700 0.53 98 Lu/Fe F, M 294 307 SarstuÂn 2117 218 (779) 26.7 4000 0.55 146 Lu/Fe R, M, F 773 653 Polochic-Dulce 5832 880 (2855) 22.6 3150 0.54 313 Lu/Fe/C CF 7140 5070 Motagua-San Francisco 13 168 1100 (3223) 21.3 1500 0.30 186 Fe/C/L/A B, A 5440 8930 Ulua-CameÂlocon 16 880 939 (2670) 22.3 1900 0.36 370 Fe/C/L/Fl B, Af, A 7110 9140

A: area. E: elevation. T: temperature. R: rainfall rate. ⌬ f/r: runoff ratio. Q: discharge. SLÐR: sediment load (RUSLE model). SLÐM: sediment load (Morehead model). Soil type: LÐleptosol, FeÐferralsol, CÐcambisol, LuÐluvisol, FlЯuvisol, AÐandisol. Land use: BÐ banana, CÐcitrus, FÐforest, SÐswamp, PÐlogging for pine, MÐmilpa (slash and burn), RÐcattle ranching, CFÐcommercial forestry, AÐagriculture, AfÐAfrican palm.

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC 992 JOURNAL OF HYDROMETEOROLOGY VOLUME 4

FIG. 5. Land cover map of the inner Gulf of Honduras. The land cover types shown include crop land (dot), forest (open diamond), irrigated land (wavy line), nonirrigated land (back-slash), grazing land (inverted v), sand (white space), wetland (circles and dots), and forested wetland (gray shade). the Motagua and the Ulua contain extensive agriculture equation. The Morehead et al. model does not explicitly ®elds and banana plantations (Fig. 5). include land use, which is represented by the C factor in A sensitivity analysis was carried out to assess the the RUSLE model. The sediment loads from the two dependence of discharge on temperature and rainfall. models were compared using a paired t test, and there An increase or decrease in the temperature [Eq. (2)] by was no signi®cant difference between the means (t ϭ 1ЊC results in a decrease or increase, respectively, in 1.46) with 9 degrees of freedom at the 95% level of the discharge by 5%. A change in the annual rainfall signi®cance. The results from the two models agreed to rate by ϩ100 and Ϫ100 mm results in a change of within Ϯ2.3%. The availability of more comprehensive ϩ8.6% and Ϫ8.4% in discharge, respectively. The use data on precipitation intensity and land use for the iGoH of the discharge model seems reasonable for the smaller watershed would presumably improve the estimation us- watersheds; whereas the larger watersheds would bene®t ing the RUSLE model. from being subdivided into additional polygons based The sediment yields for the 12 watersheds range from on elevation (Kjerfve et al. 1997). The need for river 8to1224tkmϪ2 yrϪ1 according to the RUSLE es- discharge simulations is necessitated by the lack of local timates, and from 5 to 869 t kmϪ2 yrϪ1 according to river gauging. the Morehead et al. equation. The Polochic-Dulce wa- Sediment load for the regional drainage basins was tershed has the highest calculated sediment yield (sed- calculated based on the two empirical models (Table 3). iment load normalized by basin area) among all the The total sediment load calculated from the RUSLE mod- basins, even though its sediment load is not the highest. el is 21 ϫ 106 tyrϪ1, and from the Morehead et al. model The high value results from higher elevation and pre- is 25 ϫ 106 tyrϪ1. The Morehead et al. equation results cipitation compared to the other larger watersheds. The in slightly greater sediment loads on average as compared CameÂlocon-Ulua and Motagua-San Francisco water- to the RUSLE model, but the difference could be reduced sheds have similar sediment loads (8.9 ϫ 106 and 9.1 by adjusting the values of the parameters in the RUSLE ϫ 106 tyrϪ1) and large drainage areas, and the rivers

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC DECEMBER 2003 THATTAI ET AL. 993

