Tropical Rainfall Measuring Mission Precipitation Radar precipitation features are analyzed to understand the role of storm characteristics on the seasonal and diurnal cycles of precipitation in four distinct regions in Costa Rica. The distribution of annual rainfall is highly dependent on the stratiform precipitation, driven largely by seasonal increases in stratiform area. The monthly distribution of stratiform rain is bimodal in most regions, but the timing varies regionally and is related to several important large-scale features: the Caribbean low-level jet, the ITCZ, and the Chorro del Occidente Colombiano (CHOCO) jet. The relative importance of convective precipitation increases on the Caribbean side during wintertime cold air surges. Except for the coastal Caribbean domain, most regions show a strong diurnal cycle with an afternoon peak in convection followed by an evening increase in stratiform rain. Along the Caribbean coast, the diurnal cycle is weaker, with evidence of convection associated with the sea breeze, as well as a nocturnal increase in storms. The behavior of extreme precipitation features with rain volume in the 99th percentile is also analyzed. They are most frequent from May to November, with notable differences between features at the beginning/end of the rainy season versus those in the middle, as well as between wet and dry seasons. Convective rain exceeds stratiform in winter and midsummer extreme features, while stratiform rain is larger at the beginning and end of the wet season. Given projected changes in precipitation and extreme events in Costa Rica for future climate change scenarios, the results indicate the importance of understanding both changes in total precipitation and in the storm characteristics.
Costa Rica is home to some of the world’s most valuable and diverse cloud and rain forest ecosystems that are intricately linked to the region’s abundant clouds and precipitation. These forests are strongly influenced by climatic conditions, and factors like precipitation amount, frequency, and intensity control many aspects of the forest systems, suggesting vulnerability to perturbations and fluctuations in rainfall (e.g., Hager and Dohrenbusch 2011; Jarvis and Mulligan 2011) in a changing climate. The economy also depends heavily on tourism and agriculture, the latter of which is especially sensitive to extreme precipitation events that can devastate crops.
Precipitation variability in Costa Rica is driven by interactions between the local topography and a combination of the seasonal migration of the intertropical convergence zone (ITCZ), land/sea breeze effects, monsoonal circulations, strong easterly trade winds and low-level jets, temporales (periods of extended rainfall), hurricanes and easterly waves, and nortes (cold air masses from midlatitudes) in the winter. The activity of many of the aforementioned phenomena peaks during boreal summer and autumn, bringing frequent and heavy precipitation from May through October (Waylen et al. 1996a). Heavy precipitation events due to direct landfalling tropical cyclones are rare. However, other heavy precipitation events, especially on the Pacific side, may be an indirect effect of the circulation around tropical cyclones in the Caribbean, reversing the typical easterly flow (Waylen and Harrison 2005). As the countertrade flow interacts with the topography, moderate to heavy rainfall events are more likely (Waylen and Harrison 2005) and can occasionally result in large-scale flooding with significant socioeconomic impacts. Aguilar et al. (2005) noted opposite trends in precipitation in Costa Rica over the last several decades, with increases on the northwestern Caribbean side and decreases on the Pacific sides. However, they also show that the precipitation intensity has increased, and there are fewer heavy precipitation events along the Pacific side and more heavy precipitation events on the Caribbean side.
Though there is considerable uncertainty in regional climate projections, a report on the Assessments of Impacts and Adaptations to Climate Change and Extreme Events in Central America (AIACC-LA06) shows future climate scenarios that indicate enhanced precipitation in the southern part of the country and a decrease in precipitation in the northern part, with magnitudes that are projected to continue to grow through 2100 (Alvarado et al. 2005). Model results also indicate a possible seasonal shift in precipitation, with increases from October to April resulting in a shorter dry season and less seasonality. The report also hypothesizes that an increase in extreme precipitation events, especially those due to heavy rains associated with cold surges (Schultz et al. 1998), may be part of the reason for the increase in winter precipitation in the southern part of the country. The observed increase in the frequency of extreme precipitation events and the projected changes in regional precipitation and extreme precipitation frequency suggest that a better understanding of how the seasonal variability in storm characteristics contributes to the seasonal precipitation variability is necessary.
Typical precipitating systems during the wet season may come as short-lived, intense, convective events associated with daytime heating or from large stratiform regions associated with more organized mesoscale convective systems (MCS; Houze 2004). There have been a number of studies focused on precipitation variability in Costa Rica using in situ gauge data (e.g., Hastenrath 1967; Waylen et al. 1996a); however, there is currently no ground-based radar network with publicly available data to explore the characteristics of convection underlying the precipitation variability in this region. Many tropicswide (e.g., Nesbitt et al. 2000, 2006; Schumacher and Houze 2003; Liu 2011) and regional (e.g., Xu et al. 2009; Romatschke and Houze 2011; Guy and Rutledge 2012) studies have used satellite data from the Tropical Rainfall Measuring Mission (TRMM) to explore the characteristics of convection, especially in tropical regions where ground-based radar measurements are sparse. Zuluaga and Poveda (2004) used the University of Utah TRMM precipitation feature database (Nesbitt et al. 2000; Nesbitt and Zipser 2003; Liu et al. 2008) in a study analyzing the behavior of MCS in South and Central America. Their analysis focused on the frequency, size, and contribution to total precipitation, but not on the characteristics of the MCS or on how the different convective and stratiform precipitation components contribute to the total precipitation. A number of studies using TRMM data and the TRMM precipitation feature database to study the contributions of convective and stratiform rainfall in other regions like South America, Asia, and Africa (e.g., Kodama et al. 2005; Romatschke and Houze 2010; Romatschke et al. 2010) motivate this study of precipitating systems in Costa Rica, where a unique set of local, regional, and large-scale forcings drive precipitation.
