Climate of the study area

The study was performed over West Africa, with a central focus on Ghana (Fig 1), where station data was retrieved for further assessment. West Africa is defined here as the region between 0 ^° N and 20 ^° N, 20 ^° W and 20 ^° E and designated into the various climatic zones: Guinea Coast, Sudano, Sahel and Sahara [39]. Ghana is located within 2 ^° W, 4 ^° N and 2 ^° E, 12 ^° N, and have topography consisting of highlands, mountain ranges and coastal plains. The land cover is dominated by forest in southwestern and part of the central portion and grassland in the North. There is also a network of river systems originating from the highlands such as the Volta River, which influence the climate system of the country. The climate is tropical, characterized by the wet and dry season, with rainfall patterns, mainly unimodal in the North, and bi-modal (separated by a short dry spell in August) in the south. These patterns are associated with mesoscale convective systems and controlled by the advection of moisture from the Gulf of Guinea in the low-level atmosphere [40].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of Study Area.

Map of Ghana, West Africa (green shaded region in West Africa) with the rain gauge network shown as circles on the right panel. The study region is demarcated into four agro-ecological zones as per Ghana Meteorological Agency (GMet) standards [41]. The four study locations (Bole, Kumasi, Accra, and Axim) for thunderstorm case assessments are represented by green circles on the right panel.

https://doi.org/10.1371/journal.pclm.0000703.g001

Description of datasets

Thunderstorm data, categorized into rain-associated (TS) and non-rain-associated or dry thunderstorms (Dry TS), were extracted from daily weather registers (METAR form) obtained from the Ghana Meteorological Agency (GMet). Monthly TS data covering a 19-year period (1988–2006) were used for four synoptic stations namely; Bole, Accra, Kumasi, and Axim. For Accra, an additional five years (1961–1965) of data were included. Dry TS observations were available from 1999 to 2006 for all stations except Axim. The selected meteorological stations were strategically chosen to represent Ghana’s diverse agro-climatic zones. Specifically, Bole represents the Savannah zone, Kumasi corresponds to the Forest zone, and Accra and Axim represent the Coastal zone. These selections ensure that the analysis captures the spatial variability of thunderstorm activity across the country’s major ecological regions. Data for the Transition zone were not available.

To assess rainfall characteristics and their relationship with thunderstorm occurrence, the study utilized a 23-year monthly gridded rainfall dataset (GMet v1.0) spanning 1990 to 2012, developed by the Ghana Meteorological Agency [42]. This dataset, generated through spatial interpolation of gauge observations, provides a continuous rainfall field at regional scale. It is important to distinguish this rainfall dataset from the thunderstorm occurrence data, which were derived directly from METAR records at specific synoptic stations. While the thunderstorm data capture the presence or absence of storm events (with or without rain), the gridded rainfall data quantify actual precipitation amounts. In subsequent analysis, rainfall was compared with satellite-derived lightning activity through computation of flash rate density (FRD), a measure of lightning flashes per unit area per year (km⁻² yr⁻¹). FRD is introduced here as it forms a key metric in linking lightning climatology with rainfall patterns.

For lightning assessment, the study employed the 17-year TRMM-LIS Very High-Resolution Monthly and Seasonal Climatology (VHRMC, VHRSC) dataset covering 1998–2013, obtained from NASA’s Global Hydrology Center over the West African region, with primal focus on Ghana (https://ghrc.nsstc.nasa.gov/hydro). The datasets subjected to both 49-day and 1-degree boxcar moving average to remove diurnal cycle and smooth regions with low flash rate for robust results, was obtained at 0.1-degree spatial resolution. Details on the lightning sensor and datasets can be found in [43], [44] and [45].

Atmospheric variables relevant to thunderstorm development were obtained from the ERA5 reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), under the Copernicus Climate Change Service (C3S). ERA5 provides a high-resolution dataset at horizontal resolution and 137 vertical levels. The study utilized 17-year monthly means of Convective Available Potential Energy (CAPE), daily relative humidity at 200, 500, and 925 hPa, 2-meter air temperature, and convective precipitation data spanning 1998–2013. In addition to ERA5, monthly mean values of the Lifted Index (LI) were obtained from the National Centers for Environmental Prediction (NCEP) Reanalysis dataset [46], also spanning 1998 to 2013 at resolution. The LI is a key parameter used to assess atmospheric stability and was used alongside CAPE and relative humidity in the evaluation of convective potential.

Methodology

To ensure spatial consistency, all datasets were interpolated to a common (latitude-longitude grid) resolution.For comparison with gridded rainfall data (GMet v1.0), lightning flash rate densities (FRD) were further aggregated to (latitude-longitude grid). The reanalysis datasets (ERA5 and NCEP) and TRMM-LIS were analyzed separately and evaluated against station-based TS and Dry TS observations. No bias corrections were made to the datasets. A 16-year (1998 – 2012) temporal range of analysis is selected for the TRMM-LIS, ERA-5, and NCEP Reanalysis datasets.The monthly GMet v1.0 gridded rainfall data [42] was also further averaged to obtain monthly and seasonal climatologies for 16-years (1998-2012). These rainfall characteristics were then compared with lightning occurrence patterns to explore potential relationships.

Prior to analysis, the station-based thunderstorm data underwent quality control to address missing values and inconsistencies. The cleaned dataset was then used to compute monthly means, annual medians, and total monthly frequencies for each station. This enabled the identification of peak thunderstorm months and characterization of seasonal cycles at each location. Contour plots were then generated to visualize the temporal (monthly, seasonal, and annual) distributions of thunderstorms over the 19-year (or longer) record.Spatio-temporal distributions of thunderstorm frequencies distinguishing between general thunderstorms (TS) and dry thunderstorms (Dry TS) were assessed at four selected stations: Accra and Axim (Coastal zone), Kumasi (Forest/Transition zone), and Bole (Savannah zone), representing Ghana’s main agro-climatic zones [47]. This ensured the analysis captured the full range of the country’s climatic variability. Inter-comparisons of TS and Dry TS frequencies and trends across stations were also conducted. Importantly, thunderstorm frequency data from ground stations were analyzed independently of satellite and reanalysis datasets to preserve observational integrity.

