1 Introduction

During the last decades, the need of energy in Cameroon has significantly increased due to industrial, agricultural, and demographic development. In Cameroon, the increase in electrical demand is estimated to about 8% per year (Legros et al. 2009). To fulfill this demand, Cameroon has constructed large-scale hydropower plants. The three sites, Song Loulou (384 MW), Edea (263 MW), and Lagdo (72 MW), associated with reservoirs which total storage capacity reaches 7.6 × 109 m3, are built on tributaries of the Sanaga river (Mbakaou, Mape, Bamendjin) and give a total of 719 MW (Tchanche 2014). However, the energy needs remain unsatisfied (NKUE and NJOMO 2009) due to the uneven distribution of resources and the fact that the indigenous population is too poor to afford commercial energy (Mboumboue and Njomo 2016). Moreover, the populations connected to the electrical grid are constantly affected by repeated power outages due to network shedding campaigns. Power outages and the high cost of electric power lead to a delay in gross domestic product (GDP) growth of about 1 to 2% per year (Ayompe and Duffy 2014).

Consequently, wood remains the main energy consumed in Cameroon. It covers around 72.6% of the total energy consumption and its annual demand is about 2.4%. Kerosene is also widely used in lamps for lighting, with however marked disparities between urban and rural populations (67% and 10% respectively, Kapseu et al. 2014). However, electricity generation based on fossil fuels emits greenhouse gases (GHG) that are among the causes of global warming. Therefore, to reduce GHG emission and slow down the global warming process, the use of renewable energy is the best solution to opt for.

Cameroon benefits of a high potential in renewable energy either from solar, wind Ngnikam (2006) biomass, or geothermal sources. The solar source has become economically and technologically viable (Kapseu et al. 2014). A previous study (Mas’ud et al. 2015) concerning the situation of renewable energy in Africa, and particularly in Cameroon and Nigeria, has demonstrated that the two countries benefit of high solar radiation and also excellent wind speed, which can be effectively used for electricity generation.

The solar radiation received in Cameroon in 2014 was estimated to 4282 Wh m−2, a value corresponding to a potential of 7.431 × 1011 MWh and representing 129 times the total electric production of the country evaluated at 5.761 × 1012 Wh (Mas’ud et al. 2015). These estimates confirm that Cameroon has a high potential and that this sector of energy has to be exploited (Nematchoua et al. 2015a, 2015b; Ayompe and Duffy 2014).

However, as shown by Fig. 1, the incident solar radiation in Cameroon varies significantly and increases as according to the latitude (source: www.solargis.info). Douala, the most industrialized city in the country and where the need for energy is very important, has the weakest solar resource according to this figure.

Fig. 1
figure 1

Cameroon Solar potential map for the period 1994–2010 (source: www.solargis.info). Left panel: the average annual sum of horizontal global solar radiation in (kWh/m2); right panel: the average annual sum of direct solar radiation (in kWh/m2)

Full size image

The knowledge of the solar resource variability in time is a determining factor to design, to dimension, to evaluate the performance assessment, and to manage the renewable energy conversion systems (Ayompe and Duffy 2014). The evaluation for Douala of the horizontal global solar radiation and the study of its daily and seasonal variations could be beneficial for its exploitation. These variations are influenced by local weather and climatic conditions. Therefore, the aim of this study is to picture these variations and their drivers for Douala. A first question we want to address is: what are the periods of highest or lowest solar radiation in Douala throughout the year? A second question is: what kind of daily solar radiation are recorded during these seasons? A third question is: what explains the daily variations observed and to which meteorological parameters are they linked? The paper is organized into four sections. The second section presents data and methods used. Results are presented in Sec. 3 and the conclusion is provided in Sec. 4.

2 Data and methods

2.1 Study domain

Cameroon is located in Central Africa along the Gulf of Guinea (Fig. 2). It lies between latitudes 1° 40′–13° 05′ N and longitudes 8° 30′–16° 10′ E and covers a surface area of about 475,650 km2. It is divided into two climatic regions: the humid tropical region in the south and the semi-arid region in the north.