Given the regionally varying, observed increase in precipitation intensity and the changes in heavy precipitation event frequency (Aguilar et al. 2005), it is important to determine how convective and stratiform regions contribute to precipitation variability and storm characteristics of both typical and extreme precipitation events. This study uses the University of Utah precipitation feature dataset (Nesbitt et al. 2000; Nesbitt and Zipser 2003; Liu et al. 2008) based on the TRMM Precipitation Radar (PR) to understand the role of precipitating storm characteristics on the seasonal and diurnal cycles of precipitation over land in Costa Rica. The main focus is on characterizing the frequency, area, and intensity of precipitation, as well as on the contribution of convective and stratiform rain to both regional and individual storm totals. The frequency and behavior of extreme precipitation features are also analyzed to determine how differences in the storm characteristics manifest in observed precipitation extremes.
2. Data and methods
The University of Utah TRMM precipitation feature database (Nesbitt et al. 2000; Nesbitt and Zipser 2003; Liu et al. 2008) is used here to identify all individual precipitation features for 1998–2012 over land areas in Costa Rica. The precipitation features in the database were identified using contiguous areas of one or more pixels with a TRMM PR 2A25 (Iguchi et al. 2000) rain rate that is greater than zero. The database contains a plethora of variables associated with each precipitation feature from all of the instruments onboard the TRMM satellite. However, this study uses only the PR 2A25 precipitation features because of the considerable uncertainty in passive microwave retrievals of rainfall over land. These are used to define the rain area and size of the feature, as well as the total rain volume. The precipitation feature database also contains the output of the PR 2A25 convective/stratiform classification scheme (Awaka et al. 1997) that is based on direct measurements of brightband detection and horizontal and vertical reflectivity gradients after Steiner et al. (1995). The convective and stratiform rain contributions will be used here for storm characterization. The rain area and rain volumes are then used to calculate the total, convective, and stratiform conditional rain rates (mean rain rate calculated for raining pixels only) for all features. Monthly-mean characteristics of the individual precipitation features are compared to the domain totals to determine how the individual storm characteristics contribute to the monthly total. The precipitation feature dataset is also used to identify and analyze extreme precipitation events, defined here as individual features with a total rain volume in the 99th percentile of all features.
For this analysis, Costa Rica is divided into four separate regions (Fig. 1). The regions roughly correspond to areas with distinct rainfall patterns identified by Waylen et al. (1996a) using in situ monthly-mean rainfall. While their study identified five different areas, because of sampling limitations and the small geographic size of some of these regions, some of the regions are combined here. The eastern and western domains are determined based on the Cordillera de Talamanca in the southern half of the country and along Cordillera de Guanacaste in the northern part. The eastern side of the country is further divided into a central domain (region B in Fig. 1) and a coastal Caribbean domain (region A in Fig. 1) similar to Waylen et al. (1996a). Past studies showed considerable differences in precipitation on the western side of the country, with the northern Pacific region exhibiting more distinct wet and dry seasons than the southern Pacific region (Waylen et al. 1996a). The western side of the country is divided along the rain-rate gradient into a northwestern (region C in Fig. 1) and southwestern region (region D in Fig. 1), where the Central Valley region is grouped with the northwestern domain.
For each of the defined areas, monthly-mean total raining pixels, total rain volume, and the percentage of the annual total rain volume is calculated from all of the individual precipitation features within the domain. To better characterize the storms that compose the monthly totals, mean characteristics of the individual precipitation features are also calculated for both the diurnal and seasonal cycles. Precipitation feature characteristics include storm frequency, storm area, and storm volume. Mean conditional rain rates for each feature are calculated as the sum of the rain volume divided by the total number of pixels in the precipitation feature. Finally, the relative contribution of convective and stratiform rain to the feature size, volume, and rain rate are also calculated.
a. Regional precipitation distribution
Figure 1 shows the climatological distribution of precipitation at 0.25° × 0.25° resolution from the TRMM PR 3B43 dataset (Huffman et al. 2007) for Costa Rica for 1998–2012. There are relative rainfall maxima along the northeastern Caribbean coast and the southwestern Pacific coast. The area of heavier precipitation along the Caribbean coast is persistent year-round, though it peaks in boreal summer and autumn when the northeast trade winds are strongest (Grandoso et al. 1982) and easterly wave and tropical cyclone activity is most frequent (e.g., Waylen and Harrison 2005; McAdie et al. 2009). The precipitation enhancement in Fig. 1 along the Pacific coast also peaks during June–November as sea surface temperature (SST) peaks and the ITCZ progresses through the region. Heavy precipitation events also occur, especially during late summer and early autumn when large-scale flow due to Caribbean tropical cyclones drives onshore winds on the west side of the cordillera (Vargas and Trejos 1994; Fernandez and Barrantes 1996). The precipitation peak on the Pacific side nearly completely disappears during boreal winter when a local SST minimum occurs in this region (Kessler 2002; Xie et al. 2005).