The second part of the analysis focused on flash rate density (FRD) from Very High Resolution Cloud (VHRSC) and Cloud-to-Ground (VHRMC) lightning datasets. FRD was derived by scaling flash counts by the view time within each grid cell, yielding values in flashes km⁻² yr⁻¹ [48]. These were used to produce both monthly and seasonal climatological maps. Seasonal averages were also computed for atmospheric instability and moisture variables: Convective Available Potential Energy (CAPE), Lifted Index (LI), and relative humidity (RH) at 200 hPa, 500 hPa, and 925 hPa. To characterize the atmospheric instability, CAPE and LI were adopted [49]. For CAPE, values indicate moderate instability conditions over 1000 JKg⁻¹ , very unstable over 2500 JKg⁻¹ and extremely unstable above 3500 JKg⁻¹ [50]. CAPE values less than 300 JKg⁻¹ show no or low probability of convection [51]. On the other hand, LI values between C and C represent moderately unstable conditions and, C to -9 ^°C indicate extreme instability [52]. LI above zero show low or no convective probability. The r-values and p-values between rainfall (GMet v1.0) and lightning flash rate density over the various agro-climatic zones of Ghana (Coastal, Forest, Transition and Savannah zones) were computed as assessment for the relationship between the two parameters. Similar statistical assessments were done on the stability indices and RH.

To explore the interactions between thunderstorms and atmospheric conditions across West Africa, the study employed hexbin density plots generated with Python’s matplotlib library [53]. These plots provide a compact and intuitive way to visualize multi-variable relationships by encoding data density and magnitude in color intensity. The hexbin approach was used to investigate patterns at both zonal and seasonal scales, with thunderstorm frequency (TS) used as the color-weighted variable to highlight storm activity. The analysis focused on three thematic areas: the zonal relationship between convective rainfall, near-surface air temperature, and a third variable (TS); the interaction between CAPE, relative humidity (RH), and TS; and the seasonal behavior of CAPE, convective inhibition (CIN), and TS. For the zonal analysis, data were examined across four climate zones—Guinea Coast, Sudanian, Sahel, and Sahara in the broader West African region with rainfall and temperature plotted along the axes, and color intensity representing the summed TS values to capture both distribution density and magnitude. The vertical-level analysis considered CAPE and RH at 200 hPa, 500 hPa, and 925 hPa across the same zones, again incorporating TS as a color-weighted variable. Seasonal variability was addressed by producing hexbin plots for the four climatological seasons (DJF, MAM, JJA, SON), emphasizing shifts in atmospheric stability through CAPE, CIN, and TS interrelationships (with TS as a color-weighted variable).

Distribution of lightning and instability

To start, we investigated the distribution of CAPE (Fig 2a) in the area of study. It was observed that nonzero CAPE is the exception that proves the rule: of all 960000 data points fewer than 105306 (10.96) had CAPE = 0 J kg⁻¹. The distribution of the number cases as a function of CAPE showed a strong decrease (Fig 2) for increasing CAPE values, indicating the rarity of high CAPE values even in the tropics.The joint distribution of CAPE and FRD (Fig 2) shows that the atmosphere is usually in a state of low CAPE. In addition, High values of FRD is something of a rarity in the tropics. However, maximum intensity of FRD is often associated with high CAPE.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Frequency distribution OF CAPE AND LIS.

Mean monthly climatology distribution of CAPE (a) and Lightning flash rate density over West Africa (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g002

In DJF (Fig 3), significant lightning occurrence is observed at about 400 J kg⁻¹ CAPE with peak LIS value of 0.21. Lightning occurrence persists with increasing CAPE until 1950 J kg⁻¹. MAM recorded lightning activity at 100 J kg⁻¹, with significant lightning activity between 100 J kg⁻¹ to 1250 J kg⁻¹ and peak occurrence at 0.3. In JJA, it is observed that lightning occurrence is concentrated between 100 J kg⁻¹ and 1500 J kg⁻¹. JJA also record the highest CAPE values associated with LIS occurrence. SON revealed lightning occurrence concentrated between 250 J kg⁻¹ and 1000 J kg⁻¹. In general, it is observed that the frequency of intense lightning occurrence is rare and often associated with CAPE >500 J kg⁻¹. The exception to this phenomenon is MAM where peak lightning intensity of 0.3 was observed with accompanying CAPE<500 J kg⁻¹.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Seasonal distribution of CAPE and LIS.

A seasonal distribution plot CAPE and Lightning flash rate density (FRD). Note that the color bar used in these plots are for density distribution for both variables over West Africa (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g003

In DJF (Fig 4), most of the lightning frequencies were concentrated between 500 J kg⁻¹ — 2000 J kg⁻¹ and -1 — -2.9 LI, with a peak frequency occurring between -2 — -2.9 LI and CAPE > 1500 J kg⁻¹. Furthermore, the high frequency of lightning was accompanied by high values of CAPE and LI (> -1). In MAM the environment of lightning occurrence was recorded between -1 — -4 LI and CAPE > 500 J kg⁻¹. However, a significant lightning frequency was also observed between 0 — -2 LI and sufficient CAPE (< 500 J kg⁻¹). In JJA, the frequency of lightning was observed to follow an almost linear path, increasing with increasing CAPE and LI values. Peak lightning frequency was observed around -3 LI and 1900 J kg⁻¹ CAPE. In SON, the lightning frequency was mostly concentrated between -2 — -3.2 LI and CAPE > 1000 J kg⁻¹.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Seasonal hexagonal bin plot of CAPE and Lifted index to LIS occurrence.

A seasonal hexagonal bin plot of CAPE and Lifted index with respect to lightning flash rate density over West Africa (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g004

In DJF (Fig 5) at 925 hPa, lightning occurrence increases with RH. It is observed that sufficient environment for significant lightning occurrence is between 75–100% RH and CAPE >500 J kg⁻¹ & < 1200 J kg⁻¹. At 500 h Pa, a minimum RH of >30% and CAPE between 500-1400Jkg-1 was observed to be sufficient for lightning occurrence. At 200hPa, significant LIS occurrence was observed above 50% RH and CAPE between about 300 J kg⁻¹ to 1200 J kg⁻¹. Peak frequency was recorded at about 55% RH associated with CAPE between 500 J kg⁻¹ and 800 J kg⁻¹. It is worth noting that, all peak occurrence of FRD for all three levels were recorded between 500 J kg⁻¹ and 800 J kg⁻¹.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Seasonal hexagonal bin plot of CAPE and relative humidity.

A seasonal hexagonal bin plot of CAPE and Relative humidity (at 200 hPa, 500 hPa and 925 hPa) with respect to Lightning flash rate density over West Africa (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g005

In MAM (Fig 5) at 925 h Pa lightning occurrence increases with increasing RH. The highest recorded frequency of occurrence was between 80%–100% RH and 0 J kg⁻¹ to 500 J kg⁻¹, however significant lightning occurrence is observed for a minimum RH of 20% with sufficient CAPE >500 J kg⁻¹. At 500 h Pa the environment of significant FRD occurrence was observed to be between 50%–65% RH and 0–500 J kg⁻¹. Lightning was also observed to occur with a minimum of  30% with sufficiently large CAPE (about 1000 J kg⁻¹). At 200 h Pa, significant FRD frequency environment is observed between 70% - 90% RH and CAPE between 0 J kg⁻¹–1200 J kg⁻¹.