Fig. 2
figure 2

Location of Douala (adapted from Google map). Where solar radiation data from HelioClim3v5 were extracted for the coordinates 4.050605 N–9.766213 E (P14). The cloud type data were extracted for the region 4.0–4.5 N and 9.41–9.43 E

Full size image

Douala is the economic capital of Cameroon. It is located at 9.7° E and 4.0° N at the Atlantic coast, by the Wouri River. It is subjected to an equatorial climate and influenced by the West African monsoon flux deviated to the east. It is uplifted by the Mount Cameroon Mountain, located 60 km west and culminating at 4070 m. Douala is one of the largest cities in Central Africa harboring more than 70% of Cameroon industrial productions which explains the high energy demand and its regular increase (Suchel 1987).

2.2 Data

2.2.1 Satellite products

The HelioClim3-v5 solar radiation data

HelioClim3-v5 database provides global solar irradiance and radiation estimated at ground level from February 2004 to day 1 for the Europe–Africa disk (see http://www.soda-pro.com/home). The database was created using Heliosat method (Rigollier et al. 2004) which relies on observations made by Meteosat Second Generation (MSG) and provides estimates of the global radiation at ground level. The acquisition and processing of MSG measurements to obtain solar products are explained in Qu et al. (2017), Beyer et al. (1996), Hoyer-Klick et al. (2010), and Rigollier et al. (2004). The SoDa web interface (http://www.soda-pro.com/home) allows retrieving horizontal global solar radiation (GSR) daily values or monthly averages for any location in the Europe–Africa disk. For the purpose of this work, we extracted the global, diffuse, and direct solar radiation components for Douala (4.050605 N–9.766213 E) over the study period from 2005 to 2015 with a time step of 15 min. We upscaled the data to the hourly resolution.

The SAFNWC cloud types products

The Satellite Application Facility for Supporting NoWCasting (SAFNWC) cloud classification product offers a classification of clouds into 21 classes and at 3 km and 15 min space–time resolution. It is issued from algorithms using reflectance and brightness temperatures from the Spinning Enhanced Visible and Infrared Imagery (SEVIRI) instrument on board of MSG, sun and satellite angles data, Numerical Weather Prediction forecasts, and climatological fields of sea and land surface temperature, air temperature along the atmospheric column, total water vapor content, and altitude (Derrien and Le Gléau 2005, 2010). The cloud classification data were acquired for 2008–2015 in six grid points around the city of Douala with a temporal (spatial) resolution of 15 min (0.125°lat × lon). Following Philippon et al. (2016), the 21 classes were reorganized into 5 main classes (Table 1): clear sky (classes 1–4), ultra-low and low clouds (e.g., stratus, classes 5–8), medium clouds (classes 9–10), high and very high opaque clouds (e.g., cumulonimbus, classes 11–14), and semi-transparent clouds (e.g., cirrus, classes 15–18). Pixels belonging to classes 0, 19, and 20 (missing or corrupted, fractional, and undefined) were excluded from the analyses.

Table 1 Twenty-one classes of clouds from SAFNWC rearranged into five main types after Philippon et al. (2016). Column 1: code associated to the cloud type; column 2: the class name; column 3: example of cloud type; column 4: the short name adopted
Full size table

2.2.2 In situ measurements

Daily surface temperature, daily sunshine duration, monthly global solar radiation on horizontal surface (2000 to 2016), and monthly precipitations (2011–2016) based on synoptic reports data for the meteorological station of Douala (World Meteorological Organization—WMO—index of 64,910) have been acquired. The Ogimet weather information service stores database for all meteorological services around the whole world. It is based on WMO and the International Aviation Organization (ICAO) station indices. The daily means are based directly on climate reports (FM 71-XI), usually available during the first 10 days of the next month (more details can be found at: http://www.ogimet.com/guia.phtml.en). These synoptic reports data have been used successfully by Schneider et al. (1992), 2005) to estimate the monthly precipitation. The authors confirm that these reports can be used with a good accurately. Furthermore, observed data on several climatic parameters (temperature, humidity, solar radiation, wind speed, and global luminance) from synoptic reports have been used also by many authors to characterize climatic variability (Enete et al. 2014; Kemajou et al. 2012; Nematchouaa et al. 2016), to analyze the weather conditions (Lenouo et al. 2009), and to estimate global solar radiation in some regions of Cameroon (David and Ngwa 2013). The various data are measured with frequencies of 10–15 min and in height from the ground between 3 and 10 m.

