1. Introduction
Lake Victoria is the largest tropical lake of the African Great Lakes and the second largest lake in the world after Lake Superior in the United States. With an area of 68 800 km2, the lake supports the livelihood of more than 5 million people through the largest freshwater fishing industry in the world, producing around 1 million tons of fish per year and employing more than 200 000 fishermen that go every day to the lake (African Great Lakes Information Platform, https://www.africangreatlakesinform.org/). However, the lake also represents a global hotspot for severe nocturnal thunderstorm activity (e.g., Thiery et al. 2017; Roberts et al. 2022; del Moral Méndez et al. 2023; Virts and Goodman 2020), causing weather-related accidents on the lake, especially to the fishing and boat users’ community, which contribute to the global death toll from natural disasters (e.g., Cannon 2014; Watkiss et al. 2020; Dune 2022; Ogega et al. 2023; World Health Organization 2023). Drowning rates in the African region are the highest in the world (Tyler et al. 2017), with Lake Victoria being one of the water bodies registering one of the highest drowning rates per inhabitant. For instance, the drowning fatality rate in the Tanzanian lakeside communities is 231 per 100 000 people, with 81% of fatalities occurring while fishing (Whitworth et al. 2019). Drowning rates in the Ugandan lakeside communities are even higher, with 502 deaths per 100 000 a year (Whitworth et al. 2019). Several factors are listed as direct and indirect drowning causes in Kobusingye (2020) (i.e., the use of unseaworthy boats and overloading the boats, fleeing an act of crime, and hippopotamus attacks), with bad weather as the leading cause for boat accidents and fatalities (58%). Although accurate numbers of deaths and accidents are difficult to obtain due to the lack of reports (Kobusingye 2020), it is easy to obtain an idea of the impact of single accidents in the lake community by performing a quick search on the internet (e.g., British Broadcasting Corporation 2018; FloodList News in Africa 2019; Kombe 2022; Bashir 2023). Accidents on the lake typically result in at least a handful of fatalities, which contributes to the high number of deaths.
Convective activity over the Lake Victoria basin has been studied in the past primarily using constantly evolving satellite- and model-based approaches. This has provided an important and deep understanding of the convective and hydrological cycle in this region (Lumb 1970; Kayiranga 1991; Ba and Nicholson 1998; Yin and Nicholson 1998; Bedka et al. 2010; Woodhams et al. 2019). With new radar observations, especially from the High Impact Weather Lake System (HIGHWAY) field project (Roberts et al. 2022), we are now able to describe convection initiation, propagation, organization, and main storm radar-based attributes in Lake Victoria (del Moral Méndez et al. 2023, hereafter Part I). Subjective analysis of the convective cycles over Lake Victoria using high-resolution radar observations was first performed by Waniha et al. (2019) and Wilson and Roberts (2022). An extended objective analysis of the convective cycles is presented in the first part of this work (Part I). Part I established the basis for the work presented here, which analyzes, for the first time, preconvective environments over Lake Victoria for six different convective modes identified from the 2019 data collection period. Maximum convective activity occurs overnight and over the lake, with more organized systems, sometimes growing into upscale systems. Those systems have bigger vertical developments, greater intensities, and propagation velocities, with a greater possibility of producing convective straight-line winds and heavy precipitation than the less organized systems.
After identifying the storm modes taking place over Lake Victoria during the wet seasons in 2019, the goal of this second paper is twofold. First, we aim to identify key preconvective environmental factors that might help to discern between the different storm modes identified in Part I. We evaluate how these factors change throughout the day and the two wet seasons in 2019, as well as their possible role in storm-scale processes. Second, and equally important, we aim to provide a guide for forecasters to improve marine weather forecasting. The latter goal is also presented in Part I and adds to national and international ongoing capacity-building efforts for the Lake Victoria basin. As in Part I, the analysis conducted here uses open-source software, so the analysis can be replicated with a bigger database that encompasses the broader Lake Victoria region. The paper is structured as follows: section 2 presents a short background description of environmental factors affecting Lake Victoria, as well as the possible future climate scenarios discussed in the literature; sections 3 and 4 show the data and methodology used, respectively; section 5 describes the selection of the preconvective environments pertinent to the lake based on statistical analysis and presents the results of the analysis divided into three main environmental factors: thermodynamic, moisture, and wind parameters. Section 6 shows real cases, and finally, section 7 presents the discussion and conclusions of this paper.
2. Environmental conditions in Lake Victoria for convection
Recent studies of meteorological factors by Woodhams et al. (2019) and Nicholson et al. (2021b) show the daily and seasonal variability of moisture, mesoscale wind patterns, convective available potential energy (CAPE), and convective inhibition (CIN) in storm initiation over the lake. By analyzing case studies, Woodhams et al. (2019) show that for both “long-rains season” (storm case in May) and “dry season” (storm case in July), a high CAPE and low CIN combination are the leading environmental conditions for convection initiation. However, the local moisture source contribution will vary depending on the season, with greater specific humidity values and consequently greater CAPE and lower CIN for the long-rains season (Nicholson et al. 2021b). They also demonstrate that the vertical motion over the lake, analyzed with the omega vertical profile, is stronger during this season and that the wind patterns (convergence and divergence) over the lake might not be a controlling factor for the rain seasonality, although they play a major role on enhancing the convection when the katabatic flow is stronger during the wet season. During the dry season, CAPE and CIN are shown to be lower, with a lower specific humidity budget for storm initiation. However, this same study also finds that the enhanced convergence–divergence patterns over the lake might allow for storm initiation via moisture convergence due to collisions of density currents over the lake, as also seen in Woodhams et al. (2019). Nicholson et al. (2021b) and Woodhams et al. (2022) also found a humidity maximum (“bulge”) extending above the lake to the upper layers of the atmosphere in November (“short-rains season case”) compared to April (“long-rains season case”), with the maximum occurring over the lake at nighttime.
Woodhams et al. (2019) and Van de Walle et al. (2020) also demonstrate the important role that convection from the previous day during the second season plays in triggering new storms over the lake when strong, long-lived cold pools from convection to the northwest enhance the western land breeze, triggering secondary convection. More recent convection-permitting modeling studies in Lake Victoria basin show the importance of the strength of the trade winds and the orography on the intensity, location, and diurnal cycle of precipitation over the basin. They show that these winds curl around the Gregory Ridge, creating a northerly and southerly convergence over the basin, triggering nighttime convection. Convection is then enhanced and prolonged, aided by the land breeze over the lake (Van de Walle et al. 2020). Finney et al. (2020a), Anyah et al. (2006), and Williams et al. (2015) show that land–lake breeze processes triggering convection over the lake are affected by the explicit representation of convection in models and that Lake Victoria has an important role in the diurnal cycle in mesoscale processes in the basin. These findings on the importance of local moisture budgets, storm-scale factors, the role of the lake in mesoscale processes, and the synoptic influence of the trade winds demonstrate the complexity of the region and illustrate compound-type events in which the interaction of multiple physical processes across temporal and spatial scales leads to high-impact weather events.
The complexity of these compound-nature events in Lake Victoria basin is also evident when considering future climate projections, as shown in recent studies. For instance, convection-permitting regional climate simulations over Lake Victoria basin project dramatic differences between warming for the land and lake. This differential warming would affect the current diurnal pattern by weakening the nighttime land breeze which would likely decrease the shear and triggering mechanisms and storm occurrences over the lake at night. Consequently, this would result in a decrease in the mean precipitation over the lake by the end of the century (Thiery et al. 2016; Van de Walle et al. 2021). However, this latter study shows that the strength of extreme convective events, such as precipitation peaks over 50 mm h−1 and severe wind gusts exceeding 30 m s−1, is projected to increase (Van de Walle et al. 2021), probably due to the strengthening of thunderstorm dynamics (e.g., updraft and cold pool intensity). Similar results are shown by Finney et al. (2020a), demonstrating that more convection-permitting model studies over the region are still needed, since parameterized convection does not fully capture the local drivers of extreme rainfall events, and there is not full confidence on existing climate change projections over the region. Finally, a similar response is found in novel research investigating lightning changes with climate change in Africa with convection-permitting climate change simulation and ice-based lightning parameterizations (Finney et al. 2020b), showing an increase in the frequency of strong lightning events and therefore storm intensification. All of these recent studies demonstrate that there is an imperative need to understand convection organization and dynamics from an environmental and storm-scale perspective to improve knowledge and resilience against severe weather threats for the local population as well as fishing and agriculture communities. Additionally, it underscores the necessity to anticipate and adapt to potential changes driven by climate projections.