smaller for Deep River, and on average smaller by a factor of 41. Hence, implementation of proper land use and conservation measures in the iGoH watersheds as- sumes urgent importance because a large area of the region is now subject to both commercial and slash- and-burn (milpa) agriculture, which accelerate sediment loss. The effects of varying the C factor on the sediment load and yield for the rivers in the iGoH are shown in Figs. 6a,b. The average sediment yield for the iGoH rivers is 593 tkmϪ2 yrϪ1, with the total sediment load being 25 ϫ 106 tyrϪ1. The sediment yield is very similar to a set of rivers in Colombia, located in the Caribbean basin (541 t kmϪ2 yrϪ1; Restrepo and Kjerfve 2000). For large rivers, a considerable portion of the sediment load may be trapped in the delta. This is not applicable to the iGoH rivers, and the entire sediment load will mostly be discharged into the lagoon or the Gulf of Honduras. Increased sediment loads could potentially affect the growth, or worse, the very existence, of the reefs (Wool- fe and Larcombe 1999). It is a chronic threat that delays reef recovery. Edinger et al. (1998) found that stresses from land-based sources of pollution, including sedi- ments, resulted in 40%±70% reduction in coral species diversity in Indonesia. The destructive effects of sedi- mentation on the reefs of the Bay Islands in Honduras have been documented (Harborne et al. 2001). The sed- FIG. 6. (top) Sediment load (t yrϪ1) as a function of discharge (m3 iment load carried by the Monkey, SarstuÂn, Dulce, Mo- sϪ1) and (bottom) sediment yield (t kmϪ2 yrϪ1) as a function of basin area (km2) from applying the RUSLE model to each of the 12 wa- tagua, and Ulua Rivers is a continuing threat to the reefs tersheds in Table 3 for three cases of model parameter C. Open in the southern region of the MBRS, including the Sap- diamonds denote results from C values varying by land use, as given odillas. This situation can be improved only with better in section 3c, while solid triangles and solid squares, respectively, land management practices in the watersheds. Certainly, represent universal use of C ϭ 0.3 (high erosion potential) and C ϭ 0.003 (low erosion potential). The linear best-®t line is shown for nutrients, pesticides, herbicides, and chemicals often ad- each C-value case, associated with a correlation of (top) 0.9 and here to sediment particles and constitute an additional (bottom) 0.7. threat (Stednick et al. 1995), although not treated here.

c. El NinÄo±La NinÄa variability ¯ow across ¯at lands. Sediment load has a high cor- relation with discharge (R 2 ϭ 0.9) and sediment yield The El NinÄo±La NinÄa cycle has been correlated with also correlates well with basin area (R 2 ϭ 0.7) (Figs. weather patterns worldwide (Kousky et al. 1984; Ro- 6a,b). pelewski and Halpert 1987; Rogers 1988). Ropelewski Changes in land practices, including clearing of for- and Halpert (1987) found a weak relationship between ests, agriculture, and implementation of conservation precipitation and the El NinÄo±Southern Oscillation practices have been included in the simulations of sed- (ENSO), or El NinÄo±La NinÄa, in southern Mexico, Gua- iment load by changing the C value in RUSLE. This temala, southward into Panama, and eastward into the allows the assessment of the importance of land cover Caribbean. Rogers (1988) examined 300 stations to sug- on sediment loss. Two additional simulations were run gest that signi®cant large-scale precipitation variability with C values of 0.3 (less land cover, representing high occurs over the Caribbean and tropical Americas during erosion) and 0.003 (high land cover, representing very the extremes of the El NinÄo±La NinÄa cycle. Based on little erosion) for the entire watershed, which were the a canonical correlation analysis between the Caribbean extreme limits used in the model. The additional sim- rainfall and sea level pressures (SLP) and SST over the ulations indicate that increased cover plays a great role eastern Paci®c and Atlantic Oceans (1951±80), 30% of in conserving sediment and lowering sediment load in the Caribbean rainfall is explained by the interannual the rivers (Figs. 6a,b). The increase in sediment load climatic variability of the eastern Paci®c or Atlantic when C ϭ 0.3 is a maximum of 5 times greater for Oceans and their effects on the tropical Atlantic SST Motagua, and averages 3 times the present best estimate. (Giannini et al. 2000). However, the decrease in sediment load when C ϭ 0.003 The occurrence of El NinÄo and La NinÄa is one of the is much greater; by as much as a factor of 100 times most important modulators of the Atlantic tropical storm