Consistent with previous in situ studies (e.g., Hastenrath 1967; Waylen et al. 1996a; Magaña et al. 1999), monthly-mean precipitation exhibits a strong seasonal cycle and considerable regional variability. Figure 2 depicts the mean monthly percentage of the annual total rain volume for each region derived from the TRMM 2A25 precipitation feature database. There is generally much less precipitation in the boreal winter when surrounding SSTs are lowest and trade winds are weaker, with the notable exception of the Caribbean coastal domain. Winter precipitation events are associated with cold air outbreaks or frontal systems that pick up moisture over the Caribbean and advect precipitation to northern Costa Rica and the Caribbean coasts (Schultz et al. 1998).
Most regions show a peak in annual rain volume during boreal summer and autumn, with a bimodal distribution that corresponds to the interactions between the eastern Pacific ITCZ, seasonal SST patterns, radiation, and easterly winds (Magaña et al. 1999). However, the timing of the maxima varies from region to region, likely due to interactions with other features such as the Caribbean low-level jet (Amador 2008) and the Chorro del Occidente Colombiano (CHOCO) jet (Poveda and Mesa 2000). There is also evidence in Fig. 2 of the midsummer drought (e.g., Magaña et al. 1999; Waylen and Quesada 2002; Small et al. 2007). While not a true drought, the midsummer drought mostly impacts the domains on the Pacific side and manifests as a relative minimum within the overall May–November precipitation increase (Magaña et al. 1999). As strong northeasterly trades during July and August flow onshore perpendicular to the central Costa Rican cordillera, precipitation increases in the Caribbean and central domains, with a reduction in the Pacific domains (Fig. 2).
The peak in precipitation in Pacific domains around September/October also corresponds to the most active months for Atlantic tropical cyclones (Waylen and Harrison 2005). While these cyclones very rarely make landfall in Costa Rica, the associated reversal of the large-scale flow across the Central American isthmus results in greatly enhanced precipitation on the Pacific side (Vargas and Trejos 1994; Waylen and Harrison 2005).
b. Annual storm characteristics
Figure 2 shows that there is a large amplitude of the annual cycle in precipitation. The different mechanisms responsible for the seasonal precipitation variability suggest considerable variability in the characteristics of precipitating systems across the different regions in Costa Rica. The following sections explore how the mean individual precipitation feature characteristics contribute to the annual precipitation variability for the four domains shown in Fig. 1.
1) Caribbean coastal domain
The Caribbean coastal domain of Costa Rica lies along the eastern edge of the country where adjacent SSTs in the Caribbean are above 28°C during the summer and autumn and then reach a minimum during February (Wang and Enfield 2001). Figure 3 shows the annual behavior of the precipitation features in this region. During late winter/early spring, monthly total rain volume is about half of the early summer maximum. Precipitation features are just as frequent during the winter (Fig. 3a), but the individual precipitation feature size and rain volume are much smaller. As indicated in Fig. 2, there is a bimodal distribution of total precipitation area (Fig. 3b) and rain volume (Fig. 3c) in this region. There is a weak but broad increase in total precipitating area and rain volume in May–August and a large spike in precipitation area and volume in November. Rather than being associated with more frequent storms (Fig. 3a) or higher rain rates (Fig. 3f), the solid black lines in Figs. 3d–e show that the two peaks in precipitation are a result of an increase in the area of the individual precipitating features.
The dashed and dotted lines in Figs. 3b–e indicate the relative contributions of convective and stratiform rainfall to the regional and individual storm totals. Precipitation is largely composed of convective rainfall until May. From May through October, stratiform precipitation becomes much more important and contributes about 40% of both the total (Fig. 3c) and storm (Fig. 3e) rain volume, despite lower stratiform rain rates (Fig. 3f). The increase in the stratiform contribution is the result of storms with larger stratiform areas. While the summer stratiform contribution increases, convective rainfall still contributes a larger fraction of the total and individual storm rain volume. There is some increase in the convective area as the storm size increases, but the convective rain rates also increase during the summer. During the November precipitation maximum, stratiform rain volume reaches nearly 50% of the total and is mainly due to storms with large stratiform areas.
The bimodal structure of the storm characteristics in this region seems to be partially related to the presence of the Caribbean low-level jet (e.g., Amador 1998; Poveda and Mesa 2000; Muñoz et al. 2008), which transports moisture from the Caribbean to the region. This jet exhibits significant seasonal variability, with peak strengths in easterly winds and water vapor flux divergence in the central Caribbean during July and February (Muñoz et al. 2008; Poveda et al. 2014). A number of studies (e.g., Amador et al. 2000; Whyte et al. 2008; Muñoz et al. 2008; Cook and Vizy 2010) have shown an area of enhanced precipitation along the Caribbean coasts of Costa Rica and Nicaragua during June and July. This is associated with the enhanced water vapor transport and moisture flux convergence at the jet exit region, which corresponds with the increase in rain volume here. The enhanced moisture convergence in this region yields conditions more favorable for the development of large MCS indicated by the increasing stratiform area during June and July.
In situ studies (e.g., Hastenrath 1967; Waylen et al. 1996a) along the Caribbean coast typically show a secondary precipitation peak in November–January, with a maximum in December. This winter peak has been associated with a secondary Caribbean low-level jet maximum as well as cold surges from nortes (Schultz et al. 1998). These midlatitude frontal systems can penetrate to Central America, with the air mass moistening as it crosses the Caribbean and enhancing the strength of the trade winds. The associated winter precipitation peak is reflected in the storm frequency and is noticeable in the domain totals; however, the signal is somewhat overwhelmed by the large peak in November. While small precipitation features are more common, examination of the individual precipitation features shows that the November peak is not solely due to a few anomalous events that occur in only a few anomalous years. These large coastal storms in November are a regular occurrence. Because in situ stations do not show such large November contributions to the annual total rainfall, these precipitation features might be “contamination” from oceanic storms. The features within the database are identified by the storm center latitude and longitude, so it could be that the bulk of the heavy convective precipitation is actually occurring over the ocean, with the large stratiform regions (and thus the center of the storm) extending over the land.