3. Data
a. Reanalysis data
This study utilizes ERA5 reanalysis data (Hersbach et al. 2020). The ERA5 dataset covers from 1959 to the present with a spatial and temporal resolution of 30 km and 1 h, respectively. ERA5 is obtained from a 4D-Var data assimilation and model forecast of the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System (Haiden et al. 2021). The dataset contains 137 hybrid sigma/model levels in the vertical, with the top at 0.01 hPa, that are interpolated to 16 potential temperature, 1 vorticity, and 37 pressure levels.
Lake Victoria basin is a tropical region influenced by mesoscale patterns and local terrain-lake interactions (Fig. 1). Therefore, for data managing and processing purposes, we selected 20 ERA5 pressure levels (up to 100 hPa) plus the surface level, with a higher density of levels within the planetary boundary layer. Comparisons were made with all 37 levels, and we did not find significant differences. Variables for each level, including surface1, are temperature (T; in K), dewpoint temperature (Td; in K), wind speed (wspd; in m s−1) and direction (wdir; in °), and pressure (p; in hPa). ERA5 incorporates the Freshwater Lake (FLake) model (Dutra et al. 2010), a new parameterization scheme in the land surface model that uses lake cover (proportion of a grid box covered by an inland water body) and lake depth to identify regions with inland water bodies, resolving down to subgrid scales. ERA5 does not include observed lake surface temperatures, and despite an overestimation of lake water surface temperatures, it is demonstrated to capture mesoscale land–lake breeze patterns in Lake Victoria (Minallah and Steiner 2021).
Satellite image of the region of interest of the present study. The S-band radar discussed in Part I is located in Mwanza, Tanzania (see TMA label and yellow star on the map). The black asterisk represents the Nairobi (Kenya) UAS. White lines denote country borderlines; black polygons over the lake with black Roman numerals represent sectors dividing the lake that are used in the operational marine forecasts. Detailed information about the forecasts can be found in Part I, section 2d.
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
b. Nairobi sounding
For statistical verification and correlation analysis, we initially compare ERA5-derived profiles at the Nairobi station location with the observed Nairobi upper air sounding in Fig. 1 (KMD 2020). According to the proximity-inflow method (Rasmussen and Blanchard 1998), to get the representative air mass for a weather event, a sounding should be closer than 400 km, and the elapsed time between the event and the sounding data collection should not be greater than a 9-h window (6 and 3 h before and after the event, respectively) (Rodríguez and Bech 2018). The distance from the Nairobi upper air sounding station (UAS) to Lake Victoria is ∼300 km to the eastern lake shore and ∼430 km to the center of the lake. The only other UAS available during the field project, in Lodwar, Kenya, is >500 km away from the lake center, even further away than the Nairobi station. For this reason, the use of the UASs in this work is restricted to Nairobi for the ERA5-derived profile verification. For verification, we used a total of 187 sounding profiles from Nairobi station from August to December 2019 to compare with the ERA5-derived profiles at the Nairobi UAS location. Unfortunately, we were not able to perform analysis for both wet seasons since the Kenya Meteorological Department (KMD) Nairobi UAS did not become operational and compliant with WMO requirements with twice daily launches at 1200 and 0000 UTC until August 2019 during HIGHWAY.
4. Methods
a. Convective mode classification and analysis stratification
Convective storms are identified and tracked throughout their life cycle by running the Thunderstorm Identification, Tracking, Analysis, and Nowcasting (TITAN) algorithm (Dixon and Wiener 1993) on data from the Tanzania Meteorological Authority (TMA) S-band weather radar in Mwanza (Part I). Thunderstorms are identified and tracked continuously and are classified into different convective modes depending on radar morphological attributes. There are six thunderstorm classifications ranging from small systems to bigger and more organized ones and labeled: isolated, multicell unorganized, multicell nonlinear, multicell linear, upscale nonlinear, and upscale linear (Fig. 3 in Part I). Each storm is assigned a different convective mode label for every radar volume, which might change from volume to volume due to storm evolution. Therefore, all storms are part of a bigger system that has different single labels. However, only the label with the maximum organization/bigger storm mode is considered for the analysis (see Fig. 4 in Part I for a detailed description of a storm system). The preconvective analysis is done for the totality of storms identified in Part I: 576 and 1253 isolated, 294 and 670 multicell unorganized, 1 and 23 multicell nonlinear, 20 and 41 multicell linear, 0 and 4 upscale nonlinear, and 1 and 13 upscale linear, for seasons I and II, respectively (see Fig. 5 in Part I).
Once storms are identified, tracked, and classified in convective modes, preconvective environment indices are retrieved for each storm. The analysis follows the same data stratification as in Part I, based on seasons, timeslot, lake sector, and system longevity.
-
Seasons: Season I data are from March to May 2019 and season II data are from October to December 2019.
-
Timeslots and sectors:
-
Morning (M): from 0600 to 1159 LT2 with convection in sectors III, IV, and V (southwest region of the lake).
-
Afternoon (A): from 1200 to 1759 LT with convection in sectors VI, VII, VIII, IX, and X (east and northeast region of the lake).
-
Night before midnight (NbM): from 1800 to 2359 LT with convection in sectors VII, VIII, IX, and X (eastern and central region of the lake).
-
Night after midnight (NaM): from 0000 to 0559 LT with convection in sectors I, II, III, IV, VII, and X (central and western region of the lake).
-
Although storms are tracked through their entire life cycle and can be present in different sectors and during different timeslots, we stratified the results based on the maximum organization time and location, as in Part I. This allows us to identify predominant timeslots and sectors and to conduct statistical analysis. A detailed description of the morphological and TITAN-derived attributes, the thresholds used for the classification of each storm mode, and the stratification parameters are given in Part I.
b. Retrieval of preconvective atmospheric vertical profiles from ERA5
Atmospheric vertical profiles are retrieved from the ERA5 dataset for each storm location and 6 h before the storm occurrence (event). The location is obtained from the volumetric reflectivity field centroid, retrieved from TITAN in the identification and tracking process. This location corresponds to the maximum organization of the convective system classification (Part I), and therefore, we consider it to be representative of the “event” location. For the selection of the elapsed time between the profile and the event, analysis of stability parameters evolution every 3 h (from t − 9 to t − 0 h, with t − 0 h defining the time of the event) showed that t − 6 h is the best discriminant between convective modes (not shown). Closer in time to the event, the atmosphere is already showing the precipitation occurring in the region with a saturated low troposphere and is unable to depict the preconvective environment. For this reason, we have selected atmospheric profiles 6 h before the storm as the preconvective environment sounding.
c. Selection of convective parameters
Convective parameters have been used extensively to investigate discriminators for thunderstorm initiation, organization, and associated hazards. Knowing the key factors favoring convective and severe weather environments helps forecasters around the world to identify potential severe threats to the population (Evans and Doswell 2001; Brooks 2009; Hitchens and Brooks 2014; Taszarek et al. 2020). Due to the lack of a severe weather database or case-based analysis for the Lake Victoria basin, proximity-sounding-derived parameters are unknown and are not available to identify thresholds or covariate parameters for severe weather forecasting purposes. Therefore, in this study, we consider that the maximum organization is our proxy location for storm maturity, and we identify key parameter precursors of the different organization modes over Lake Victoria as previously determined in Part I. This can help identify the likelihood of severe weather over the lake. CAPE, CIN, midlevel lapse rate, and low- and midlevel humidity values allow for the identification of convection initiation potential, updraft strength, and potential thunderstorm intensity (Johns et al. 1993; Doswell et al. 1996; Rasmussen and Blanchard 1998; Thompson et al. 2003; Craven and Brooks 2004; Gensini et al. 2014; Púčik et al. 2015; Prein and Holland 2018; Chen et al. 2020). Kinematic parameters, such as deep-layer bulk shear, allow for the identification of storm development, organization, and propagation (Weisman and Klemp 1982; Bunkers et al. 2000; Coniglio et al. 2007; Coffer et al. 2019). A good summary of discriminants for severe weather events in Europe and the United States can be found in Taszarek et al. (2017). The work presented here tries to identify those discriminants for the first time in an East African region. The indices presented in this work are calculated using the open-source sounding analysis toolkit SHARPpy (Blumberg et al. 2017).