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC 994 JOURNAL OF HYDROMETEOROLOGY VOLUME 4 activity (Fitzpatrick 1999). An average of ®ve hurri- uation. Applying C ϭ 0.3 corresponds to well-devel- canes occurred annually between 1851 and 2000, and oped agriculture (high erosion) with an increase in the there were an additional three tropical storms each year. sediment load by as much as a factor of 5 as compared Examination of the data shows that during El NinÄo to the actual situation. The present best estimates are years, hurricanes occur four or less times, and during thus closer to the high-erosion scenario, stressing the La NinÄa years, they occur six or more times. The re- need for implementation of conservation strategies to duced frequency of tropical storms during El NinÄo years reduce erosion. The high sediment load is a potential is more consistent than the increased frequency during threat to the MesoAmerican Barrier Reef (MBRS) which La NinÄa years. extends through the iGoH. The MBRS is the site of a The regional meteorological stations exhibited higher number of marine conservation areas, and is also one and more variable precipitation between July and De- of the most important tourism destinations in Central cember, whereas the ®rst half of the year is relatively America. dry. The yearly data was split into four sets of three Time series of precipitation for the iGoH exhibits months each: January±February±March (JFM), April± bimodal distribution with maxima in May±June and in May±June (AMJ), July±August±September (JAS), Oc- September±October, similar to the distributions in the tober±November±December (OND). There is no spe- rest of the Caribbean and eastern Paci®c regions. Local ci®c pattern seen in the correlations, but temperature orography, however, destroys the bimodal pattern in the shows higher correlation with SOI than precipitation. southeastern coast of Belize and the northern coasts of Guatemala City has 5%±15% of its precipitation cor- Guatemala and Honduras. The SOI explains only 7%± related with SOI throughout the year, whereas 8%±22% 15% of the precipitation and temperature variability for of Kingston's rainfall during nine months is from the the iGoH. Less than 5% of all tropical storms that pass in¯uence of SOI. A majority of the temperature±SOI through the western Caribbean impact the iGoH. correlations lie between 4% and 23%; 22%±32% of Kingston's temperature and 5%±27% of Guatemala Acknowledgments. This work was supported by a City's temperature are in¯uenced by SOI, whereas only grant from the Nature Conservancy's Ecosystem Grant/ the JFM temperature of Punta Gorda is strongly in¯u- Mellon Foundation (``Physical±biological coupling in a enced by SOI (41%). La Ceiba's temperature correlates tropical bay with implications for conservation and en- well over six months (January±June, 29%±86%). The vironmental management: Inner Gulf of Honduras'' to iGoH lies between the Gulf of Mexico, which shows Kjerfve and Heyman). Thanks are due to Dr. Venkat positive rainfall correlations with the El NinÄo phase, Lakshmi for comments on the manuscript, Mr. Francisco and northern South America, which shows negative AgrenÄal, UTOH/SERNA, for the isotherms of Hondu- rainfall correlations with El NinÄo (Ropelewski and Hal- ras, Mr. Justin Hulse, National Meteorological Service, pert 1987). It thus exhibits the characteristics of both Belize, for the isohyets and isotherms of Belize, E. Ma- the regions and is also affected by the climatic vari- jzlik, K. Sattison, and B. Glett for help with map graph- abilities of the Atlantic and eastern Paci®c (Giannini et ics, and to Sahadeb De for help with the L factor cal- al. 2000; En®eld and Alvaro 1999). Our results con®rm culation in the RUSLE model. The base maps for the that the variability in the western Caribbean does not ®gures have been adapted from ESRI online data strongly depend on ENSO events. Local or regional (http://www.esri.com/data/online/index.html). orographic and climatic conditions in¯uence the tem- perature and precipitation patterns more than global weather cycles. REFERENCES

5. Conclusions Boggs, G., C. Devonport, K. Evans, and P. Puig, 2001: GIS based rapid assessment of erosion risk in a small catchment in the wet/ The inner Gulf of Honduras (iGoH) receives on av- dry tropics of Australia. Land Degrad. Dev., 12, 417±434. erage 1232 m3 sϪ1 of freshwater discharge from 42 408 Bood, N. N., 2001: Ecological status of Belize's southern reef sys- temsÐImpacts of hurricane Iris. Tech. Rep., Belize Coastal Zone 2 km of drainage area, including from seven rivers with Management Authority and Institute, Belize City, Belize. a mean discharge in excess of 100 m3 sϪ1. The drainage CINDI, 1999: Digital Atlas of Central America: Prepared in response basins are characterized by calculated runoff ratios of to Hurricane Mitch, version 2.0. Center for Integration of Natural 0.30±0.55. The composite sediment load from the wa- Disaster Information, CD-ROM disks 1 and 2. 6 Ϫ1 Edinger, E. N., J. Jompa, G. V. Limmon, W. Widjatmoko, and M. J. tersheds is 25 ϫ 10 tyr , and the sediment yield Risk, 1998: Reef degradation and coral biodiversity in Indonesia: for the 12 main watersheds ranges from 8 to 1224 t Effects of land-based pollution, destructive ®shing practices and kmϪ2 yrϪ1. The sediment load and yield are sensitive changes over time. Mar. Pollut. Bull., 36, 617±630. to the degree of land cover as estimated through the C El-Swaify, S. A., and E. W. Dangler, 1982: Rainfall erosion in the factor in the Revised Universal Soil Loss Equation (RU- tropics: A state-of-the-art. Soil Erosion and Conservation in the Tropics, W. Kussow et al., Eds., ASA Special Publ. 43, 1±25. SLE). Applying C ϭ 0.003 corresponds to natural land En®eld, D. B., and E. J. Alvaro, 1999: The dependence of Caribbean cover (less erosion). It decreases sediment load by as rainfall on the interaction of the tropical Atlantic and Paci®c much as a factor of 100 as compared to the actual sit- Oceans. J. Climate, 12, 2093±2103.