In JJA (Fig 5), at 925 h Pa, significant FRD occurrence is observed at  30%–50% at CAPE 0 J kg⁻¹–100 J kg⁻¹. At 60%–90% RH, lightning occurrence increases with RH and CAPE up to 1000 J kg⁻¹ and then reduces with increasing CAPE. At 500 h Pa significant lightning occurrence is observed at 50%–60% RH with little to no CAPE (0 J kg⁻¹ – 100 J kg⁻¹). The environment of concentrated FRD occurrence is observed between  60%–78% RH and 0 J kg⁻¹–1500 J kg⁻¹. Although there is some presence of LIS occurrence with increasing CAPE , these phenomena are not common. At 200 h Pa, the environment of concentrated FRD occurrence is between 68%–85% RH with associated CAPE values of 0 J kg⁻¹–1500 J kg⁻¹.

In SON (Fig 5) at 925 h Pa, the majority of FRD occurrence is observe between about 60%–100% RH and 400 J kg⁻¹ to 1500 J kg⁻¹. It is however worth noting that LIS occurrence was observed for a minimum of  20%–60% RH with sufficient CAPE (0 J kg⁻¹ –  1400 J kg⁻¹). At 500 h Pa, significant FRD occurrence was observed between 40%–70% and associated CAPE of 400 J kg⁻¹–1200 J kg⁻¹. Peak occurrence between 50%–60% and CAPE 400 J kg⁻¹–800 J kg⁻¹. At 200 h Pa, the environment of significant FRD occurrence was between 60%–80% with associated CAPE of 400 J kg⁻¹–1000 J kg⁻¹.

In DJF (Fig 6), the peak occurrence of LIS was observed between 500 J kg⁻¹—-1000 J kg⁻¹ and -50 J kg⁻¹ to -150 J kg⁻¹. In general, LIS occurrence was observed at environments -150 J kg⁻¹ and CAPE 1000 J kg⁻¹ (Higher CAPE is needed for peak concentration).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Occurrence of lightning flash rate density under CAPE and CIN (West Africa).

Hexbin plot of lightning flash rate density for the combination of CAPE and CIN over West Africa (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g006

In MAM (Fig 6), the environment of peak LIS occurrence was between 0 J kg⁻¹–500 J kg⁻¹ CAPE and -50 J kg⁻¹ to -150 J kg⁻¹. It is worth noting that LIS activity continues to occur with increasing CIN -150 J kg⁻¹ and CAPE 500 J kg⁻¹ (less CAPE values are required for the initiation of most LIS occurrence during the season).

In JJA (Fig 6), the environment of concentrated LIS occurrence is observed between -300 J kg⁻¹ and -500 J kg⁻¹ CIN and 0 J kg⁻¹ -800 J kg⁻¹. It is also observed that a majority of LIS occurrence are associated with large CAPE 500 J kg⁻¹ and -200 J kg⁻¹ CIN with little or no LIS occurrence associated with CIN below 100 J kg⁻¹. It is worth noting that, though LIS associated with extremely large CIN ( 600 J kg⁻¹) is rare. These LIS occurrences require very little to no CAPE.

In SON (Fig 6), peak LIS environment was observed between -50 J kg⁻¹ to -250 J kg⁻¹ CIN and 250 J kg⁻¹ to 800 J kg⁻¹ CAPE. Peak occurrence was recorded at around 500 J kg⁻¹ CAPE and -100 J kg⁻¹ CIN. The rest of LIS occurrence in the season is associated with either large CAPE 800 J kg⁻¹ to 1500 J kg⁻¹ or CIN 200 J kg⁻¹ -550 J kg⁻¹. It was observed that LIS activity requires a minimum 300 J kg⁻¹ CAPE and  -50 J kg⁻¹ to occur. However, this is not an absolute rule as some occurrence was observed at CIN -100 J kg⁻¹ with CAPE close to zero.

Zonal distribution of lightning and instability

In the Sahara (Fig 7), very little lightning occurrence is observed above 700 J kg⁻¹ regardless of LI value. Compared to the GC/SOUD & SAHEL, the Sahara recorded the lowest lightning frequency for the climatological period. It was also observed that very little CAPE is need for the initiation of lightning activity. The plot further revealed that the environment for lightning occurrence in the Sahara is often between 50 J kg⁻¹ to 600 J kg⁻¹ & 0 to -2.5 lifted index.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Lightning flash rate density under CAPE and LI (West African Zones).

Zonal Hexbin plot of Lightning flash rate density for the combination of CAPE and LI over West Africa (1998–2012). Note that the values for fL are directly proportional to the frequency of lightning.

https://doi.org/10.1371/journal.pclm.0000703.g007

The Sahel region showed a gradual increase in lightning occurrence with increasing CAPE and negative LI. Peak concentration of lightning occurrence was observed between 1000 J kg⁻¹ to 1200 J kg⁻¹ & an LI of about -3.5 to -4.5. A decline in lightning occurrence was observed at 1000 J kg⁻¹ CAPE and -1 LI, and gradually decrease with increasing CAPE and LI. The highest LI value observed for the Sahel was -5.5, associated with medium frequency of occurrence between 1200 J kg⁻¹ to 1600 J kg⁻¹ and very little to no lightning occurrence after 1600 J kg⁻¹.

Significant environment for lightning occurrence in the Soud region (Fig 7) begun at -1.5 LI and about 200 J kg⁻¹ CAPE. The occurrence of LIS in the region then gradually increased with increasing CAPE (up to 1400 J kg⁻¹) and LI (-1.5 to -4.5). The frequency of occurrence then decreased with increasing CAPE (> 1400 J kg⁻¹) and LI (> -4.5). It should be noted that the combination of large values of CAPE (> 1400 J kg⁻¹) and LI (>-4.5) is rarely associated with high lightning frequency (this is an indication that in the Soud region, severe convective storms associated with intense lightning phenomena is a rather rare occurrence)

In the GC (Fig 7), the plot revealed significant occurrence of FRD between -1 to -3.5 LI and 120 J kg⁻¹ to 1200 J kg⁻¹ CAPE. Very little to no lightning occurrence was record below -1 LI even with increasing large CAPE values. Furthermore, with CAPE values greater than 1400 J kg⁻¹, very little to no lightning occurrence is recorded even at increasing LI (up to -4.8).