A prior comparison of the ERA5-derived atmospheric sounding with the observed Nairobi sounding was performed to identify possible dispersion of the data that could propagate into the calculation of the stability indices studied in this work. This was done by considering the Nairobi profile as ground truth and computing the standard deviation of the ERA5 profile for temperature, dewpoint, virtual temperature, and wind direction and speed (plots not shown for brevity). We presume that the Nairobi sounding data are being assimilated into ERA5 after becoming operational, as part of the in situ data provided by the WMO Information System (WIS). The analysis was performed for 187 soundings and considering pressure levels from 850 to 100 hPa. In general, ERA5 profiles present a warmer and moister atmosphere at 850 hPa: the mean temperature at 850 hPa is 1°C higher for ERA5 than for the Nairobi sounding, with a standard deviation ranging from −1° to 2°C. The profile is even warmer for the dewpoint and potential temperatures, with mean values up to 2°C more than the observed ones. Similar behavior is shown for the upper levels (from 300 to 100 hPa) where ERA5 presents a moister atmosphere in general. ERA5 uses surface elevation data interpolated from a combination of SRTM30 and other surface datasets. The boundary layer is parameterized, as well as subgrid orography features that are too small to be resolved. We anticipate that this impacts the accuracy of results over complex terrain areas. In our case, the comparison is made in a mountainous region (Nairobi sounding and Lake Victoria are at 1795 and 1134 m MSL, respectively), which could explain the differences described above.
Figure 2 shows some of the ERA5-derived parameters tested for representativeness for this study. The R2 in Fig. 2 represents the Spearman’s rank correlation coefficient (Spearman 1904) for nonnormal data, previously tested with the Shapiro–Wilk normality test (Shapiro and Wilk 1965). The best correlation for CAPE values is obtained when considering the mean layer (ML) parcel (Fig. 2b), with a correlation of 0.8 compared to 0.61 and 0.52 for the most unstable (MU, Fig. 2a) or surface-based (SB; Fig. 2c) parcels, respectively. This aligns with ERA5 presenting a warmer and moister low-level atmosphere, which impacts the retrieved indices for those parcels when using layers closer to the surface. An analysis of the temporal evolution of the different parameters also showed that MLCAPE would increase as expected closer to the event time, depicting a more unstable atmosphere and a decreasing ML level of free convection (MLLFC) (not shown). Figures 2d and 2e also show good correlations (0.73 and 0.84) on the downdraft CAPE (DCAPE) and the ML lifting condensation level (MLLCL), respectively. DCAPE is calculated using the minimal potential temperature of the lowest 400 hPa. ML is calculated by averaging over the lowest 100 hPa.
ERA5 (reanalysis) vs Nairobi radiosonde (observed) calculations of (a) MU; (b) ML; (c) SB CAPE and (d) DCAPE (J kg−1); (e) LCL (m AGL); (f),(g),(h) LR (°C km−1) for the 700–500-hPa, 0–3-km, and 3–6-km layers; (i),(j) RH (%) for the 850–500- and 700–500-hPa layers; (k),(l) bulk wind shear (Wind Shear; m s−1) for the 0–3- and 0–6-km layers; (m),(n) SRH (m2 s−2) for the 0–1- and 0–3-km layers; and (o)–(r) wind components (m s−1) for the 0–1- and 0–3-km layers. The R2 represents Spearman’s rank correlation coefficient for nonnormal data. The best fit and one-to-one lines are respectively shown in blue and black.
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
Moisture and lapse rate parameters (Figs. 2f–j) showed correlations > 0.66. Of particular interest are the two different groups in the low-level lapse rate (LR) (LR0–3km, Fig. 2g), where one group of points is close to 6°C km km−1, which is distinctly different from the group of points located around 8.5°–9°C km−1. The storms within the second group of points, showing steep lapse rates greater than 7°C km−1, have greater potential to become severe, with a high probability for strong downdrafts.
The best correlation in wind shear is found for the deep layer shear (Wind Shear0–6km, Fig. 2l), with a correlation of 0.644, which suggests potential convective organization. Also important is the correlation found for the wind components (Figs. 2o–r). This may help to identify the prevailing winds before the different storm types and between the main regions of the lake. Of special interest is the good correlation found for the wind components for the layer up to 3 km (0.92 and 0.93 in Figs. 2q,r, respectively), which allows the identification of wind patterns in the lower troposphere characteristic of the lake–land cycle.
Finally, although not high, both storm-relative helicity (SRH) from the 0–1- and 0–3-km layers (SRH0–1km and SRH0–3km, Figs. 2m,n, respectively) are close to 0.5. The latter is also considered in this study to identify if some times of the day present a more favorable atmosphere for storm rotation. Other parameters were tested, such as the bulk Richardson number (BRN), CIN, and the level of free convection (LFC), although none of them presented a correlation parameter greater than 0.3. As explained previously, ERA5 does not depict surface values correctly, especially over the lake. This could lead to the overestimation of CIN found in our test, which would also explain the overestimation found for MLLFC: the higher the LFC, the greater the CIN. Because of the low correlation found for these parameters and the surfaced-based parcels, these are not considered in the present analysis.
Resulting from the statistical intercomparison analysis, the following parameters are considered in this study: MLCAPE, DCAPE, MLLCL, relative humidity (RH)Surface–850hPa, RH850–500hPa, RH700–500hPa, LR0–3km, Wind Shear0–6m, and SRH0–3km. Lake Victoria is at ∼1135 MSL, and therefore, the RHSurface–850hPa represents the ∼200 m above the lake surface, depicting the near-surface boundary layer humidity. These statistical analyses give us the confidence to use the ERA5 reanalysis and retrieved soundings to describe the representative preconvective environments.
5. Results
a. Instability
Figure 3 shows MLCAPE, DCAPE, and MLLCL values for the different convective modes (colors), separated by season (pattern) and time of the day (A, NbM, NaM, and M). Values are aggregated by sector within each timeslot due to minimal differences (Part I). For instance, M timeslot will include sectors III, IV, and V. A first difference is depicted between seasons with season I (March–May, patterned boxplots in Fig. 3) presenting generally higher values and with higher variability in the parameters throughout the day, especially for MLLCL and DCAPE mean values. MLCAPE (Fig. 3a) mean values are higher for the night timeslots (NbM and NaM) and season I, with values between 900 and 1000 J kg−1, compared to the mean values between 500 and 800 J kg−1 found in season II, for almost all convective modes in that same timeslot. CAPE is dependent on moisture and temperature, accumulating daytime heating. Nicholson et al. (2021b) demonstrate that February presents the highest CAPE values of all seasons, although also the highest CIN values, preventing convection over the lake. The high values of MLCAPE in season I may be a reflection of the large energy budget available that has not yet been consumed. Of particular interest is the lack of significant difference between seasons during morning and afternoon; however, in those timeslots, we do observe differences between convective modes. The isolated, multicell unorganized, and multicell nonlinear modes are the ones with higher MLCAPE values compared to the more organized systems at nighttime and for season I. Again, this may be the reflection of a greater energy budget available at the beginning of season I convective activity in 2019, early at night. Also of interest is the high MLCAPE mean value for the upscale linear cases in NaM timeslot (compared to the other modes for the same season) and for the multicell linear cases in the afternoon, both for season II. Other high values are also present but not considered relevant due to the lack of large enough sample cases (e.g., upscale linear in season I and M or multicell linear in season II and A).
MLCAPE (J kg−1), DCAPE (J kg−1), MLLCL (m AGL), and difference between MLLFC and MLLCL (m AGL), for the six different modes (colors) and the two seasons (patterns). Indices are stratified depending on the time of the day: A, NbM, NaM, and M. The boxplot’s lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles), and the upper and lower whiskers extend from the corresponding hinges to the largest/smallest values, at most 1.5 times the IQR. Dots in boxplots indicate the mean value.
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
The biggest differences are found for DCAPE (Fig. 3b), especially regarding seasonality, being 200–400 J kg−1 higher on average for season I for all the convective modes and timeslots (750–950 J kg−1 compared to 400–550 J kg−1, on average). Furthermore, higher DCAPE values are found during nighttime and morning timeslots (NbM, NaM, and M) but with no significant difference between convective modes. DCAPE over 1000 J kg−1, which is found for several cases within isolated and multicell unorganized at NaM and M timeslots, and especially the multicell linear at NbM and M timeslots in season I are often significant values to be considered. These might imply the downward transport of higher momentum air to the surface, which could potentially result in strong and gusty winds at the surface.