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC DECEMBER 2003 THATTAI ET AL. 995

Fitzpatrick, P. J., 1999: Natural DisastersÐHurricanesÐA Reference Middleton, G. V., 2000: Data Analysis in the Earth Sciences Using Book. ABC-Clio, Inc., 286 pp. Matlab. Prentice Hall, 260 pp. Giannini, A., Y. Kushnir, and M. A. Cane, 2000: Interannual vari- Mikhailova, E. A., R. B. Bryant, S. J. Schwager, and S. D. Smith, ability of Caribbean rainfall, ENSO, and the Atlantic Ocean. J. 1997: Predicting rainfall erosivity in Honduras. Soil. Sci. Soc. Climate, 13, 297±311. Amer. J., 61, 273±279. Gibson, J., M. McField, and S. Wells, 1998: Coral reef management Millward, A. A., and J. E. Mersey, 1999: Adapting the RUSLE to in Belize: An approach through Integrated Coastal Zone Man- model soil erosion potential in a mountainous tropical watershed. agement. Ocean Coastal Manage., 39, 229±244. Catena, 38, 109±129. Glantz, M. H., 1997: Currents of Change: El NinÄo's Impact on Cli- Morehead, M. D., J. P. M. Syvitski, E. W. H. Hutton, and S. D. mate and Society. Cambridge University Press, 194 pp. Peckham, 2003: Modeling the inter-annual variability in the ¯ux GLOBE Task Team and Coeditors, cited 1999: The Global Land One- of sediment in ungauged river basins. Global Planet. Change, kilometer Base Elevation (GLOBE) Digital Elevation Model, in press. Version 1.0. Digital database, NOAA. [Available online at Neumann, C. J., B. R. Jarvinen, C. J. McAdie, and J. D. Elms, 1993: http://www.ngdc.noaa.gov/seg/topo/globe.shtml.] Tropical Cyclones of the Atlantic Ocean, 1871±1992. Historical Harborne, A. R., D. C. Afzal, and M. J. Andrews, 2001: Honduras: Climatology Series, Vol. 6-2, National Climatic Data Center, and Caribbean Coast. Mar. Pollut. Bull., 42, 1221±1235. National Hurricane Center, 193 pp. Heyman, W. D., 1996: Integrated coastal zone management and sus- Nieuwolt, S., 1977: Tropical Climatology: An Introduction to the tainable development for tropical estuarine ecosystems: A case Climates of the Low Latitudes. John Wiley and Sons, 207 pp. study of Port Honduras, Belize. Ph.D. thesis, University of South Perkins, J. S., and A. Carr III, 1985: The : Status Carolina, 272 pp. and prospects for conservation management. Biol. Conserv., 31, 291±301. ÐÐ, and B. Kjerfve, 1999: Hydrological and oceanographic con- Portig, W. H., 1976: The climate of Central America. Climates of siderations for Integrated Coastal Zone Management in southern Central and South America, W. Schwerdtfeger, Ed., World Sur- Belize. Environ. Manage., 24, 229±245. vey of Climatology, Vol. 12, Elsevier, 405±478. ÐÐ, and ÐÐ, 2001: Gulf of Honduras. Coastal Marine Ecosystems Rappaport, E. N., and J. Fernandez-Partagas, cited 1997: The dead- of Latin America, U. Seeliger and B. Kjerfve, Eds., Springer- liest Atlantic tropical cyclones, 1492±present. NOAA/NCEP. Verlag, 17±32. [Available online at http://www.nhc.noaa.gov/pastdeadlytx1.html.] Holland, H. D., 1978: The Chemistry of the Atmosphere and Oceans. Rasmusson, E. M., R. E. Dickinson, J. E. Kutzbach, and M. K. Cleav- John Wiley and Sons, 351 pp. eland, 1992: Climatology. Handbook of Hydrology, D. R. Maid- Kaczmarek, Z., N. W. Arnell, and L. Starkel, 1996: Climate, hy- ment, Ed., McGraw-Hill, 2.1±2.44. drology and water resources. Water Resources Management in Renard, K. G., L. D. Meyer, and G. R. Foster, 1997: Predicting soil the Face of Climatic/Hydrologic Uncertainties, Z. Kaczmarek erosion by water: A guide to conservation planning with the et al., Eds., Kluwer Academic Publishers, 3±29. Revised Universal Soil Loss