In GC (Fig 8) at 925 h Pa, lightning activity is observed to be most frequent between 0 to 500 J kg⁻¹ CAPE and 85% to 100% RH. Frequency of occurrence begins to decrease with increasing CAPE between > 500 J kg⁻¹ to about 1200 J kg⁻¹ and RH > 70%. At 500 h Pa, Significant frequency of lightning occurs between 50%–75% RH and 0 J kg⁻¹–1200 J kg⁻¹ with peak frequency of occurrence recorded at about 70% RH and 400Jkg-1. Significant lightning environment at 200 h Pa is recorded between 65%–85% RH and 0 J kg⁻¹–1000 J kg⁻¹. The highest frequency of lightning was recorded between 75%–80% RH and about 300 J kg⁻¹ to 500 J kg⁻¹ CAPE.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Lightning flash rate density under RH and CAPE (West African Zones).

Zonal hexbin plot of Lightning flash rate density for the combination of RH and CAPE over West Africa (1998–2012). Note that the values for fL are directly proportional to the frequency of lightning.

https://doi.org/10.1371/journal.pclm.0000703.g008

In the Soud region (Fig 8), the environment of lightning occurrence follows a seemingly linear path as CAPE values increase. Significant frequency of lightning is observed between 500 J kg⁻¹ and 1000 J kg⁻¹. It is worth noting that some lightning activity was also observed at about 25% RH with CAPE values of about 200 J kg⁻¹. A significant frequency of lightning was also observed between 40% to about 92% RH and CAPE > 1000 J kg⁻¹. At 500 h Pa the environment of lightning activity was observed between 400 J kg⁻¹ to 1000 J kg⁻¹ and 40%–70%. Peak frequency of lightning was observed at about 45% and 700 J kg⁻¹. At 200 h Pa, significant frequency of lightning was recorded between 65%–75% and 500 J kg⁻¹–1000 J kg⁻¹ with peak values occurring around 67% RH and 700 J kg⁻¹–900 J kg⁻¹.

In the Sahel region (Fig 8), at 925 h Pa lighting frequency was observed to decrease with increasing CAPE. The environment of lightning occurrence was recorded between 0 J kg⁻¹–500 J kg⁻¹. It is worth noting that high lightning frequency was observed for RH as low as 10%–30% RH. At 500 h Pa, peak lightning environment was observed between 25%–45% RH and 0 J kg⁻¹–900 J kg⁻¹ CAPE. 200 h Pa recorded significant lighting activity at about 45% RH and peak frequency at around 50% RH. It is observed that the environment of lightning occurrence exists between 45%–75% RH and 0 J kg⁻¹–1000 J kg⁻¹

In the Sahara, significant lightning frequency is recorded between 10%–20% RH at 925 h Pa. At 500 h Pa, the lightning-occurrence environment is observed between 20%–35% RH. The lightning occurrence then shifts to about 37% to 50% RH at 200 h Pa. It is worth noting that all of the significant frequencies were accompanied by CAPE  100 J kg⁻¹. Furthermore, lightning frequencies although minimal were observed with increasing values of CAPE(> 100 J kg⁻¹) and RH (> 10%).

In GC (Fig 9), the frequency of lightning is mostly concentrated between 100 mm RR to 250 mm and 22.5 (^°C) to 30 (^°C). Temperatures below 22 (^°C) regardless of convective rainfall, were observed to yield little to no lightning frequency. Furthermore, Peak occurrence of lightning was recorded between 25 (^°C)–27 (^°C). Although temperatures above 30 (^°C) were accompanied by some lightning occurrence, these events were not associated with intense convective rainfall.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Lightning flash rate density under 2m temperature and convective rainfall (West African Zones).

Zonal hexbin plot of Lightning flash rate density for the combination of 2m temperature and convective rainfall over West Africa (1998–2012). Note that the values for fL are directly proportional to the frequency of lightning.

https://doi.org/10.1371/journal.pclm.0000703.g009

In the Soud region (Fig 9), the environment of high lightning frequency was observed between 23 (^°C)–30 (^°C) and 50 mm–250 mm. Peak frequencies of lightning were recorded between 26 (^°C)–27 (^°C) and 150 mm – 200 mm. Temperatures above 30 (^°C) also recorded slightly higher frequencies associated with minimal rainfall (0–100 mm). It is also worth noting that temperatures between 22.5 (^°C)–27 (^°C) were associated with large rainfall events accompanied by minimal frequency of lightning.

In the Sahel region (Fig 9), concentrated lightning activity was observed between 28 (^°C)–35 (^°C) and 0–130 mm. Peak occurrence was recorded at about 32.5 (^°C) and 20 mm of rain. It should be noted that little to no lightning frequency was observed below 25 (^°C). Furthermore, temperatures between 25 (^°C)–29 (^°C) were associated with intense rainfall and lightning frequency decreasing with increasing rainfall.

In the Sahara (Fig 9), most of the lightning frequency was concentrated between 30 (^°C) to 35 (^°C) and 0–15 mm. Temperatures below 27 C were associated with minimal rainfall and little to no lightning frequency. Peak frequency of lightning was observed at about 35 (^°C) and 0–15 mm of rain. Some lightning frequency was also recorded between 25 (^°C)–30 C and 20 mm–65 mm, however the frequency was very low.

Spatial and temporal variability of FRD, CAPE, RH and LI climatology

Lightning activities were observed to start around the coastal regions of West Africa as early as January (Fig 10a) and February (Fig 10 and 10b) with peak mean FRD between 0.10 - 0.12 fl km⁻²day⁻¹ and at this time inland areas experiencing almost no lightning activity. Fig 11a–11b captured moderate to low CAPE values across mostly the entire region, with peak monthly CAPE values observed mainly along the Guinean coasts. Moderate lightning activities observed within March, April and May (Fig 10c–10e) gradually moving northward toward the sudano regions with maximum peaks occurring over the south-eastern boarders of Nigeria. CAPE values began to rise and advance toward the northern parts of the study region (Sudano and lower Sahel) in March-April-May (Fig 11c–11e). By June (Fig 10 and 11f) moderate to high CAPE values seen across the entire Sudano, Sahel and Guinea coast with maximum peaks occurring in the Sudano-Sahel zones.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Mean monthly FRD Climatology over West Africa.

Mean monthly FRD Climatology over West Africa, January (a) to December (l) in progressive order (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g010

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Mean monthly CAPE Climatology over West Africa.