Finally, following a similar pattern, but even more apparent, the MLLCL (Fig. 3c) level in season I is, on average, between 100 and 200 m AGL higher than in season II in nighttime slots and even higher (500 m) during the afternoon and morning. Mean values also vary depending on the timeslot. For example, for season I, MLLCL for all convective modes oscillates around 1000–1300 m AGL. However, in the afternoon and morning, MLLCL is higher for the multicell unorganized and the multicell linear, respectively. On the other hand, season II presents lower values of MLLCL on average, with a more pronounced difference between timeslots. In this case, mean values are lower during the morning and afternoon (between 800 and 1000 m AGL, on average) in comparison with the nighttime timeslots (between 1000 and 1300 m AGL, on average). A similar pattern as MLLCL is found for MLLFC values (not shown), with a marked diurnal cycle and higher values overnight. The differences between seasons are almost imperceptible, especially between the isolated and multicell unorganized cases, although for season II, MLLFC is higher in the morning than in season I. Figure 3d shows the differences between MLLFC and MLLCL for which the same daily cycle is found, although the nighttime timeslots present the higher differences between both levels (between 650 and 950 m, compared to 450–700 m in the morning timeslot). This indicates that more mechanically forced ascents are needed at nighttime compared to daytime. These are probably helped not only by the land breeze, the orographic flow, and moisture convergence due to strong trade winds overnight, as highlighted in previous research but also by cold pools and outflows from preceding convection. Although a bigger difference between MLLFC and MLLCL overnight indicates the possibility of more dry air entrainment, this might be reduced by a deeper humidity layer, as is seen in Fig. 4 and discussed in the following section.
As in Fig. 3, but for RH from (a) the surface–850-, (b) 850–500-, and (c) 700–500-hPa layers (RHSurface–850hPa, RH850–500hPa, and RH700–500hPa, respectively; %) and (d) LR0–3km (°C km−1).
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
b. Moisture and lapse rate
Figure 4 shows RHSurface–850hPa, RH850–500hPa, RH700–500hPa, and LR0–3km values for the different convective modes (colors), separated by season (pattern) and time of the day (A, NbM, NaM, and M). As in Fig. 3, values are aggregated by sectors within each timeslot. Similar to the results from Fig. 3, the first noticeable difference is shown between seasons, although in this case, season II presents higher values for relative humidity (Figs. 4a–c). Lake Victoria presents high low-level moisture values (RHSurface–850hPa, in Fig. 4a), in both seasons and for all timeslots. Mean RHSurface–850hPa values oscillate around 75%, with even higher values (around 85%) during the insolated timeslots, morning and afternoon. These two timeslots also show greater differences between seasons. Season II has mean values higher than 80% for almost all convective modes in the insolated timeslots, except for the upscale nonlinear mode in the morning, whereas mean values tend to be lower than 80% for all timeslots during season I. Midlevel moisture values for the two layers considered (RH850–500hPa and RH700–500hPa, in Figs. 4b,c, respectively) show a similar pattern to those of low-level moisture values, although the diurnal cycle is much more pronounced, as well as the season differences. Differences between the two layers are minimal, showing the same diurnal pattern and depicting the higher low-level RH values for the layer from 850 hPa. In the case of season I, mean values oscillate from 55%, for the multicell linear case at NbM (the lowest value) to 73% for the isolated mode during A, being this timeslot the one with the highest values. The same diurnal variability is found for season II, although values are 5%–10% higher, especially during the M and A timeslots, the latter being the one with the highest values (Figs. 4b,c). These results partially agree with Nicholson et al. (2021b) and Woodhams et al. (2019, 2022), the latter including in situ aircraft observations. They found a humidity maximum (bulge) extending above the lake to the upper layers of the atmosphere in November compared to April; similarly, we found that season II (OND) has higher values than season I (MAM). However, Nicholson et al. (2021b) and Woodhams et al. (2022) show that this bulge of humidity is maximum over the lake at nighttime, unlike our results showing higher values during the afternoon. This could be the result of including profiles with already initiated convection in the analysis and the ERA5 reanalysis depicting higher humidity values than expected. Furthermore, when considering differences between convective modes, of particular interest are the low mean and median RH700–500hPa values (52% and 45%, respectively) during the NbM timeslot for the multicell linear convective mode in season I (Fig. 4c). This presents an interesting difference of potential hazards from similar convective modes at different times of the day. While high-humidity profiles in a linear mode during the morning would be associated with higher precipitation rates, the dry midlevel layer found at night could lead to evaporative cooling, potentially resulting in strong winds at the surface, posing a risk for fishermen operating on the lake.
Low-level LRs (LR0–3km in Fig. 4d) also show a noticeable difference between the two wet seasons, with season I presenting steeper lapse rates varying between 0.5° and 1°C km−1 higher than season II. In general, the lapse rate is steeper in NaM and M timeslots, with mean values between 7.5° and 7.7°C km−1 for season I and between 6.8° and 7.3°C km−1 for season II. Values between 6° and 9.8°C km−1 imply a conditionally stable atmosphere, usually in need of enough moisture to lift the air parcels to reach free convection, which results to be sufficient in both seasons. The high variability [interquartile range (IQR)] found for specific modes in the morning timeslot is also interesting. Some cases of isolated, multi-unorganized, and organized nonlinear modes present values up to 8.25°C km−1, sometimes even greater (isolated and multicell unorganized with values higher than 8.5°C km−1), indicating that those modes in the morning hours, especially during season I, could present strongest downdrafts due to the well-mixed boundary layer. Finally, it is worth mentioning that upscale modes (linear and nonlinear) present lower low-level lapse rate values in all timeslots, compared to the rest of the modes (0.5°C km−1 lower on average). This could be a result of those modes being in the later stages of isolated or unorganized convection and, therefore, in a less well-mixed low-level environment.
c. Kinematic parameters
Figure 5 shows the bulk layer wind difference (SHEAR0–6km in m s−1, Fig. 5a) and SRH for the layer 0–3 km (SRH0–3km, in m2 s−2, in Fig. 5b). The first noticeable result is the difference between the nighttime and daytime bulk shear mean values and variability for the two seasons studied in 2019. In general, mean values of bulk shear are higher for season I in the A, NaM, and M timeslots while mean values are higher in season II for all convective modes during NbM. During NbM in season II, mean values for isolated and unorganized modes are around 12 m s−1 and between 12.5 and 17 m s−1 for the rest of the modes. Values between 10 and 20 m s−1 representing moderate shear (Markowski and Richardson 2010) indicate the capacity to organize storms into multicell systems with longer life cycles. Therefore, all convective modes during the NbM timeslot and season II have the optimal environment to organize in bigger systems. Furthermore, there exists a high variability in all timeslots regardless of the season, with the NbM timeslot containing some cases with up to 20 m s−1 or greater of bulk shear (Fig. 5a). Although there is no absolute threshold value, this could indicate the capacity of rotating updrafts and the capacity for some storms to become severe, a major risk for fishermen (e.g., Markowski and Richardson 2010; Púčik et al. 2015).
As in Fig. 4, but for bulk wind difference (Shear0–6km; m s−1) and SRH0–3km (m2 s−2).
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
Figure 5b shows storm-relative helicity for the layer 0–3 km, indicating the potential for tornadic phenomena in the different convective modes. As expected, SRH0–3km is not as high in Lake Victoria as it would be in other regions of the world, such as on the Great Plains in the United States. SRH0–3km mean values oscillate around 50–70 m2 s−2, indicating that there is almost no potential for supercell tornadogenesis over the lake if we compare those values to the United States thresholds, where SRH0–3km needs to be higher than 250 m2 s−2 (NOAA/NWS Storm Prediction Center). However, there are aspects worth mentioning. First, there exists some difference between seasons during NaM and M timeslots, especially for the isolated, unorganized, and multicell linear modes. Season I presents higher values for those modes than season II. Furthermore, some of those cases reached up to 160 m2 s−2. This is evident in the case of multicell linear, presenting a mean value during the NaM timeslot of 100 m2 s−2 and higher magnitudes of 170 m2 s−2 during the M timeslot. This, combined with the high bulk shear for the same timeslots, may lead to embedded rotations in more dynamically complex systems. Kiwanuka-Tondo et al. (2019) show that one of the major concerns for fishermen operating in the lake at night is waterspouts. The precursors presented here show how, indeed, storms at nighttime have the necessary ingredients to organize in linear systems that could have embedded rotations in them.