On the other hand, it is unlikely that only 1 month of the year would be impacted by this type of oceanic storm contamination. Because the Caribbean low-level jet is near its minimum during October–November (Muñoz et al. 2008), this suggests that late autumn/early winter cold surges may be responsible. The higher temperatures and moisture availability in the Caribbean results in conditions more conducive to the development of MCS with large stratiform regions associated with cold surges during this period. It may also account for why the storm characteristics are very different compared to the colder, late winter surges that produce smaller storm systems and much less stratiform precipitation in February and March. Despite a secondary peak of the Caribbean low-level jet during February, a decrease is observed in both storm and total rain area and volume, partially because the easterly winds are slightly weaker than those in July, but primarily because SST and specific humidity in the Caribbean are much lower (Muñoz et al. 2008), leading to reduced surface moisture fluxes and thus lower moisture availability for the development of MCS.
2) Central domain
In the Waylen et al. (1996b) clustering analysis of in situ data, stations in the central domain were grouped with the Caribbean because they are generally under the influence of Caribbean air masses. However, the central portion of Costa Rica is characterized by highly variable topography, with large elevation gradients along the eastern slope of the cordillera. Because this could potentially influence the characteristics of the convection, this region is considered separately here.
The easterly trades generally flow perpendicular to the mountain range, transporting moisture inland from the Caribbean. Unlike for the Caribbean coastal region, the annual rain volume in Fig. 2 is largely reflected in the variability in the frequency of central domain storms (Fig. 4a). Storms are nearly twice as frequent in May–October than during February–April. The domain totals exhibit a bimodal distribution peaking in August and October/November with storm activity slightly less frequent in June and September. During the wet season, the storm characteristics are less variable than the coastal storms, although there are slight peaks in August and November. When compared with the Caribbean coast domain, storms in the central domain, especially in the late summer, are much smaller and thus have considerably lower storm rain volume despite comparable rain rates.
Convective rain rates are very similar to the rain rates for the coastal storms. They also peak in the same months as the storm size, which is why the convective rain volume per storm stays higher than the stratiform despite the lack of change in the convective area of the storm. There is almost no seasonal variability in the convective area per storm, but Figs. 4b and 4d show a large increase in the stratiform area during early summer that continues to increase during the autumn, indicative of the development of afternoon deep convection and MCS typically found over tropical land areas (e.g., Dai 2001; Nesbitt and Zipser 2003).
The strong easterly trade winds in the months of maximum solar insolation combined with the large elevation gradient due to the Central American cordillera result in frequent storms and large rain volumes throughout the summer and autumn. The large topographic variability in this region may also be reflected in the characteristics of the storms. The smaller feature size could be due to more numerous, smaller-scale, orographic precipitation features induced by local mountain valley circulations. While precipitation in the Caribbean domain is largely dependent on the characteristics of the storms, farther inland in the central domain the variability in storm frequency becomes much more important to the annual distribution of rainfall.
3) Northwestern domain
Shifting the focus to regions on the Pacific side, the northwestern domain lies on the leeward side of the cordillera, but adjacent to the eastern Pacific warm pool. The easterly trades produce heavy precipitation on the windward side and a rain shadow effect leaves this region relatively dry compared with the regions on the Caribbean side. Figure 5 shows that storms on the Pacific coast are less frequent than the Caribbean coast, with few storms and little precipitation during the winter season. Storm frequency increases and is relatively high from May through November. There is a weak bimodal distribution of increasing storm frequency during May/June and October, but domain totals peak in June and September when storm sizes are largest. This increase in storm area and storm volume is primarily due to the fivefold increase in stratiform area. During June, stratiform rain volume actually equals the convective rain volume, despite the increased convective rain rates. During the second rain volume peak in September, convective rain volume again dominates due to slightly larger convective to stratiform area ratios and a secondary peak in convective rain rates. This bimodal distribution is associated with the seasonal movement of the east Pacific ITCZ (Hastenrath 2002) and its interactions with radiative effects of deep convection, as well as seasonal variability in the easterly trade winds (Magaña et al. 1999).
The large peaks in the storm characteristics in June and September are also consistent with past studies that indicate an increase in temporales (e.g., Portig 1965; Hastenrath 1967; Fernandez and Barrantes 1996). Southwesterly flow from the Pacific transports moisture to the region and has been associated with the reversal of trade winds when tropical storms are present in the Caribbean (Vargas and Trejos 1994; Waylen and Laporte 1999; Peña and Douglas 2002). The persistent supply of moisture from the Pacific and orographic lift provided by the Cordillera de Guanacaste and Tilaran in this region can result in rainy conditions that may last for days. These peaks are not as noticeable in storm frequency, but the storm characteristics and domain totals indicate a significant difference in stratiform rain production between months most affected by temporales compared with the rest of the “wet” season.