2.3 Methods

2.3.1 Determination of the main types of diurnal cycles of GSR

Following the approach developed by Philippon et al. (2016) for northern Congo, the diurnal variations of solar radiation at Douala and their seasonal evolution are pictured thanks to a classification of the diurnal cycles into main types and then the computation of their percentage of occurrence during the annual cycle. The main types of diurnal cycles of solar radiation were detected using a hierarchical cluster analysis (HCA). HCA has been set with Euclidean distance as dissimilarity measure and the Ward method as aggregation process. The number of classes was varied from three to seven, and a partition into five classes was selected after analyzing the clustering tree.

To satisfy the input conditions of the software (XLSTAT) used for the clustering process, we organized the HelioClim3-v5 data in days × years. After the application of the clustering process to the GSR component, we obtained the five main diurnal cycles for Douala (cf. Fig. 8a). Then, the corresponding diurnal cycles of the direct and diffuse components were obtained in two steps. First, we gathered all days corresponding to each class. Then, we calculate the average of x days for each class to obtain the corresponding mean diurnal cycle (cf. Fig. 8b, c).

Lastly, the computation of the occurrence percentages of the 5 diurnal cycles throughout the annual cycle, enables to picture interactions between the daily and the seasonal time scales (cf. Fig. 11).

2.3.2 Assessment to the relationships between GSR and cloudiness

Cloudiness mean seasonal Nm evolution has been estimated using the in situ sunshine duration data based on relation (1) (Besharat et al. 2013):

$$ \frac{S}{S_0}=1+\frac{N_m}{8} $$
(1)

where S the effective sunshine duration and S0 the theoretical sunshine duration given by (2):

$$ {S}_0=\frac{2}{15}{\cos}^{-1}\left(\tan \delta \tan \varphi \right) $$
(2)

φ being the latitude and δ the solar declination (cf. Fig. 6).

Then, to assess the type of cloud cover associated with each type of GSR diurnal cycle, the total percentage of the diurnal cycle for the five types of cloud cover were computed.

First, we determine the mean percentage for each time step (1/4 h) and for each type of cloud by the relation (3):

$$ {p}_i=\frac{n_i}{6};1\le i<5 $$
(3)

where pi is the percentage associated with the encoded cloud I and ni the cloud i occurrence among the 6 pixels documenting Douala.

Second, we computed the average over the time steps to obtain the mean daily percentage p associated with each cloud type (4).

$$ p=\frac{\sum {p}_i}{24X4};1\le i<5 $$
(4)

Third, we projected the five diurnal cycles of solar radiation into the five types of cloud cover. We extracted all the days belonging to the same class of GSR diurnal cycle and then calculate the clouds mean percentage (cf. Table 2 and Fig. 3) and clouds mean hourly percentage (cf. Fig. 9a–e). Lastly, we computed the Pearson’s correlation coefficient r to determine how well the GSR and cloud diurnal cycles are related (cf. Fig. 10).

Table 2 Statistics for the five types of GSR diurnal cycles: Per., percentage of occurrence (4017 days available); difSR, daily diffuse solar radiation; dirSR, daily direct solar radiation; GSR, horizontal global solar radiation. Associated occurrence percentage of the five types of cloud cover: CS, clear sky; uLow, ultra-low and low clouds; Med, Medium cloud; H_vHop, high and very high opaque clouds; sTp, semi-transparent clouds
Full size table
Fig. 3
figure 3

Monthly average of daily horizontal global solar radiation in Douala (HelioClim3v5, period 2005–2015

Full size image

2.3.3 Simulation of the GSR mean seasonal cycle using the observed temperature

Simulation of the mean seasonal cycle of GSR was carried out using the observed temperature in order to determine to what extent it could be predicted using a theoretical model.