6. Real examples
As stated in the introduction section, one of the goals of this study is to provide some guidance for the Lake Victoria marine forecasters. Therefore, it is of vital importance to provide some more concise information on how the analysis here presented should be used. For example, we encourage East African forecasters to pay attention to parameters related to instability and kinematics on those days when convection is forecast. These could be in the form of retrieved soundings with indices calculated or in the form of charts for a better spatial location. Forecasters should pay attention to the evolution of MLCAPE and DCAPE before convection is forecast over the lake. Convection might be in isolated or unorganized mode (clustered) with nonthreatening weather indices in model outputs. However, the higher those values, the higher the potential for deeper updrafts. This will increase the potential for severe wind gusts and heavier rain over the lake. Those storms may produce stronger downdrafts and cold pools and therefore also originate strong outflows causing high waves some kilometers away from the storm core and initiating new storms nearby. An example of an upscale linear system, seen with the Tanzania S-band radar, is shown in Fig. 6 which depicts a north–south line of strong precipitation (Fig. 6c) and inflow notches depicting shearing instabilities, and possible later rotations, along the leading edges of the squall line (squares in Fig. 6f). Time steps before the maximum organization is depicted in Figs. 6a–e to show the system evolution. The red dot shows the radar location, and the black dot shows the ERA5-retrieved sounding 6 h before (1900 UTC) and the maximum organization (0100 UTC the day after). CAPE and DCAPE values are high showing the possibility for deep convection with hazardous downdrafts 6 h later in the sounding location (black dot at 1900 UTC, Fig. 6a). Kinematic indices are lower than expected for which moderate values of bulk shear between 10 and 12 m s−1 and SRH0–3km higher than 100 m2 s−2 would be needed for this type of long-lasting convective mode capable of producing embedded rotations in the forward flank. This is probably due to ERA5 not fully representing surface wind parameters. Detailed case analysis and increased vertical representation of the atmosphere are needed.
Upscale linear convective system on 28 Mar 2019, as seen from the S-band radar in Mwanza, Tanzania (red dot), in 3-h intervals: (a)–(c) reflectivity factor Z (dBZ), (d)–(f) radial velocity (m s−1), and (g) ERA5-retrieved sounding and hodograph at 1900 UTC. The black polygon line shows the Lake Victoria shoreline, and the black dot represents the ERA5-retrieved sounding location at 1900 UTC. The image is centered in the southwestern region of the lake.
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
On the other hand, if convection is forecast overnight, we encourage forecasters to pay extra attention to midlevel moisture values and kinematics. High bulk shear values might be expected, which combined with low midlevel (700 hPa) moisture values and/or high storm-relative helicity could indicate the capacity of convection to organize into linear systems with embedded, dangerous wind-related phenomena. An example of a case with that type of convection is shown in Fig. 7. The reflectivity thin line (in yellow in Fig. 7c) represents the leading edge of the outflow associated with the intense multicell convective system. Figure 7f shows the radial velocity field associated with that convective system, with very high wind speeds just above the surface of the water. Hazards may include strong linear convective winds at the surface and even embedded waterspouts. These phenomena can produce damaging winds and high waves over the lake, affecting a large region over longer periods, and increasing exposure of fishermen operating on the lake.
As in Fig. 6, but for convective activity on 9 Oct 2019, as seen from the S-band radar in Mwanza, Tanzania (red dot): (a)–(c) reflectivity factor Z (dBZ); (d)–(f) radial velocity (m s−1); (g) ERA5-retrieved sounding and hodograph at 2100 UTC. The black polygon line shows the Lake Victoria shoreline, and the black dot represents the ERA5-retrieved sounding location at 1900 UTC. The image is centered in the southwestern region of the lake.
Citation: Monthly Weather Review 153, 6; 10.1175/MWR-D-24-0074.1
Finally, wind speed and direction are equally important, not only during preconvective periods and forecasting but also when monitoring convection that has already been initiated. Lake breeze interactions with existing cold pools from previous convection can create new convection overnight and until the early morning, especially in the western and center regions of the lake (e.g., Woodhams et al. 2019, Odongo et al. 2022).
7. Discussion and conclusions
The work presented here constitutes the second part of a two-part paper analyzing environment attributes and convective activity over Lake Victoria, one of the hotspots for severe convection in the world. In this paper, we analyze ERA5-derived preconvective environments of the six different convective modes identified in Part I, contributing to the small group of papers that have studied the variability of meteorological fields over the lake (e.g., Woodhams et al. 2019; Van de Walle et al. 2020; Nicholson et al. 2021a,b). The results presented here are consistent with those presented in Part I, showing that the most severe preconvective indicators are found during the nighttime periods, preceding the most organized convective modes and with some differences between the two wet seasons. Preconvective environments during the two wet seasons in 2019 are analyzed for the convective modes previously identified by radar (Part I). The environmental parameters were divided into four main groups listed in three categories for this study: 1) instability, 2) moisture, and 3) kinematics. The main findings for those three groups, based on seasonality and time of the day, are described below. We have not found significant differences between lake sectors shown in Fig. 1, and therefore, we do not discriminate our results based on these weather service forecast zones.
a. Instability
-
Season I [March–May (MAM)] presents generally higher instability values and variability throughout the diurnal cycle than season II [October–December (OND)], especially for MLLCL and DCAPE.
-
The isolated, multicell unorganized, and multicell nonlinear convective modes exhibit higher MLCAPE values compared to the more organized systems at nighttime and for season I.
-
Several cases of the less organized modes (isolated and multicell unorganized) at NaM and M present MLCAPE larger than 1000 J kg−1, a suitable ingredient to develop deeper storms with bigger downdraft depths (assuming storms with the same tropopause level and no dilution).
-
DCAPE is higher on average for season I for all the convective modes and timeslots. This increases the potential for strong and damaging winds at the surface, posing a major risk to fishermen. This is the result of higher downward transport of higher momentum air to the surface.
-
MLLCL, and the difference between MLLCL and MLLFC, is higher during nighttime for all convective modes. This indicates the greater need for nighttime forcing mechanisms (e.g., cold pools from preceding convection) to initiate storms. This may be overcome by the orographic flow converging on the lake at night, and deeper levels of humidity in low-to-mid levels.
b. Moisture
-
Season II presents higher (850–500 and 700–500 hPa) RH values than season I, especially for diurnal timeslots (A, M, related to heating over the lake). These findings are partially consistent with Woodhams et al. (2019) and Nicholson et al. (2021a,b), which describe a humidity maximum (bulge) extending above the lake to the upper layers of the atmosphere in November compared to April (season II compared to season I).
-
The multicell linear mode in season II and at nighttime (NbM) presents significantly low RH700–500hPa (median of 45% and mean near 53% with some values as low as 41%). Those values indicate the potential for evaporative cooling, possibly resulting in strong winds at the surface, increasing the risk for fishermen.
-
Low-level lapse rates (0–3 km) are steeper in season I than in season II but are more variable for some modes during the M timeslot. The isolated, multicell unorganized, and multicell nonlinear modes present values around or greater than 8.25°C km−1, which could indicate stronger updrafts.
c. Kinematics
-
Bulk shear is higher in season I for all timeslots except for NbM, which is higher in season II. At NbM, isolated and multicell unorganized modes present bulk shear values that indicate the capacity to organize into multicell systems, with some events indicating the capacity to have rotating updrafts. This represents an extended risk for fishermen operating in the lake at night, since they can encounter systems that last longer, cover larger areas, and could present severe phenomena, such as strong wind gusts or even waterspouts (Part I; Roberts et al. 2022).
-
Mean values for SRH0–3km oscillate around 50–70 m2 s−2, with no potential for tornadogenesis over the lake. However, season I presents higher values for the isolated, unorganized, and multicell linear modes during NaM and M timeslots.
The multicell linear mode at NbM and NaM can reach more than 160 m2 s−2, even higher during the M, which combined with the high bulk shear may lead to embedded rotations in more dynamically complex systems.
e. Limitations of current work and future research
We acknowledge that limitations are present in the current study, also encountered in Part I. First, ERA5 retrievals have coarse spatial and temporal resolution, and the lake is parameterized. This is not the ideal final data source to represent storm-scale parameters with the best accuracy to discriminate between indices. Additionally, the data sample is small for the storm modes with higher organization having only one year of data available. This affects the results in two ways: first, we cannot properly test the significance of the differences between seasons and time of the day of the analyzed indices, which might lead to a misrepresentation of the truth in those groups with a small sample (i.e., upscale linear). Second, we cannot screen out ERA5-retrieved profiles that already contain precipitation 6 h before the storm maximum organization since we would reduce the sample even more. This might introduce errors in the analysis since some profiles might have greater moisture values than they would originally, but it could also reduce instability and kinematic indices from profiles representing ongoing convection.