Another interesting result here is that the reduction in precipitation related to the midsummer drought is more noticeable in the characteristics of the storms than in the storm frequency. The leeside rain shadow effect likely contributes to the slight July decrease in storms; however, storms are nearly as frequent during July and August as in May and June, but their characteristics change considerably during the period when the Caribbean low-level jet is strongest. When the jet is strong, on the Caribbean side the storms are larger with enhanced stratiform areas, as well as increasing convective rain rates. On the leeward side, storms in the northwestern domain are about 30%–40% smaller during July and August compared with storms in June or September, mostly due to the decrease in the amount of stratiform precipitation and in convective rain rates. Strong easterly winds accelerate through gaps in the cordillera along the northern border with Nicaragua and continue offshore in the eastern Pacific (e.g., Clarke 1988; Chelton et al. 2000). This results in enhanced surface upwelling and can lower SSTs by several degrees in the Gulf of Papagayo (Chelton et al. 2000; Xie et al. 2005). Combined with the decrease in SSTs due to the reduction in surface shortwave radiation by ITCZ deep convection in June (Magaña et al. 1999), adjacent SSTs decrease below the threshold for deep convection (Webster 1994). The lower SSTs as well as the strong divergence and subsiding air off the northwestern side of Costa Rica (Amador 2008) result in a more unfavorable environment for the development of MCS and leads to the reduction in stratiform storm area and convective rain rates observed in this region during the midsummer drought.
4) Southwestern domain
The southwestern domain of Costa Rica is bounded on the western side by the warm waters in the eastern Pacific warm pool and by some of the highest mountains in the Cordillera de Talamanca on the eastern side. Unlike the northwestern domain, the southwestern domain experiences westerly winds for part of the year, bringing in moisture off the Pacific (Durán-Quesada et al. 2010). The southwestern region also shows a bimodal distribution in the domain totals in Fig. 6, with peaks in May and October when storms are most frequent as the ITCZ migrates through. There is also a bimodal distribution in the storm characteristics, although the peaks are 1 month later than in storm frequency. From January through October, storm area increases by more than a factor of 4, resulting in a similar increase in storm volume. Combined with the increased frequency of storms, domain total rain volume increases by more than a factor of 5. Conditional rain rates are slightly larger during the late winter and early spring and do not exhibit the bimodal distribution observed in the northwestern domain. Evidence of the precipitation suppression during the midsummer drought is bit more apparent in the storm frequency compared with the northern Pacific domains. However, storms during the midsummer drought also show a large change in the storm characteristics during July and August, with smaller storms and lower rain volume manifesting in reduced domain totals.
The seasonal variability in domain and storm area is largely a function of the massive increase in stratiform area during the early summer and especially autumn. Domain total and storm convective and stratiform rain volume variability is similar, with stratiform volumes equal to convective during the autumn precipitation peak. Both convective and stratiform rain rates show less seasonal variability than the other regions. The large enhancement of precipitation during the summer and autumn is likely related to the transport of moisture to southwestern Costa Rica by the CHOCO jet (Poveda and Mesa 2000; Poveda et al. 2014). As the ITCZ moves farther north, the southwestern region is under the influence of the cross-equatorial westerly winds and the low-level CHOCO jet, which peaks in strength during October and November. Using a Lagrangian trajectory technique to identify sources of moisture for Central America, Durán-Quesada et al. (2010, 2012) show that the Pacific is an important source of moisture in southern Costa Rica during summer and autumn. As the moisture is transported inland, it undergoes forced ascent over the high Cordillera de Talamanca, allowing deep convection to develop. Zuluaga and Poveda (2004) and Poveda et al. (2006) show frequent MCS in the general region during June–October. Although somewhat difficult to discern from their figures, it also appears that they find larger MCS during early summer and autumn, with smaller MCS during the middle of the summer. The results here agree, showing storms reaching very large sizes with well-developed large stratiform regions, indicating the presence of MCS during the early summer as the ITCZ moves through and nearby SSTs reach a maximum and during autumn when the CHOCO jet is strongest.
c. Diurnal storm characteristics
A number of studies have shown distinct differences in the diurnal cycle of precipitation over ocean and land surfaces. Over land, daytime heating fuels an afternoon peak in precipitation, while over ocean there is a smaller diurnal cycle, although precipitation typically peaks at night (Gray and Jacobsen 1977; Dai 2001; Nesbitt and Zipser 2003). Additionally, large diurnal variability is expected in regions like Costa Rica where there is a strong land/sea breeze and topographic features that are generally perpendicular to the flow. Figure 7 shows the mean diurnal cycle in storm frequency, rain rate, storm area, and percentage of convective rain volume.
Throughout the year, precipitation features in nearly all regions are most frequent between 1200 and 1800 LT (Fig. 7a). Maximum storm average rain rates occur in the early hours of the peak, although the highest convective rain rates (gray lines, Fig. 7b) are about 2 h later during peak afternoon heating. The stratiform rain rates are not shown because they vary little over the diurnal cycle. The storm area in Fig. 7c is generally the smallest in the late morning and early afternoon from about 1100 to 1500 LT, as the systems are primarily convective (Fig. 7d) during the development of convection. Just after sunset, storm areas become 3–4 times larger, and the stratiform fraction (Fig. 7d) increases to nearly 50% of the total rain volume. This shift also explains the time offset between storm mean and convective rain rate since a larger fraction of the precipitation comes from lighter stratiform rainfall later in the afternoon. This progression suggests the typical evolution of deep convection from convective toward more stratiform rainfall (Houze 2004). It is interesting that most of the regions show a small peak in storm frequency around 1700 LT. Zuluaga and Poveda (2004) showed MCS peak over land in this region at 1700 LT, which is consistent with the peak in convective rain rates and increasing areas in Fig. 7.