Equation (RUSLE). Agriculture Katz, A., 1989: Coastal resource management in Belize: Potentials Handbook 703, U.S. Department of Agriculture, 404 pp. and problems. Ambio, 18, 139±141. Restrepo, J. D., and B. Kjerfve, 2000: Water discharge and sediment Kinnell, P. I. A., 2000: AGNPS-UM: Applying the USLE-M within load from the western slopes of the Colombian Andes with focus the agricultural non point source pollution model. Environ. Mod- on Rio San Juan. J. Geol., 108, 17±33. el. Software, 15, 331±341. Rogers, J. C., 1988: Precipitation variability over the Caribbean and Kjerfve, B., 1990: Manual for investigation of hydrological processes tropical Americas associated with the Southern Oscillation. J. in mangrove ecosystems. UNESCO Rep., RAS/79/002 and RAS/ Climate, 1, 172±182. 86/120, Mangrove Ecosystems in Asia and the Paci®c, Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional scale UNESCO/UNDP, New Delhi, India, 79 pp. precipitation patterns associated with the El NinÄo/Southern Os- ÐÐ, C. H. A. Ribeiro, G. T. M. Dias, A. M. Filippo, and V. S. cillation. Mon. Wea. Rev., 115, 1606±1626. Quaresma, 1997: Oceanic characteristics of an impacted coastal Schortman, E. M., and P. A. Urban, 1995: Late classic society in the bay: BaõÂa de Guanabara, Rio de Janeiro, Brazil. Cont. Shelf Res., Rio Ulua drainage, Honduras. J. Field Archaeol., 22, 439±457. 17, 1609±1643. Schreiber, P., 1904: UÈ ber die Beziehungen zwischen dem Nieder- Kousky, V. E., M. T. Kagano, and I. F. A. Cavalcanti, 1984: A review schlag und der Wasserfhrung der FluÈsse in Mittleeuropa (About of the Southern Oscillation: Oceanic-atmosphere circulation the relations between the precipitation and the channel ¯ows of changes and related rainfall anomalies. Tellus, 36A, 490±504. the rivers in Mittleeuropa). Meteor. Z., 21, 441±452. Kundzewicz, Z. W., D. Rosbjerg, S. P. Simonovic, and K. Takeuchi, Shen, H. W., and P. Y. Julien, 1992: Erosion and sediment transport. 1993: Extreme hydrological events in perspective. Extreme Hy- Handbook of Hydrology, D. R. Maidment, Ed., McGraw-Hill, drological Events: Precipitation, Floods and Droughts, Z. W. 12.1±12.61. Spangler, W. M. L., and R. L. Jenne, 1990: World monthly surface Kundzewicz et al., Eds., IAHS Publ. 213, 1±7. station climatology (TD-9645). World Weather Disc User Man- LoÂpez, R., and C. Scoseria, 1996: Environmental sustainability and ual, Version 2.0, WeatherDisc Associates, R1±R4. poverty in Belize: A policy paper. Environ. Dev. Econ., 1, 289± Stednick, J. D., D. M. Gilbert, and M. D. Lee, 1995: Development 307. of a national water-quality monitoring program for Belize. Water MaganÄa, V., J. A. Amador, and S. Medina, 1999: The midsummer Resources at Risk, D. R. Hotchkiss et al., Eds., American In- drought over Mexico and Central America. J. Climate, 12, 1577± stitute of Hydrology, RA16±22. 1588. Woolfe, K. J., and P. Larcombe, 1999: Terrigenous sedimentation and Martyn, D., 1992: Climates of the World, Vol. 18, Developments in coral reef growth: A conceptual framework. Mar. Geol., 155, Atmospheric Science, Elsevier, 435 pp. (Translated from Polish 331±345. by P. Senn.) YaÂnÄez-Arancibia, A., L. D. ZaÂrate, C. M. GoÂmez, O. R. GodõÂnez, McField, M., S. Wells, and J. Gibson, Eds., 1996: State of the coastal and F. V. Santiago, 1999: The ecosystem framework for planning zone report, Belize. Coastal Zone Management Project Bze/92/ and management of the Atlantic coast of Guatemala. Ocean G31, Belize City, Belize, 255 pp. Coastal Manage., 42, 283±317.

Unauthenticated | Downloaded 10/03/21 12:02 AM UTC