Numerous models have been previously proposed for various locations in Africa to estimate global solar radiation on the horizontal surface on both daily and monthly time scales. Therefore, various input parameters and different functional forms have been used. Some of these models are described in Erbs et al. (1982). In this work, we use a model based on the temperature observed. This model was proposed by Hargreaves et al. (1982) who demonstrated that the GSR can be estimated from the difference between the maximum and the minimum air temperature. The model has been used successfully by Zhou et al. (2006). Later, Annandale et al. (2002) introduced a correlation factor in the relationship between the GSR and the zone elevation given by Eq. (5).

$$ \frac{G}{G_o}=a\left(1+2.7\times {10}^{-5}Z\right){\Delta T}^{0.5} $$
(5)

where G is the horizontal global solar radiation in Watt-hour per square meter per day, ∆T = Tmax − Tmin is the difference between the daily maximum and minimum air temperature, and a is an empirical coefficient, whose value should be 0.16 for interior regions and 0.9 for coastal regions. As Douala is a coastal city, it is set to 0.9 Douala (Ayompe and Duffy 2014; Antón et al. 2017; Pecenak et al. 2016; Boch et al. 1981); Z is the elevation in meters. Go is the extraterrestrial radiation in Watt-hour per square meter per day, which can be calculated for any day of the year and any latitude (details can be found in Duffie et al. 2013; Almorox et al. 2011; Erbs et al. 1982; Odutayo et al. 2016; Kerkouche et al. 2013).

For this model, the evaluation criteria adopted is the coefficient of determination R2 (Mas’ud et al. 2015).

3 Results

3.1 Global solar radiation mean seasonal evolution in Douala

This part presents the mean seasonal solar cycle of GSR in Douala. A few meteorological parameters such as sunshine duration, surface temperature, precipitations, and cloudiness throughout the year are also pictured to explain the mean seasonal cycle of GSR.

3.1.1 GSR mean seasonal cycle as pictured from HelioClim3-v5 database

To efficiently use the solar resources in Douala, it is important to assess the periods of the year which have the highest and lowest GSR and also the meteorological parameters associated with this evolution. Figure 3 presents the monthly average of daily GSR in Douala obtained from HelioClim3-v5 while Figs. 4, 5, and 6 correspond to those of sunshine duration, temperature, and cloudiness, respectively. Four seasons are clearly visible on these figures.

Fig. 4
figure 4

Monthly average of daily sunshine duration (left axis) and monthly average of daily temperature (right axis, Tmin, Tmean, and Tmax) in Douala (www.ogimet.com), period 2011–2016

Full size image
Fig. 5
figure 5

Monthly mean rainfall amount in Douala (www.ogimet.com, period 2011–2016)

Full size image
Fig. 6
figure 6

Monthly mean of cloudiness computed using sunshine duration for 2000–2016 (source: www.ogimet.com)

Full size image

In December–January–February (DJF), GSR is around 5000 Wh/m2/day (Fig. 3), a period when sunshine duration remains above 6 h and the temperatures (min, max, and mean) reach their maximum (Fig. 4). The strong illumination during this period is also related to little precipitations (Fig. 5) and low cloudiness, less than 5 octas (Fig. 6). In a recent study of the weather conditions for aviation safety in Douala, Lenouo et al. (2009) find a frequency of rainy days less than 5 days/month and refer to this period as the great dry season. Nematchoua et al. (2015a, 2015b) investigated on climate change and its role in the future energy demand of buildings in Douala and found a maximum of cooling energy needs in DJF. The need of cooling in this period is due to the high level of temperature (Fig. 4) (Molua 2006).

In March–April–May period (MAM), GSR is lower and about 4200 Wh/m2/day (Fig. 3) in agreement with a decrease in sunshine duration (5–6 h) and temperatures (Fig. 4) and an increase in cloudiness from 5 to 7 octas (Fig. 6) and precipitations (> 150 mm/month; Fig. 5). This period is locally called the “small rainy season” and is important for agriculture as a preparatory phase before the great rainy season of June–July–August–September (JJAS) (Molua 2006).