Most of the regions show similar diurnal patterns in storm frequency, although there is a considerably weaker diurnal cycle in the Caribbean coastal region. There is evidence of two peaks in storm frequency and area in the Caribbean domain, with an early afternoon peak and a second peak during the nighttime between 0000 and 0300 LT. The peak in frequency around 1200 LT followed by the peak in storm area a few hours later is likely related to convection developing along the sea breeze front as it moves inland. This peak is consistent with the results of Kikuchi and Wang (2008) who classify land areas of Costa Rica as a “landside coastal” region where precipitation begins along the coast during the late morning, propagates inland, and is reduced in the evening. The latter peak is more difficult to interpret; however, examination of the seasonal distribution shows that this nighttime peak is most frequent during winter from November to March, which is consistent with Fig. 7d showing little stratiform precipitation. This suggests that it could be related to the nighttime peak of the Caribbean low-level jet during the winter (Muñoz et al. 2008).
The Pacific coastal domains also show a small increase in storm area and stratiform precipitation, although they occur later than in the Caribbean domain. One alternative explanation for the nighttime precipitation peak along the coasts is that it is connected to nighttime oceanic precipitation associated with the land breeze that develops along the coast and moves offshore. Kikuchi and Wang (2008) identified seaside coastal regions offshore of Costa Rica with nighttime peaks in precipitation from late evening through the early morning. In their analysis of the diurnal cycle of precipitation using high-resolution TRMM PR data, Biasutti et al. (2012) also show enhanced nighttime precipitation frequency along these coastal regions.
The diurnal cycle in storm characteristics is most pronounced in boreal summer and autumn and is weaker during the late winter and early spring. Figure 8 shows the general overall seasonal pattern in the diurnal cycle for the central domain, which is fairly representative of most of the domains presented here. Daytime heating in the summer and autumn helps set up strong land/sea breeze circulations, which are further enhanced especially in regions with large orographic effects. During these months, rain rates peak in the early afternoon and fall off as storm area increases associated with the development of stratiform rain. During winter there is essentially no diurnal cycle in storm area or in convective/stratiform rain fractions. As suggested in Figs. 3–6, storms are less frequent and small, with primarily convective cells present throughout the diurnal cycle during the winter.
d. Extreme precipitation features
To get a sense of how extreme precipitation features are distributed within the different domains, extreme features are identified here as those with rain volume in the 99th percentile of all features in the country for 1998–2012. The sensitivity of the definition of extreme rainfall events was also tested using rain area and rain rates. The results for separately identifying extremes as a function of rain rate or area are qualitatively similar to those shown here using both criteria; however, there are some differences in magnitude. Rain volume is chosen because it represents the combination of both high rain rates and large rain area, which is more likely to be associated with large-scale flooding events resulting in significant human and socioeconomic impacts. Despite quantitative differences, the overall conclusions are the same regardless of identification criteria. Figure 9 shows the average monthly frequency of extreme precipitation features for each domain for the entire 15-yr dataset. Extreme precipitation features are most common during the wet season with about 85% of all extreme features occurring from May to November. Extreme precipitation features in the northwestern and southwestern domains in September and October coincide with the months of most frequent tropical storm activity in the Atlantic and are consistent with the findings of Waylen and Laporte (1999) for flood events on the Pacific side. An increase in the frequency of extreme events in the Caribbean and central domains occurs multiple times during the year, with summer and autumn maxima.
Figure 10 shows the rain area, rain volume, conditional rain rates, and convective/stratiform volume fraction for the 99th percentile precipitation features. It is immediately obvious that the size of these features is much larger than typical events and that the stratiform area is more than twice as large as the convective area for May–December extreme features (Fig. 10a). The convective area of extreme features shows only a small amount of monthly variability, with only slightly higher precipitation area during winter. There is also a noticeable decrease in the rain area for extreme features during August–October, driven primarily by the decrease in stratiform area. Since most of the extreme storms during these months are located in the central and Pacific domains, this suggests that extremes in these regions are smaller and able to support less robust stratiform areas than extremes in the other domains and at other times of the year. The maxima in the rain area and rain volume occur during June/July and November, when extreme storms are most frequent in the Caribbean and central domains. These extreme events in the regions on the windward side of the cordillera seem to correspond with months of the increasing strength of Caribbean low-level jet and moisture convergence (Alfaro 2002). It is also possible these maxima are due to large MCS that interact with the local topography as they move inland, especially in the southern portion of the Caribbean and central domains. The exceedingly large stratiform areas are consistent with the suggestion by Houze (2012) that upward motion and stratiform regions are enhanced when MCS with existing large stratiform regions interact with higher terrain.
Correspondingly, rain volumes associated with winter extreme events are generally lower than the rain volume during the peak of the wet season (Fig. 10b). Rain volumes are highest during early summer and late autumn with the decrease in rain volume during late summer following the rain area curve, despite the increase in the frequency of extreme events. Unlike the rain area, there is considerable monthly variability in the convective rain volume of extreme features. Convective rain volume is larger than stratiform during the midsummer and winter extremes, while stratiform rain volume makes up over 50% of the total rain volume during the early summer and autumn extreme events. This difference in convective rain volume with a lack of change in rain area points to differences in the intensity of convective rainfall for extreme events.