Mean monthly CAPE Climatology over West Africa (1998–2012), with similar description as Fig 10.

https://doi.org/10.1371/journal.pclm.0000703.g011

It is worth mentioning that the most prominent peaks in mean LIS FRD monthly climatology over West Africa during 1998-2013 occurred between the months of June and September (Fig 10f–10i). The coastal regions (Liberia, Cote d’Ivore, Ghana and some parts of Nigeria) recorded very low lightning activities in July and August (Fig 10g–10h). During July-August and September (Fig 11g–11i), a gradual decrease in CAPE values was observed along the Guinea coasts, with a reversal IN the Sudano-Sahel zones and peaks occurring mainly across the Sahel zone. The retreat of lightning activities toward the coastal regions of Southern West Africa in October, November and December (Fig 10j–10l) was accompanied by a sharp maximum of mean FRD between 0.12–0.14 fl km⁻²day⁻¹ along the coastal lines with little to no lightning events occurring in areas further inland. A similar southward retreat of convective activity is seen in CAPE (Fig 11j–11l) with maximum CAPE values recorded along southern-most portions of the Guinea Coast. During this period, moderate to low CAPE values in the Sudano region and almost zero CAPE in the Sahel zone were observed. It is interesting to note that relatively high convective activities was observed along the Volta river with accompanying lightning activity in all months.

The seasonal peak of relative humidity (Figs 10 and 13) was observed mainly in the lower tropospheric region (925 hPa) and decreasing with increasing altitude (500 hPa to 200 hPa). In December-February season, high RH was observed mostly along the coastal regions in the lower troposphere. Moderate to low RH were observed to spread across the entire West African region at 500 hPa and 200 hPa, with the latter being the secondary maxima. In Figs 10 and 12, moderate lightning activities were observed over the landmasses near the coastal regions with peak occurrence mainly along the coasts of Liberia, Cote d’Ivore and Ghana (Guinea Coast). DJF in Figs 10 and 14 also recorded high LI values along the Guinea coast zone, with a gradual northward decrease in intensity. Active lightning occurrences and increasing RH were recorded at all levels along the coastal landmasses (Guinea coast zone) advancing northward toward the Sudano-Sahel zones in MAM (Figs 12 and 13). Besides, significantly high LI values observed in the the Sudano, Guinea coasts during the MAM season (Figs 10 and 14) with the Sahel zone recording minimal LI values across the entire West African region.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 12. Seasonal Mean Flash Rate Density Climatology over West Africa.

Seasonal Mean LIS Flash Rate Density Climatology over West Africa, December -January -February (DJF) to September-October-November (SON) in progressive order (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g012

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 13. Seasonal RH Climatology over West Africa (1998–2012).

Seasonal RH Climatology over West Africa (1998–2012), with similar description as Figs 10 and 12 at 200 hPa, 500 hPa, 925 hPa.

https://doi.org/10.1371/journal.pclm.0000703.g013

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 14. Seasonal Lifted Index Climatology over West Africa.

Seasonal Lifted Index Climatology over West Africa (1998–2012), with similar description as Figs 10 and 12.

https://doi.org/10.1371/journal.pclm.0000703.g014

The most prominent lightning peaks occurred in JJA (Figs 10 and 12) with significant lightning activities concentrated in the Sudano and lower Sahel boundaries. Mid and upper tropospheric levels also recorded very high RH further inland to the Sudano zone, while its magnitude decreased along the northern parts of the Sahel zone. RH values at 200 and 500 hPa were observed to be relatively high and concentrated mainly in the Sudano and Guinea coasts. Isolated high values of LI were also recorded in the Sudano and lower Sahel, whereas, all other regions observed slightly lower intensities in JJA (Figs 10 and 14). It is, however, necessary to note that RH values at 200 hPa during the JJA season (Figs 10 and 13) were relatively high and spread out across the entire West African region. SON season showed moderate to high LI propagating toward the Guinea coast with the Sahel region experiencing a decline in LI. High RH values observed over the Guinea coasts, however, the retreat of RH southwards resulted in a reduction of lightning activities in the Sudano-Sahel zones. In Figs 10 and 12 relatively low lightning activities were observed to be distributed across a large portion of the West African region in SON, with the maximum peak of 0.25–0.28 fl km⁻²day⁻¹ occurring in Guinea.

Thunderstorm frequency

Temporal evolution of thunderstorm frequencies.

Fig 15b[0pt][-1pc] Figure 15 - The font size of the image is below 6 pt which affects the readability of the image. Hence, please supply a corrected version with font size above 6 pt. shows the TS frequency, as well as monthly and annual trends of TS activities for Axim. From Fig 15b(ii), MAM had the highest observed TS frequencies over the period of 1988-1995, followed by a sharp decline in 1996 to 1999.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 15. Storm frequency and trends over Accra, Axim, Bole and Kumasi.

The Storm frequency, monthly and annual trend of TS activities in Accra, Axim, Bole and Kumasi.

https://doi.org/10.1371/journal.pclm.0000703.g015

The storm frequencies (Fig 15b(ii)) showed a gradual rise from 2000 to 2006 with SON and DJF (1996–2006) exhibiting similar. JJA recorded the lowest TS frequencies across the entire study period for Axim (Fig 15b(ii)), however, September and January also experienced relatively low TS frequencies over this period. The annual trend plot in Fig 15b(i) (the median representation of the frequency data) showed a relatively high frequency from 1988 to 1993 and 2002 to 2006, with the lowest peak in 1997 and a gradual increasing trend on either side of the peak. The monthly trend exhibited double maxima (bimodal), with the highest peak in April and the lowest in November. In Accra (Fig 15a), the inter-seasonal variability of total TS occurrences(Fig 15(ii)) revealed concentrated TS activity during the MAM and SON seasons from 1961 to 1965. Similar activities were observed throughout the study period with JJA and DJF showing the lowest total TS occurrence. The annual trends (Fig 15a(i)) from 1988 to 2006 showed an almost uniform interannual variability with an isolated peak in 1993. However, a high interannual variability was observed between 1961–1965. The monthly trend (Fig 15a(iii)) for the entire period was bimodal with significant peaks in April, May and October.