In JJAS, the great rainy season, the daily GSR is at its lowest levels, less than 3500 Wh/m2/day (Fig. 3). Temperature and sunshine duration are also at their lowest levels (Tmax < 29 °C and sunshine duration less than 4 h) (Fig. 4). On the other hand, precipitation and cloudiness reach their maximum values (> 500 mm/month, Fig. 5 and > 7 octas; Fig. 6). This period corresponds to the monsoon season in Douala: it is characterized by a high frequency of rainy days, more than 20 days/month (Lenouo et al. 2009), and of high intensity. This confirms the results on Fig. 5 and agrees with those of Molua (2006) and Nematchoua et al. (2015a, 2015b) who estimated precipitation rate over 600 mm/month in JJAS, using different rainfall data.

Lastly, in October–November (ON) period, GSR mean values increase to about 4400 Wh/m2/day (Fig. 3), i.e., nearly the same level as in March–May. This is an intermediate season between the great rainy reason in JJAS and the great dry season in DJF. It is associated with a rapid decrease in precipitations (Fig. 5) and cloudiness (Fig. 6) (Guévart et al. 2010) and an increase in temperatures (Fig. 4) (Lenouo et al. 2009; Guévart et al. 2010).

3.1.2 GSR mean seasonal cycle according to observations data and a mathematical model

Figure 7 presents GSR annual cycle using ground data measurements and its estimation from a mathematical model (Annandale et al. 2002) (see Sect. 2.3 for model description).

Fig. 7
figure 7

Comparison between the monthly average of GSR simulated by the model of Annandale et al. (GSRAnan) and the satellite estimation of HelioClim3-v5 (GSRhelio) (a mean seasonal cycles, b scatter plots) and between the in situ measurements data (GSRmes) and the satellite estimation of HelioClim3-v5 (GSRhelio) (c mean seasonal cycles, d scatter plots). In gray is the limit of the confidence intervals, and regressions in black

Full size image

Results from model have been compared to GSR obtained from HelioClim3-v5 data (Fig. 7a, b). The seasonal profile (Fig. 7a) and the scatterplots (Fig. 7b) with R2 around 96.4% confirm that in situ observations of temperature can be used with confidence to approximate the mean seasonal evolution of GSR (Hassanalian et al. 2018) in Douala. It is however noteworthy that slight differences exist between GSR estimated from in situ temperatures and from HelioClim3-v5. In February to May, GSR estimated from temperatures (GSRAnan) is higher than GSR estimated from HelioClim3-v5 (GSRhelio) and in June–September GSRAnan is lower than GSRhelio.

Results from ground data measurements have also been compared to GSR obtained from HelioClim3-v5 data (Fig. 7c, d). R2 is around 96% (Fig. 7d); this confirms a good coincidence between satellite estimation and the ground measurements although the mean monthly solar radiation ground data is slightly high compared to the mean seasonal cycle estimated from satellite data base in DJF (Fig. 7c).

3.2 GSR diurnal cycles and cloud types

In the preceding section, we have defined the four main seasons in Douala based on GSR and some meteorological parameters. Now we examine the behaviour of GSR diurnal cycles during these four seasons.

Many authors consider clouds as the most important factor which influences incoming solar radiation (Ramachandra et al. 2011; Guévart et al. 2010; Besharat et al. 2013). In this section, diurnal solar cycles are presented with the associated daily percentages of cloud types. Pearson’s correlation coefficients are used as the metrics to determine the cloud types which better explain these cycles.

Applying the HCA method (see Sec. 2.3), we extracted five different types of GSR diurnal cycles referred to C1 (very bright), C2 (moderately bright), C3 (dark in the afternoon), C4 (dark in the morning), and C5 (obscure). The five types are presented in Fig. 8a, with the corresponding diurnal cycles for the direct (Fig. 8b) and diffuse (Fig. 8c) components. Figure 9 presents the mean hourly percentages of the five types of clouds (Fig. 9a–e).