Figure 10c shows the convective rain intensity for extreme events is 3–4 times larger than the typical convective rainfall for any of the domains in Figs. 3–6. There is a notable difference in seasonal extreme feature convective rain intensity, which nearly doubles from February/March to June/July. Stratiform rain intensity for wet and dry season events is nearly constant, showing that the peaks in rain volume for extreme events are driven by the large stratiform areas and heavy convective rain rates rather than changes in stratiform rain intensity and convective area. This is also consistent with the findings from Houze (2012) for topographically enhanced extreme precipitation events.
There are fewer extreme features in December–April, with the majority occurring in the Caribbean and central domains where cold surges from the midlatitudes are bounded by the cordillera. There are noticeable differences between extreme events in the winter and those that occur during the wet season. Winter season, especially late winter, extreme features are smaller than their summer/autumn counterparts. As in the monthly-mean precipitation features, convective rainfall still dominates winter extreme rain volume (Fig. 10d). Although extreme events that lead to flooding are typically associated with large areas of prolonged stratiform precipitation (e.g., Houze et al. 2011), the winter extreme features identified here still show convective rain volumes that exceed stratiform precipitation. For example, the January 2005 precipitation feature located in the Caribbean domain was associated with one of the largest winter flood events in recent years, displacing thousands of Costa Ricans and destroying many of the country’s banana plantations. Examination of the individual precipitation feature characteristics showed that it had nearly 60% convective rainfall, with above average convective rain rates. However, the precipitation feature associated with the extreme flooding event in the northwestern domains related to Tropical Storm Tomas in November 2010 had a much larger area, with over 80% of the rainfall from the large, well-developed stratiform region.
It is also notable that that convective rain rates for early winter extreme events are similar to summer and autumn rain rates, while late winter extreme events are smaller and have considerably lower convective rain rates. The smaller size and lower rain rates are likely why few large-scale flood events are noted during these months (Waylen and Laporte 1999). Because many of the extreme precipitation features identified during winter are likely associated with cold surges (Schultz et al. 1998), the results again suggest a possible difference in the behavior of convection associated with early relative to late winter cold air outbreaks.
It is unclear from these results whether the late summer decrease in rain area and rain volume is due to actual changes in the characteristics of the convection associated with these extremes or to the more frequent occurrence of extreme events in these months, which could wash out some of the features through averaging. Examination of the precipitation features in the 75th, 90th, and 95th percentiles (not shown), which change the monthly relative distribution of extreme feature frequency, also indicated this decrease during August–October. The corresponding slight decrease in rain area in the mean storm features for these months points to an actual difference in the characteristics of both typical and extreme precipitation features.
TRMM PR precipitation features were used to understand the role of storm characteristics on the seasonal and diurnal cycles of precipitation in Costa Rica. The annual cycle in TRMM PR precipitation features for the four domains analyzed here is consistent with previous in situ studies (e.g., Hastenrath 1967; Waylen et al. 1996a; Magaña et al. 1999), with a wet season spanning May–November. Domain total rain volumes during the wet season are 4–5 times higher than the dry season from December to April. The wet season increase in domain total rain volume is due to a combination of increased storm frequency and storm area, although the relative contributions vary regionally. One of the interesting features to note is how the annual precipitation variability in different regions is driven by different aspects of the convection. For example, the annual precipitation variability for the Caribbean coastal domain is not primarily driven by the frequency of storms, but is more of a function of the characteristics of the individual storms, especially the stratiform storm area. In other regions, it appears to be more of a combination of the storm frequency and characteristics, which tend to have generally concurrent peaks.
It is also interesting to note that there is a slight peak in storm-scale conditional rain intensity in all regions prior to the onset of the rainy season as in Biasutti and Yuter (2013). They found higher conditional rain intensities prior to the core of the rainy season in monsoon regions because there was less stratiform precipitation included in the conditional rain-rate averages. The results here are similar; however, stratiform rain rates are also slightly higher during the dry season, though there are relatively few stratiform pixels.
The characteristics of storms, especially storm sizes and stratiform precipitation, appear to be related to several important large-scale features: the Caribbean low-level jet, the location of the ITCZ, the CHOCO jet, and cold air surges from midlatitudes. The Caribbean low-level jet that peaks in strength during the summer (Amador 2008) coincides with peaks in the storm area and rain volume for the Caribbean and central domains. The strong moisture transport from the Caribbean supports the development of MCS with large stratiform areas and suggests that they are responsible for a large fraction of the annual total rainfall. This is consistent with Zuluaga and Poveda (2004) and Poveda et al. (2006) who find that precipitation from MCS composes up to 70% of the annual total for the greater tropical Americas region. The increase in large MCS during the wet season is also likely related to convection coupled with easterly wave activity that peaks during June–November (Roundy and Frank 2004). Serra et al. (2010) show an area of enhanced easterly wave tracks that extends from the central Caribbean across Costa Rica during the summer. Provided that the Caribbean low-level jet is not too strong (Méndez and Magaña 2010; Serra et al. 2010), interactions with the jet may help to intensify these easterly waves (Serra et al. 2010) and could contribute to the noticeable increase in rain volume during the wet season that results from more intense storms (as evidenced by the higher convective rain rates) with large stratiform areas.