In Fig 15d (Kumasi), the lowest seasonal TS occurrences (Fig 15d(ii)) were recorded in DJF with JJA following as a close second. The frequencies further intensified during the SON and MAM with the highest frequency occurring in October 1997. However, the year with the most intense TS activity was 1995. A strong interannual variability (Fig 15d(i)) was also observed within 1988 to 1997, with 1989 having the lowest recorded peak. TS activity thereafter intensified at a steady rate till 2001 and relaxed from 2002 to 2005 with a sharp increase in 2006. Although the highest monthly TS (Fig 15d(ii)) was observed in 1997, a relatively low annual frequency (Fig 15d(i)) was recorded as compared to 1995, which had the highest annual TS frequency. The monthly trend plot (Fig 15d(iii)) exhibited a double maximum with the highest peaks occurring in October and May and the lowest peaks occurring in August and July. Significant thunderstorm activities in Bole (Fig 15c(ii)) began in March through to November from 1988 to 2006, except for 1992 which experienced very little activity as compared to the rest of the years. The month of May in 1989 recorded the most intense monthly TS activity, whereas DJF recorded the lowest TS activity over the entire period. The annual trends (Fig 15c(i)) showed intense TS activity from 1988 to 2006 with a sharp decline in 1992 and 1998. Fig 15c(iii) showed an almost uni-modal monthly trend, with the highest peaks occurring in May and September. Fig 16 shows an illustration of TS and Dry TS frequencies against months from 1990 to 2006. A bimodal trend revealed across all stations except for Bole, which depicted a more uni-modal trend of dry TS activity. TS activity for Axim intensified in MAM and SON followed by relaxed frequency of occurrence in JJA and DJF, as dry TS frequencies (Fig 16) slightly mirrored the monthly TS trend. Relatively low dry thunderstorms observed for Accra except in October and November where the dry TS activity was relatively high. Higher dry TS activities were found in Bole as compared to the other two stations. However, these frequencies were relatively low for the TS activity in the station. Kumasi also recorded relatively low dry TS activities to TS activities over the area.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 16. Boxplot of TS and Dry TS frequency.

Boxplot of TS and Dry TS frequency over four stations (Accra, Axim, Bole, Kumasi). Solid lines within the boxes represent medians. Upper and lower values of boxes indicate 75th and 25th percentile; upper and lower whiskers 90th and 10th percentiles.

https://doi.org/10.1371/journal.pclm.0000703.g016

Statistical analysis

In Table 1, an overview of the correlation between convective rainfall amount and flash rate density (FRD) has been provided, along with the significance of the correlation at 95% confidence level. The p-values above 0.05 indicate low/ no significant correlation between the two variables, whereas lower p-values indicate a strong correlation. The Savannah zone had a significant correlation of approximately 78% with the transition, forest and coastal zones recording 49%, 32% and 20% showing no significance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The r-values and p-values between convective rainfall amount and Lightning flash rate density over the various agro-climatic zones.

P-values above 0.05 indicate low/no significant correlation between the two variables.

https://doi.org/10.1371/journal.pclm.0000703.t001

Table 2 revealed a strong correlation (84%) between CAPE and FRD, and an even stronger relationship (87%) between LI and FRD. At 500 hPa, a correlation of about 89% was found between RH and FRD, with a lower correlation of 72% and 70% at 200 hPa and 925 hPa respectively. Furthermore, a noteworthy negative correlation of about 96% was found between mid-level RH (500 hPa) and LI, and about 86% for mid-level RH and CAPE. Additionally, the monthly mean values for convective available potential energy (CAPE) and Lightning flash rate density (FRD) over West Africa is presented in Table 3. From the results, maximum mean CAPE (644 J kg⁻¹) and FRD (0.057 fl km⁻²day⁻¹) were observed in September and June respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The r-values and p-values between lightning flash rate density (FRD), convective available potential energy (CAPE), lifted index (LI) and relative humidity (RH) over the study area.

P-values above 0.05 indicate low/no significant correlation between the two variables.

https://doi.org/10.1371/journal.pclm.0000703.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. The monthly mean values for convective available potential energy (CAPE) and Lightning flash rate density (FRD) over West Africa.

https://doi.org/10.1371/journal.pclm.0000703.t003

Seasonal mean FRD and rainfall climatology

Figs 17 and 18a shows very few lightning activities for DJF in the northern parts of the country. However, intense lightning activities were observed in the southern regions of the country, which intensified towards the south-west. The middle belt recorded relatively average flash rate densities. Correspondingly, minimal rainfall (Figs 17 and 18a) activities were recorded over the southern parts. The lightning and rainfall activities in MAM (Figs 17b and 18b) were on average intense across the country, but, the most significant recordings were concentrated within the southern and north-western parts of the country. The north-eastern sector, as well as the coastal regions recorded relatively fewer lightning activity than the rest of the country.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 17. Seasonal climatology of lightning over Ghana Mean seasonal FRD Climatology over Ghana, December -January -February (DJF) to September-October-November (SON) in progressive order (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g017

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 18. Mean seasonal rainfall climatology over Ghana Mean seasonal rainfall climatology over Ghana, December -January -February (DJF) to September-October-November (SON) in progressive order (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g018

Nonetheless, rainfall activities were relatively high in the coastal regions. Relatively high lightning observed in JJA (Figs 17 and 18c) across the entire country. The northern sector exhibited high lightning activities accompanied by intense rainfall (Figs 17 and 18c) with the upper western sector recording the most intense amounts. Towards the eastern side of the middle belt, lightning activities were observed to intensify significantly. However, relaxed lightning and rainfall activities were observed toward the southern regions of the country, with the coastal regions receiving very few lightning activity. During SON, lightning activities (Figs 17 and 18d) and concentrated rainfall  (Figs 17 and 18d) showed in the lower north-eastern section of the country extending to the entire southern belt. Fewer lightning strikes were observed in the north, accompanied by slightly reduced rainfall intensities.

Spatial and temporal variability of mean monthly FRD and rainfall climatology

Lightning and rainfall activities observed in the southern parts of the country extending towards the coastal and middle sector for January (Figs 19a and 20a). Few rainfall activities were recorded for the middle part of the northern sector, with no significant lightning activities. Intense lightning and rainfall events observed for February (Figs 19b and 20b) in the south-western sector which extended across almost the entire southern sector.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 19. Monthly Mean FRD Climatology over Ghana Monthly Mean FRD Climatology over Ghana, January (a) to December (l) in progressive order (1998–2012).

https://doi.org/10.1371/journal.pclm.0000703.g019

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 20. Monthly Rainfall Climatology over Ghana Monthly Rainfall Climatology over Ghana (1998–2012), with similar description as Fig 19.

https://doi.org/10.1371/journal.pclm.0000703.g020

The northern section of the country recorded low to no lightning activities over the period, however moderately-high rainfall activities were observed in the northern sector. Significantly high rainfall and lightning activities recorded in March (Figs 19c and 20c) across the southern parts of the country, with peak activities (FRD > 0.18 fl km⁻²day⁻¹ and RR > 160 mm) occurring towards the south-western sector. Moderate to low lightning and rainfall activities were recorded across the northern sector. Besides, very high lightning (FRD > 0.22 fl km⁻²day⁻¹) and rainfall activities (RR > 200 mm) were also experienced in Fig 19d and 20d for the month of April, across the entire country. The most significant activities were concentrated around the southern , middle and north-western regions of the country, while, the coastal and north-eastern parts recorded moderate lightning and rainfall activities.