It is important to note that the results of this study are restricted to analyzing the radar data from 2019, which had an extremely wet “short rains” season (OND), which is our season II (Nicholson et al. 2022; Wainwright et al. 2021; Palmer et al. 2023). This means that some of our results might be strongly affected by interannual variability and hence would explain the big difference noted for this year among both seasons. Rainfall variability during the “long rains” (here season I) and short rains (here season II) is influenced by climate variability modes like El Niño–Southern Oscillation (ENSO), the quasi-biennial oscillation (QBO), the Indian Ocean dipole (IOD), and the Madden–Julian oscillation (MJO) (e.g., Black et al. 2003; Indeje et al. 2000; MacLeod et al. 2021; Palmer et al. 2023; Shaaban and Roundy 2017), as well as tropical cyclones (e.g., Finney et al. 2020c; Kebacho 2022; Kilavi et al. 2018), and the Congo airmass (Dyer and Washington 2021). Season II in 2019 over East Africa was one of the wettest seasons on record since 1985, which might be influenced by the warm SST anomaly in the western Indian Ocean (strong Indian Ocean dipole event, Nicholson et al. 2022; Wainwright et al. 2021). Interannual variability could have influenced our results too, especially if season I was found to not be that abnormal in 2019, compared to season II. Further analysis to connect the differences found in the instability, moisture, and kinematic indices, with interannual variability, is needed to conclude if the findings correspond to a more general pattern and not just for 2019.
Despite the limitations, it is important also to remark that this is the first time that convective mode–based analysis is presented for the East African region, showing preconvective indices for different convective modes and for possible severe weather. Global models (e.g., ECMWF, GFS) are widely used in East African forecasting offices, and although reanalysis data are not produced in real time for full forecasting guidance, the results presented here can be extrapolated to other global model outputs in the operational arena. Furthermore, the results reflect the need for increased observations in the region. Not only are more soundings needed but also surface stations, with collocated instrumentation that would help improve the monitoring and forecasting capacity in the basin. This will also improve regional and global models, as more data from the region will be assimilated, therefore improving the forecast, not only locally but also globally.
The tropical Africa (TA4) convective-permitting version of the Met Office’s Unified Model (UM) regional model (Hanley et al. 2021) is also used in the forecasting offices in East Africa and would have been ideal for our research. Unfortunately, we only had access to four pressure levels of information for the HIGHWAY 2019 field data collection, and it was impossible to retrieve vertical profiles. Current research is being conducted to link convective environments and modes with storm-scale severe weather indicators for three cases for which we have all pressure levels. This is needed to validate our results and to obtain a more accurate representation of the impact of storm-scale and local dynamics of the complex storm evolution and weather hazards over the lake and how this might change in the future. However, more stations covering the entire lake shore and buoys over the lake would be necessary to fully observe the surface wind variability, especially at the storm scale, and detect increasing wave heights. Furthermore, and as already discussed in Part I, the lack of a severe weather database in the region does not allow for a comparison of surface weather effects with radar imagery and preconvective environments, although the continuous reports in the media demonstrate that this region is prone to severe and fatal weather events, which could be enhanced in future climate scenarios (Van de Walle et al. 2021).
Acknowledgments.
This material is based upon work supported by the NSF National Center for Atmospheric Research, which is a major facility sponsored by the National Science Foundation under Cooperative Agreement 1852977. We are grateful to three anonymous reviewers whose comments greatly improved an earlier version of this manuscript.
Data availability statement.
The Nairobi sounding data from the HIGHWAY field project (Kenya Meteorological Department 2020) are accessible upon request from the NSF NCAR Earth Observing Laboratory (EOL) Data Archive (https://data.eol.ucar.edu/). The ERA5 reanalysis data are publicly available at the Copernicus Climate Data Store (CDS) (https://doi.org/10.24381/cds.adbb2d47).
Footnotes
For surface: T and Td are retrieved at 2 m AGL, wspd and wdir, at 10 m AGL, and p follows the surface topography.
East African time (EAT) = coordinated universal time (UTC) + 3 h.
REFERENCES
Anyah, R. O., F. H. M. Semazzi, and L. Xie, 2006: Simulated physical mechanisms associated with climate variability over Lake Victoria basin in East Africa. Mon. Wea. Rev., 134, 3588–3609, https://doi.org/10.1175/MWR3266.1.
Ba, M. B., and S. E. Nicholson, 1998: Analysis of convective activity and its relationship to the rainfall over the Rift Valley lakes of East Africa during 1983–90 using the Meteosat infrared channel. J. Appl. Meteor., 37, 1250–1264, https://doi.org/10.1175/1520-0450(1998)037<1250:AOCAAI>2.0.CO;2.
Bashir, B. M., 2023: Two boat accidents on Lake Victoria: Search and rescue operation underway. Uganda Police Force, 5 July, https://www.upf.go.ug/two-boat-accidents-on-lake-victoria-search-and-rescue-operation-underway/.
British Broadcasting Corporation, 2018: Uganda party boat capsizes on Lake Victoria, killing 29. BBC, 25 November, https://www.bbc.com/news/world-africa-46334275.
Bedka, K., J. Brunner, R. Dworak, W. Feltz, J. Otkin, and T. Greenwald, 2010: Objective satellite-based detection of overshooting tops using infrared window channel brightness temperature gradients. J. Appl. Meteor. Climatol., 49, 181–202, https://doi.org/10.1175/2009JAMC2286.1.
Black, E., J. Slingo, and K. R. Sperber, 2003: An observational study of the relationship between excessively strong short rains in coastal East Africa and Indian Ocean SST. Mon. Wea. Rev., 131, 74–94, https://doi.org/10.1175/1520-0493(2003)131<0074:AOSOTR>2.0.CO;2.
Blumberg, W. G., K. T. Halbert, T. A. Supinie, P. T. Marsh, R. L. Thompson, and J. A. Hart, 2017: SHARPpy: An open-source sounding analysis toolkit for the atmospheric sciences. Bull. Amer. Meteor. Soc., 98, 1625–1636, https://doi.org/10.1175/BAMS-D-15-00309.1.
Brooks, H. E., 2009: Proximity soundings for severe convection for Europe and the United States from reanalysis data. Atmos. Res., 93, 546–553, https://doi.org/10.1016/j.atmosres.2008.10.005.
Bunkers, M. J., B. A. Klimowski, J. W. Zeitler, R. L. Thompson, and M. L. Weisman, 2000: Predicting supercell motion using a new hodograph technique. Wea. Forecasting, 15, 61–79, https://doi.org/10.1175/1520-0434(2000)015<0061:PSMUAN>2.0.CO;2.
Cannon, T., 2014: World disasters report 2014: Focus on culture and risk. International Federation of Red Cross and Red Crescent Societies Rep., 276 pp., https://www.ifrc.org/sites/default/files/WDR-2014.pdf.
Chen, J., A. Dai, Y. Zhang, and K. L. Rasmussen, 2020: Changes in convective available potential energy and convective inhibition under global warming. J. Climate, 33, 2025–2050, https://doi.org/10.1175/JCLI-D-19-0461.1.
Coffer, B. E., M. D. Parker, R. L. Thompson, B. T. Smith, and R. E. Jewell, 2019: Using near-ground storm relative helicity in supercell tornado forecasting. Wea. Forecasting, 34, 1417–1435, https://doi.org/10.1175/WAF-D-19-0115.1.
Coniglio, M. C., H. E. Brooks, S. J. Weiss, and S. F. Corfidi, 2007: Forecasting the maintenance of quasi-linear mesoscale convective systems. Wea. Forecasting, 22, 556–570, https://doi.org/10.1175/WAF1006.1.
Craven, J. P., and H. E. Brooks, 2004: Baseline climatology of sounding derived parameters associated with deep moist convection. Natl. Wea. Dig., 28, 13–24.
del Moral Méndez, A., T. M. Weckwerth, R. D. Roberts, and J. W. Wilson, 2023: Toward improved short-term forecasting for Lake Victoria Basin. Part I: A radar-based convective mode analysis. Wea. Forecasting, 38, 2509–2526, https://doi.org/10.1175/WAF-D-23-0039.1.