The importance of the jet also manifests in the characteristics of precipitation features on the Pacific side of the country during July and August (Magaña et al. 1999; Waylen and Quesada 2002; Small et al. 2007). A midsummer drought appears on the lee side of the cordillera, with rain volume and storm areas decreasing in the northwestern and southwestern domains as the strength of the jet increases. It is interesting to note that the decrease in rain volume is strongly dependent on the decrease in storm area rather than in storm frequency, indicating the importance of storm characteristics for rainfall during this period.
Another feature of importance is the location of the ITCZ. Storm frequencies in domains on the Pacific side are largely dependent on the northward migration of the eastern Pacific ITCZ. Convergence along the eastern Pacific ITCZ gives rise to convection with very large stratiform areas on the Pacific side during early and late summer, interrupted by the decrease associated with reduced Pacific SSTs and weaker low-level convergence (Magaña et al. 1999) during the midsummer drought. The late summer peak in precipitation in the northwestern region is also likely due to the influence of tropical cyclones as they move into the Caribbean and weaken or reverse the typical easterly flow. The peak in convective precipitation during September also corresponds to the peak precipitation in Liberia (on Costa Rica’s Pacific side), which Waylen and Harrison (2005) suggest is related to the peak in the North Atlantic tropical storms resulting in a circulation change that enhances precipitation on the Pacific side.
As the ITCZ moves farther north, it also allows another important feature to interact with the southwestern domain. During autumn, the low-level westerly CHOCO jet (Poveda and Mesa 2000) peaks in strength and brings moisture from the Pacific to Panama and southwestern Costa Rica. This results in a large increase in stratiform precipitation with convective and stratiform contributions becoming nearly equal.
Winter season precipitation occurs mostly in the Caribbean and central domains and is dominated by convective rainfall, a feature that is also obvious for this region in the results of Yang and Smith (2008) and somewhat in Schumacher and Houze (2003). However, in the latter study, the regional average for Central America includes a large oceanic area, resulting in higher stratiform fractions during boreal winter and spring than observed here and in Yang and Smith (2008).
For most regions except the Caribbean coastal domain, the diurnal cycle of convection illustrates features typical of land areas. Convection is most frequent in the early afternoon during the hours of peak daytime heating. Stratiform rain area and volume peaks a few hours later, suggesting the canonical evolution of MCS. The Caribbean coastal domain shows a different pattern, with a weak diurnal cycle and two peaks in rainfall. The daytime peak is likely related to the sea breeze, while the nighttime peak may be related to precipitation that forms during the nighttime secondary maximum of the Caribbean low-level jet or storms that form along the coast and move offshore.
The TRMM PR precipitation feature dataset was also used to explore the frequency and characteristics of extreme events, defined here as features with rain volumes in the 99th percentile. The frequency of extreme precipitation features peaks from May to November; however, there is a noticeable difference between extreme features at the beginning/end of the rainy season versus extreme features in the middle. Stratiform rain areas are much larger in extreme events at the beginning when extremes features are more frequent as the ITCZ moves through the southwestern domain and at the end of the wet season when extremes occur more frequently on the Caribbean side. Stratiform volume surpasses convective rain volume during these months, with no noticeable shift in stratiform rain rates for extreme features. In late summer and early autumn, extreme events are most frequent in the northwestern domains; however, they are typically 30% smaller with about 30% less rain volume than during early summer and late fall. It is unknown if this late summer decrease is a result of averaging more frequent extreme events or if other characteristics of the regional circulation result in smaller stratiform regions that are perhaps confined by topography during these months.
There are also differences between wet and dry season extreme events and between early and late dry season extreme features. Like the seasonal averages, convective rainfall still dominates during the winter extreme events even though convective rain rates are generally lower than in summer and autumn extremes. Early winter extremes, mostly in the Caribbean and central domains, are also different from late winter extremes that are primarily along the Caribbean coast. Extreme features have larger stratiform rainfall contributions and much higher convective rain rates during December–January relative to February–April. Schultz et al. (1998) also point to fundamental differences between the cold surges that make it to Costa Rica during the early and late winter, which could manifest in differences in the precipitation formation processes for these extreme events. They show a tendency for milder early winter cold surges with more maritime influence, which perhaps results in more moisture available for convection when there are stronger trade winds associated with the early winter cold surge events.
The results shown here suggest that the characteristics of the precipitating systems are important for understanding the variability in seasonal and diurnal rainfall, as well as extreme precipitation events. Dai (2006) shows that general circulation models have difficulty simulating convective/stratiform ratios, which has clear implications for accurately representing the regional rainfall totals in a region like Costa Rica where the relative contribution of stratiform precipitation is important but highly variable seasonally and regionally. Both the seasonal cycle of precipitation and frequency of extreme events in Costa Rica are projected to change in future climate scenarios (Alvarado et al. 2005). Giorgi (2006) indicate Central America as the primary tropical climate change “hot spot,” resulting from the large expected reduction in precipitation and increase in precipitation variability. It is important to understand how these projected precipitation changes will manifest not only in terms of total precipitation, but how the characteristics of these systems may change, especially as it relates to the potential for flooding rains in this region.
Funding for Alexander Peterson was provided by the Texas A&M University “Eco-hydrology of a Tropical Montane Cloud Forest” Research Experiences for Undergraduates (REU) program sponsored by the National Science Foundation under NSF Grant EAR-1004874. The authors would also like to thank Courtney Schumacher and Arelis Rivera for initial discussions on this project and Chuntao Liu for discussions on the precipitation feature database.