High lightning and rainfall activities covered almost the entire country for the month of May (Fig 19e and 20e). High lightning concentrations were observed in the north-western and south-eastern sectors of the country. Rainfall events were almost uniformly distributed across the country with significant activities concentrated in the lower south sectors. From Fig 19f and 20f, rainfall and lightning activities in June seen almost evenly distributed across the country with concentrated areas occurring in the central and lower west sections of the southern region for rainfall and upper south-east sector for lightning. June recorded one of the highest lightning (FRD > 0.24 fl km⁻²day⁻¹) and rainfall (RR > 480 mm) amounts over the south-western coast. In July (Fig 19g and 20g), intense rainfall and lightning activities were experienced mainly in the northern sector of the country. However, significant rainfall activities observed in the southern regions accompanied by moderate to low lightning activities. August (Fig 19 and 20h) received relatively intense rainfall activities distributed across the middle to the northern part of the country with intense activities concentrated in the northern sector. Lightning activities (Fig 19 and 20h) experienced over this period were highest in the northern regions of the country with little or no lightning activities in the south. Extreme rainfall activities were also recorded across the northern sector extending south of the country for September (Fig 19 and 20i). Isolated intense rainfall activities were observed in the lower north-eastern sector and some parts of the southern Isolated intense rainfall activities experienced in the lower north-eastern sector and some parts of the southern sector Lightning activities (Fig 19 and 20i) mirrored the rainfall distribution patterns, as well as its peak.

Peak intensity of lightning activities found across the lower north-eastern sector propagating towards the southern sector with slightly lower intensity for October (Fig 19 and 20j). All other regions recorded a much lower concentration of lightning activities. The rainfall activities (Fig 19 and 20j) across the country followed the lightning distribution patterns with isolated concentrations occurring in the south. Relatively low to no rainfall distribution was observed in November (Fig 19 and 20k) across the country whereas, some rainfall activities were recorded across the lower southern belt and an isolated event in the north. The lightning activities (Fig 19 and 20k) mirrored the rainfall patterns with intense concentration in the lower southern sector extending across the southern belt. In December (Fig 19l and 20l), rainfall and lightning activities were distributed across the south and a small portion of the north, with mildly intense activities (RR > 100 mm, FRD < 0.14 fl km⁻²day⁻¹) concentrated in the south-west extending towards the outer south-eastern parts of the country.

Instability

CAPE, a conditional instability parameter of the tropical atmosphere, appears to play a vital role in the occurrence of lightning in the West African region, as it exhibited strong positive correlation over the observed timescales. This finding is in agreement with the findings of [54], who investigated the relationship between CAPE and lightning at every timescale. A reasonably strong relationship between lightning activity and CAPE on monthly scale (r=0.84 at 99% confidence interval) was obtained. This result provides support for the work of [55], who theorized the possibility of predicting the lightning class of occurrence using CAPE with high accuracy. It is noted in scientific literature that, this relationship is often orography and local meteorological conditions dependent [56] with lower elevations contributing to higher CAPE, larger precipitating areas [57] and more lightning activity. This assessment also applies to low-lying areas in West Africa (e.g., countries within the Guinean-Sudan regions) and is consistent with results obtained by [58] for the mean annual total flash rate over Africa. A concomitant increase of lightning with larger CAPE was also observed. Although this is consistent with CAPE and lightning behavior elsewhere [14,59], the observed lag between larger CAPE in April and maximum lightning activity in May by [54] was not evident. However, a seemingly similar phenomenon was observed between May and June instead. A variety of factors may be responsible for this shift. These include seasonal variations of surface temperature [60,61], orography and prevailing synoptic conditions [62,63], and the behavior of the ITCZ in both regions [64].

From the results, CAPE increases from March to May and migrates northward to the Sudan and Sahel regions. Interestingly, over the West African region, the maximum mean CAPE occurs in September. This is in contrast with the findings of [65] who found the maximum median CAPE over Bangladesh to occur in April. June recording the highest mean lightning activity indicates, convective storms in the West African region are at their peak in June. Our finding of a secondary peak in September during the monsoon is probably attributable to the merging of the two rainy seasons farther north to give a single wet season between June-July and September. In addition, the West African monsoon may also play a factor as it exhibits a clear seasonal cycle, reaching a peak between July and September. This observed peaks in June-July and September are also consistent with the findings of [66] over Maharashtra, India. Although the influence of CAPE on lightning occurrence is consistent with several studies [54,58], our results contradict the findings of [63] who reported little to no effect of CAPE on lightning during pre-monsoon in the Indian region. This disparity between accounts could be a factor of analysis scale, synoptic systems [62] and local climatic influences.

From the spatial and statistical analysis, there exists a strong association between lightning and relative humidity, especially RH at 500 (r=0.89). It is known in literature that temperature and humidity are the dominant factors influencing the spatial and temporal distribution of thunderstorms [58,67]. Since the inflow of moisture is strongly associated with the movement of the ITCZ and CAPE in the West African region, countries like Mali, Niger, Nigeria and Sudan experienced maximum lighting activity during the north-most reach of the ITCZ, and an almost complete lack of lightning at its retreat. This retreat of the ITD southward, advects moisture (increasing humidity) into the atmosphere which provides a favourable environment for thunderstorm activity along the Guinean coast [68]. This result corroborates the findings of [69] and [70] that TS occurrence increases from the north to south.

Thus, RH evidently plays a key role in the occurrence of lightning within the West African region as well as a suppressive effect on convective initiation in the mid-troposphere. Even so, more than CAPE (r=0.84) or Lifted index(r=0.87). This finding is consistent with the study of [18,22,59]. For instance, [18] established a close relationship between tropospheric water vapor variability and global lightning activity. [22] summarizing several studies concluded that, dry mid-tropospheric profiles (dry air) acts to suppress deep convection. [59] also arrived at a similar conclusion in their investigation into favorable environments for thunderstorms over Europe. The strong negative correlation found between LI and flash rate density is also in agreement with literature [71]. That is, convective activity decreases as LI increases positively and with it, a decline in lightning occurrence [52].