Dixon, M., and G. Wiener, 1993: TITAN: Thunderstorm Identification, Tracking, Analysis, and Nowcasting—A radar-based methodology. J. Atmos. Oceanic Technol., 10, 785–797, https://doi.org/10.1175/1520-0426(1993)010<0785:TTITAA>2.0.CO;2.
Doswell, C. A., III, H. E. Brooks, and R. A. Maddox, 1996: Flash flood forecasting: An ingredients-based methodology. Wea. Forecasting, 11, 560–581, https://doi.org/10.1175/1520-0434(1996)011<0560:FFFAIB>2.0.CO;2.
Dune, D., 2022: Analysis: Africa’s unreported extreme weather in 2022 and climate change. CarbonBrief, 26 October, https://www.carbonbrief.org/analysis-africas-unreported-extreme-weather-in-2022-and-climate-change/.
Dutra, E., V. M. Stepanenko, G. Balsamo, P. Viterbo, P. M. A. Miranda, D. Mironov, and C. Schar, 2010: An offline study of the impact of lakes on the performance of the ECMWF surface scheme. Boreal Environ. Res., 15, 100–112.
Dyer, E., and R. Washington, 2021: Kenyan long rains: A subseasonal approach to process-based diagnostics. J. Climate, 34, 1–48, https://doi.org/10.1175/JCLI-D-19-0914.1.
Evans, J. S., and C. A. Doswell III, 2001: Examination of derecho environments using proximity soundings. Wea. Forecasting, 16, 329–342, https://doi.org/10.1175/1520-0434(2001)016<0329:EODEUP>2.0.CO;2.
Finney, D. L., J. H. Marsham, D. P. Rowell, E. J. Kendon, S. O. Tucker, R. A. Stratton, and L. S. Jackson, 2020a: Effects of explicit convection on future projections of mesoscale circulations, rainfall, and rainfall extremes over eastern Africa. J. Climate, 33, 2701–2718, https://doi.org/10.1175/JCLI-D-19-0328.1.
Finney, D. L., and Coauthors, 2020b: African lightning and its relation to rainfall and climate change in a convection-permitting model. Geophys. Res. Lett., 47, e2020GL088163, https://doi.org/10.1029/2020GL088163.
Finney, D. L., J. H. Marsham, D. P. Walker, C. E. Birch, B. J. Woodhams, L. S. Jackson, and S. Hardy, 2020c: The effect of westerlies on East African rainfall and the associated role of tropical cyclones and the Madden–Julian Oscillation. Quart. J. Roy. Meteor. Soc., 146, 647–664, https://doi.org/10.1002/qj.3698.
FloodList News in Africa, 2019: Uganda—Deadly storm and floods in eastern region. FloodList, 24 April, https://floodlist.com/africa/uganda-floods-eastern-region-april-2019.
Gensini, V. A., T. L. Mote, and H. E. Brooks, 2014: Severe-thunderstorm reanalysis environments and collocated radiosonde observations. J. Appl. Meteor. Climatol., 53, 742–751, https://doi.org/10.1175/JAMC-D-13-0263.1.
Haiden, T., M. Janousek, F. Vitart, Z. B. Bouallègue, L. Ferranti, F. Prates, and D. Richardson, 2021: Evaluation of ECMWF forecasts, including the 2021 upgrade. ECMWF Tech. Memo. 884, 56 pp., https://doi.org/10.21957/90pgicjk4.
Hanley, K. E., J. S. R. Pirret, C. L. Bain, A. J. Hartley, H. W. Lean, S. Webster, and B. J. Woodhams, 2021: Assessment of convection‐permitting versions of the Unified Model over the Lake Victoria basin region. Quart. J. Roy. Meteor. Soc., 147, 1642–1660, https://doi.org/10.1002/qj.3988.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hitchens, N. M., and H. E. Brooks, 2014: Evaluation of the Storm Prediction Center’s convective outlooks from day 3 through day 1. Wea. Forecasting, 29, 1134–1142, https://doi.org/10.1175/WAF-D-13-00132.1.
Indeje, M., F. H. M. Semazzi, and L. Ogallo, 2000: ENSO signals in East African rainfall seasons. Int. J. Climatol., 20, 19–46 https://doi.org/10.1002/(SICI)1097-0088(200001)20:1<19::AID-JOC449>3.0.CO;2-0.
Johns, R. H., J. M. Davies, and P. W. Leftwich, 1993: Some wind and instability parameters associated with strong and violent tornadoes: 2. Variations in the combinations of wind and instability parameters. The Tornado: Its Structure, Dynamics, Prediction, and Hazards, Geophys. Monogr., Vol. 79, Amer. Geophys. Union, 583–583, https://doi.org/10.1029/GM079p0583.
Kayiranga, T., 1991: Observation of convective activity from satellite data over the Lake Victoria region in April 1985 (in French). Veille Climatique Satellitaire, 37, 44–55.
Kebacho, L. L., 2022: The role of tropical cyclones Idai and Kenneth in modulating rainfall performance of 2019 long rains over East Africa. Pure Appl. Geophys., 179, 1387–1401, https://doi.org/10.1007/s00024-022-02993-2.
Kilavi, M., and, Coauthors, 2018: Extreme rainfall and flooding over central Kenya including Nairobi city during the long-rains season 2018: Causes, predictability, and potential for early warning and actions. Atmosphere, 9, 472, https://doi.org/10.3390/atmos9120472.
Kiwanuka-Tondo, J., S. Fredrick, and K. Pettiway, 2019: Climate risk communication of navigation safety and climate conditions over Lake Victoria basin: Exploring perceptions and knowledge of indigenous communities. Cogent Soc. Sci., 5, 1588485, https://doi.org/10.1080/23311886.2019.1588485.
KMD, 2020: Nairobi, Kenya high resolution BUFR radiosonde data. Version 1.0. UCAR/NCAR—Earth Observing Laboratory, accessed 13 December 2023, https://doi.org/10.26023/3NMW-GCXD-600M.
Kobusingye, O., 2020: Understanding and preventing drowning in Uganda: Final dissemination report for stakeholders. Tech. Rep., 68 pp., https://news.mak.ac.ug/wp-content/uploads/2021/12/MakSPH-Understanding-and-Preventing-Drowning-in-Uganda-2020-Report.pdf.
Kombe, C., 2022: 19 Dead After Plane Crashes in Lake Victoria in Tanzania. Voice of America (VOA) News, 6 November, https://www.voanews.com/a/passenger-plane-plunges-into-lake-victoria-in-tanzania-/6822341.html.
Lumb, F. E., 1970: Topographic influences on thunderstorm activity near Lake Victoria. Weather, 25, 404–410, https://doi.org/10.1002/j.1477-8696.1970.tb04129.x.
MacLeod, D., R. Graham, C. O’Reilly, G. Otieno, and M. Todd, 2021: Causal pathways linking different flavours of ENSO with the Greater Horn of Africa short rains. Atmos. Sci. Lett., 22, e1015, https://doi.org/10.1002/asl.1015.
Markowski, P., and Y. Richardson, 2010: Mesoscale Meteorology in Midlatitudes. Wiley-Blackwell, 430 pp., https://doi.org/10.1002/9780470682104.
Minallah, S., and A. L. Steiner, 2021: The effects of lake representation on the regional hydroclimate in the ECMWF reanalyses. Mon. Wea. Rev., 149, 1747–1766, https://doi.org/10.1175/MWR-D-20-0421.1.
Nicholson, S. E., A. T., Hartman, and D. A. Klotter, 2021a: On the diurnal cycle of rainfall and convection over Lake Victoria and its catchment. Part I: Rainfall and mesoscale convective systems. J. Hydrometeor., 22, 3037–3047, https://doi.org/10.1175/JHM-D-21-0083.1.
Nicholson, S. E., A. T., Hartman, and D. A. Klotter, 2021b: On the diurnal cycle of rainfall and convection over Lake Victoria and its catchment. Part II: Meteorological factors in the diurnal and seasonal cycles. J. Hydrometeor., 22, 3049–3064, https://doi.org/10.1175/JHM-D-21-0085.1.
Nicholson, S. E., A. H. Fink, C. Funk, D. A. Klotter, and A. Rasheeda Satheesh, 2022: Meteorological causes of the catastrophic rains of October/November 2019 in equatorial Africa. Global Planet. Change, 208, 103687, https://doi.org/10.1016/j.gloplacha.2021.103687.
Odongo, R. I., B. A. Ogwang, J. Kisembe, and H. N. Ngoma, 2022: An observational study of Lake Breeze over the Victoria basin in Uganda. North Amer. Acad. Res., 5, 109–123, https://doi.org/10.5281/zenodo.7158583.
Ogega, O. M., E. Scoccimarro, H. Misiani, and J. Mbugua, 2023: Extreme climatic events to intensify over the Lake Victoria Basin under global warming. Sci. Rep., 13, 9729 https://doi.org/10.1038/s41598-023-36756-3.
Palmer, P. I., and Coauthors, 2023: Drivers and impacts of Eastern African rainfall variability. Nat. Rev. Earth Environ., 4, 254–270, https://doi.org/10.1038/s43017-023-00397-x.
Prein, A. F., and G. J. Holland, 2018: Global estimates of damaging hail hazard. Wea. Climate Extremes, 22, 10–23, https://doi.org/10.1016/j.wace.2018.10.004.
Púčik, T., P. Groenemeijer, D. Rýva, and M. Kolář, 2015: Proximity soundings of severe and nonsevere thunderstorms in central Europe. Mon. Wea. Rev., 143, 4805–4821, https://doi.org/10.1175/MWR-D-15-0104.1.
Rasmussen, E. N., and D. O. Blanchard, 1998: A baseline climatology of sounding-derived supercell and tornado forecast parameters. Wea. Forecasting, 13, 1148–1164, https://doi.org/10.1175/1520-0434(1998)013<1148:ABCOSD>2.0.CO;2.
Roberts, R. D., and Coauthors, 2022: Taking the HIGHWAY to save lives on Lake Victoria. Bull. Amer. Meteor. Soc., 103, E485–E510, https://doi.org/10.1175/BAMS-D-20-0290.1.
Rodríguez, O., and J. Bech, 2018: Sounding-derived parameters associated with tornadic storms in Catalonia. Int. J. Climatol., 38, 2400–2414, https://doi.org/10.1002/joc.5343.
Shaaban, A. A., and P. E. Roundy, 2017: OLR perspective on the Indian Ocean dipole with application to East African precipitation. Quart. J. Roy. Meteor. Soc., 143, 1828–1843, https://doi.org/10.1002/qj.3045.
Shapiro, S. S., and M. B. Wilk, 1965: An analysis of variance test for normality (complete samples). Biometrika, 52, 591–611, https://doi.org/10.1093/biomet/52.3-4.591.
Spearman, C., 1904: The proof and measurement of association between two things. Amer. J. Psychol., 15, 72–101, https://doi.org/10.2307/1412159.
Taszarek, M., H. E. Brooks, and B. Czernecki, 2017: Sounding-derived parameters associated with convective hazards in Europe. Mon. Wea. Rev., 145, 1511–1528, https://doi.org/10.1175/MWR-D-16-0384.1.
Taszarek, M., J. T. Allen, T. Púčik, K. A. Hoogewind, and H. E. Brooks, 2020: Severe convective storms across Europe and the United States. Part II: ERA5 environments associated with lightning, large hail, severe wind, and tornadoes. J. Climate, 33, 10 263–10 286, https://doi.org/10.1175/JCLI-D-20-0346.1.
Thiery, W., E. L. Davin, S. I. Seneviratne, K. Bedka, S. Lhermitte, and N. P. M. van Lipzing, 2016: Hazardous thunderstorm intensification over Lake Victoria. Nat. Commun., 7, 12786, https://doi.org/10.1038/ncomms12786.
Thiery, W., L. Gudmundsson, K. Bedka, F. H. M. Semazzi, S. Lhermitte, P. Willems, N. P. M. van Lipzig, and S. I. Seneviratne, 2017: Early warnings of hazardous thunderstorms over Lake Victoria. Environ. Res. Lett., 12, 074012, https://doi.org/10.1088/1748-9326/aa7521.
Thompson, R. L., R. Edwards, J. A. Hart, K. L. Elmore, and P. Markowski, 2003: Close proximity soundings within supercell environments obtained from the Rapid Update Cycle. Wea. Forecasting, 18, 1243–1261, https://doi.org/10.1175/1520-0434(2003)018<1243:CPSWSE>2.0.CO;2.
Tyler, M. D., D. B. Richards, C. Reske-Nielsen, O. Saghafi, E. A. Morse, R. Carey, and G. A. Jacquet, 2017: The epidemiology of drowning in low- and middle-income countries: A systematic review. BMC Public Health, 17, 413–417, https://doi.org/10.1186/s12889-017-4239-2.
Van de Walle, J., W. Thiery, O. Brousse, N. Souverijns, M. Demuzere, and N. P. M. van Lipzig, 2020: A convection-permitting model for the Lake Victoria Basin: Evaluation and insight into the mesoscale versus synoptic atmospheric dynamics. Climate Dyn., 54, 1779–1799, https://doi.org/10.1007/s00382-019-05088-2.
Van de Walle, J., W. Thiery, R. Brogli, O. Martius, J. Zscheischler, and N. P. M. van Lipzig, 2021: Future intensification of precipitation and wind gust associated thunderstorms over Lake Victoria. Wea. Climate Extremes, 34, 100391, https://doi.org/10.1016/j.wace.2021.100391.
Virts, K. S., and S. J. Goodman, 2020: Prolific lightning and thunderstorm initiation over the Lake Victoria basin in East Africa. Mon. Wea. Rev., 148, 1971–1985, https://doi.org/10.1175/MWR-D-19-0260.1.
Wainwright, C. M., D. L. Finney, M. Kilavi, E. Black, and J. H. Marsham, 2021: Extreme rainfall in East Africa, October 2019–January 2020 and context under future climate change. Weather, 76, 26–31, https://doi.org/10.1002/wea.3824.
Waniha, P. F., R. D. Roberts, J. W. Wilson, A. Kijazi, and B. Katole, 2019: Dual-polarization radar observations of deep convection over Lake Victoria Basin in East Africa. Atmosphere, 10, 706, https://doi.org/10.3390/atmos10110706.
Watkiss, P., R. Powell, A. Hunt, and F. Cimato, 2020: The socio-economic benefits of the HIGHWAY project. Tech. Rep., 89 pp., https://www.metoffice.gov.uk/binaries/content/assets/metofficegovuk/pdf/business/international/wiser/wiser0274_highway_seb_report.pdf.
Weisman, M. L., and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110, 504–520, https://doi.org/10.1175/1520-0493(1982)110<0504:TDONSC>2.0.CO;2.
Whitworth, H. S., and Coauthors, 2019: Drowning among fishing communities on the Tanzanian shore of Lake Victoria: A mixed-methods study to examine incidence, risk factors and socioeconomic impact. BMJ Open, 9, e032428, https://doi.org/10.1136/bmjopen-2019-032428.
Williams, K., J. Chamberlain, C. Buontempo, and C. Bain, 2015: Regional climate model performance in the Lake Victoria Basin. Climate Dyn., 44, 1699–1713, https://doi.org/10.1007/s00382-014-2201-x.
Wilson, J., and R. D. Roberts, 2022: Lake Victoria thunderstorms: Radar-observed initiation and storm evolution modes. Mon. Wea. Rev., 150, 2485–2502, https://doi.org/10.1175/MWR-D-21-0283.1.
Woodhams, B. J., C. E. Birch, J. H. Marsham, T. P. Lane, C. L. Bain, and S. Webster, 2019: Identifying key controls on storm formation over the Lake Victoria basin. Mon. Wea. Rev., 147, 3365–3390, https://doi.org/10.1175/MWR-D-19-0069.1.
Woodhams, B. J., and Coauthors, 2022: Aircraft observations and sub-km modelling of the lake–land breeze circulation over Lake Victoria. Quart. J. Roy. Meteor. Soc., 148, 557–580, https://doi.org/10.1002/qj.4155.
World Health Organization, 2023: Preventing drowning in the fishing industry. Accessed 21 November 2023, https://www.who.int/news-room/feature-stories/detail/preventing-drowning-in-the-fishing-industry.
Yin, X., and S. E. Nicholson, 1998: The water balance of Lake Victoria. Hydrol. Sci. J., 43, 789–811, https://doi.org/10.1080/02626669809492173.