Given adequate mean CAPE (>400 J kg⁻¹) and sufficient low-mid tropospheric relative humidity, the mean occurrence of lightning activity is significantly increased. From the results it can be observed that, high lightning occurrence is achieved only for a combination of both sufficient CAPE, LI and high RH values. This dependence of lightning occurrence on CAPE and LI indicates that a vast majority of thunderstorms occur with some instability. Lightning activity was also observed to increase with both CAPE and LI. This finding agrees with [72]’s study but contradicts the study of [59]. [59] noted that, the relative frequency of lightning only increased until a particular level of instability(-4.5 ^° C), regardless of the instability parameter (CAPE or LI). This disparity in accounts maybe attributable to our data set not including LI values below -5 ^° C and/or difference in regime. From Table 2 we have demonstrated that mid-level humidity is closely linked to other instability parameters like CAPE, Lifted index, and Lightning occurrence. This is consistent with the findings of [59] as they investigated conditions favorable for the occurrence of electrified convection. Our finding also corroborates the results of [73] and [74] that, increases in CAPE are associated with increases in near-surface temperature and/or water vapor (humidity). The relationships between RH, CAPE, LI and lightning occurrence is of significant relevance as many traditional predictors for thunderstorms do not account for mid-level humidity, or worse, predict a higher probability of thunder with decreasing humidity in the mid-troposphere. This is the case for indices like the Bradbury index [75], the KO-index [76] and the Potential Instability Index [77] that represent potential instability as the vertical gradient of equivalent potential temperature (θ e). Consequently, such indices yield poor to mediocre performance when compared to other thunderstorm indices [72,78,79]. It is also worth noting that conditional instability is nearly always present along the Volta river. This is associated with lightning activities resulting from orographic lifting and evaporation from the Volta lake, making it a conducive area for convective initiation [41]. A similar assertion was made by [80], who found lightning density to be often enhanced over the Lake Victoria in Congo as a result of evaporation and moisture at low levels over the area.

TS frequency analysis

From Sect , TS frequency in MAM were observed to be extremely high across all stations and observed years with maximum frequency at Axim occurring in April and, Accra and Bole occurring in March. SON season also showed high TS frequencies across all stations over the study period, with most TS activity in Kumasi occurring in October. These seasonal trends (uni-modal and bimodal) observed in the monthly trend, attributable to the migration of the ITD across the country, which equally produces the two rainy seasons in the south (Axim, Accra and Kumasi) and one rainy season in the north (Bole) [40]. Furthermore, the low TS activities recorded for DJF and JJA can be attributed to the dryness associated with the period and insufficient atmospheric instability respectively, since TS formation requires significant inflow of humidity and sufficient atmospheric instability as suggested by [81]. Generally, it can be inferred that the seasonal climate pattern of the country influences the TS activities. High interannual variabilities were observed over the study period, with recorded rise in TS activities over the later 2000s. However, the total TS activity in the latter part of the 80s and 90s were higher across the entire period, and consistent with the findings of [41] who attributed these variabilities to the effects of strong El Niño and La Niña episodes on the region. However, it is unclear if this trend results from a cyclic recurrence that characterises all meteorological processes or by the strengthening of anthropogenic activities on the climate [82]. Furthermore, the collectively low TS activities observed over the four stations assessed are in agreement with the studies of [83,84] that found decreasing trends in precipitation over Ghana for 1980-2000, particularly in the south.

From Figs 15 and 16, the TS frequencies for Bole were relatively higher than all other stations, accompanied with moderate dry TS occurrences. It, therefore, implies that although TS activities are incredibly high over Bole, the relatively dry nature of the terrain limits convective precipitation thus explaining the higher dry TS activity [41,68]. Although the dry TS activity over Axim was unavailable, the dry TS of Kumasi serves as a useful baseline for intercomparison as they both lie in the forest zone. It can therefore be concluded that about 25% of the TS activities over Accra, Axim and Kumasi are not always accompanied with rainfall and somewhat attributable to single TS clouds observed to develop during the mid- afternoons, later producing one or two thunderclaps and finally dissipating [85]. The boxplot (Fig 16) showed dry thunderstorm events over all the stations during the months of July and August, consistent with the characteristic dryness associated with the northern-most movement of the Inter-Tropical Discontinuity (ITD) [86]. From Table 1 it is evident that the Savannah zone receives the highest TS activities among the four stations. In contrast, fewer rainfall activities in the Transition, Forest (Kumasi and Axim) and Coastal zones (Accra) linked with lightning, as evident in Bole (Figs 15c and 19) which is located in the Savannah zone.

Rainfall and lightning activity

Although rainfall and lightning activities were found to be associated, on a monthly basis, with a correlation coefficient of r = 0.66, the relationship at seasonal scale was low, relative to monthly scale with no significance. However, our finding is consistent with [54] whose result also revealed moderate correlation between lightning and rainfall. [87] further elaborating on the relationship between lightning and rainfall pointed out that the relationship varied with season, climatic conditions and geographic location. Lightning activities observed mostly mirrored rainfall activities across the study region with relatively high lightning flash rate densities often accompanied by intense rainfall and vice versa. This finding contradicts that of previous storm-scale studies which reported reduced rainfall when thunderstorms yielded more lightning (e.g., [54,60,88]).Although the results are not directly comparable to those noted above, we believe that the different accounts may have been caused by precipitation pattern and characteristics introducing subtle variation as rain yield varies according to climatic regime.

Lightning and rainfall activities in June were observed to be quite intense and evenly distributed across the country with isolated systems over the middle and southern belt, consistent with the peak of the so-called major rainy season over southern Ghana. The recorded intense rainfall and lightning activities in June are attributable to intense monsoon flow initiated by the ITD movement consistent with [69]’s findings. [89] in their work over Kolkata, India also reported that, intense lightning flashes are related to the active monsoon phase, which play a role in the generation of convective activities, and lightning. This assessment is consistent with Fig 17b and 17c, marking the onset and peak of the monsoon period. Low lightning intensity and distribution across the country in December, January and February are linked to the Harmattan dominating this period, as can be seen in the minimal rainfall amounts and distributions. Comparison of rainfall estimates and lightning strikes by [86] revealed that more lightning occur over land than the coastal terrains. This is consistent with results obtained in this study. The sea breeze effect along the coast and the warming process over the ocean may account for the reduced lightning activities over the area.

The present study is limited in a number of ways. First, this paper does not cover the annual and diurnal variations over Ghana, which is crucial in understanding the characteristics of lightning over the area. Thus, future works will explore the diurnal and annual variation of lightning flash rate densities, and consider more instability and convective parameters (Total Totals, K-index and Convective inhibition) to gain in-depth understanding into their relationship with lightning over the West African region. Secondly, although the number of available stations used in the local thunderstorm studies span for a significant number of years, it is not sufficient to draw conclusive remarks on thunderstorm occurrence in the country as only four weather station data was used in the analysis.