Fig. 8
figure 8

a The five types of GSR diurnal cycles (C1 to C5). b, c The corresponding diurnal cycles of direct and diffuse components. The mean diurnal cycle obtained as the average of the 4017 available days is displayed in green

Full size image
Fig. 9
figure 9

Mean hourly occurrence percentages of the five types of cloud cover (CS clear sky, uLow ultra-low and low clouds, Med medium clouds, H_vHop high and very high opaque clouds, sTp semi-transparent clouds) for each of the five (C1–C5, panels a–e) diurnal cycles of GSR

Full size image

Class 1 which represents 42% (Table 2) of the total number of days analyzed (Table 2) corresponds to very illuminated days (GSR = 5200 Wh/m2/day) with, as expected, a maximum value at 12H in the three components, reaching 700 Wh/m2 (Fig. 7). These very bright days are associated with a predominance of clear sky (CS) ranging from 40 to 60% (Fig. 9a) and to a lesser extent transparent and semitransparent clouds (sTp) (cirrus and cirrostratus) at 30%. The high opaque clouds (H_vHop), responsible for great solar attenuation, are infrequent.

Class 2 represents days with an illumination close to the average (Fig. 7a). They have a lesser percentage (14%) and lesser illuminated than class 1 (GSR = 4200 Wh/m2/day; Table 2), especially in the mornings (Fig. 7a). During these days, clear sky conditions are predominant in the afternoons (> 40%), while ultra-low and low clouds (uLow) are frequent in the mornings (~ 30%). These variations are in opposite directions. The percentage of CS increases and the percentage of uLow decreases; this explains the variation of hourly global solar radiation from 480 Wh/m2 at 11 am to 645 Wh/m2 at 14H.

Class 3 (18% of the total number of days analyzed and with GSR = 3800 Wh/m2/day; Table 2) corresponds to days which are dark in the afternoons, i.e., an opposite situation of class 2 particularly obvious for the direct component (Fig. 7). However, in terms of diffuse component, class 3 shows higher levels, so that the direct to diffuse components ratio is much lower for class 3 than class 2 (0.54 for C2 and 1.06 for C3). This class is associated with a percentage of uLow which increases from 13 to 27% at 10H to 19H while the percentage of clear sky decreases and that of high and very high opaque clouds (H_vHOp) slightly increases at 13H to 19H (Fig. 8d). This suggests that small variations in the percentage of H_vHOp have strong influence on solar radiation (Zhuravleva et al. 2017).

Class 4 corresponds to days darker than classes 1–3 (2800 Wh/m2/day; Table 2), especially in the morning (the hourly GSR is less than 300 Wh/m2; Fig. 7d). They account for 16% of the total number of days analyzed. They are also characterized by a low direct radiation as compared to the diffuse one so that the direct to diffuse ratio equals 0.51. The darkness of the sky, especially in the morning, is due to very week percentage in clear sky (< 20% note that it increases between 13H and 15H with the top of atmosphere solar (Antón et al., 2017) and high percentage of uLow, but mainly medium and H_vHop clouds as compared to classes 1 to 3.

Lastly, class 5 with the least percentage (10% of the total number of days) corresponds to very obscure days (GSR is 1900 Wh/m2/day) with the lowest hourly global solar radiation (< 300 Wh/m2) during the whole day (Fig. 7). Compared to class 1, the brightest days, the percentage of clear sky and sTp is less than 10%, while the percentage of uLow and H_vHOp are above 50% (Table 2). The sky is very cloudy with a slight increase of the percentage of high opaque clouds during the day at the expense of low clouds. The small increase in the percentage of the H_vHOp seems to affect solar radiation more than variations in the uLow percentage. This implies that the high opaque clouds return solar energy and therefore contribute more to the albedo effect (Coley and Jonas 1997). The direct radiation is lower than the diffuse one, so that the ratio equals 0.52. The sun rays have undergone a strong attenuation throughout the atmospheric layers during these types of days.

The Pearson correlation coefficients between the diurnal cycles of GSR and the cloud types are shown in Fig. 10. Results can be classified into two categories: those with r positive and those with r negative. The category with r positive corresponds to sky that do not attenuate incoming solar radiation, i.e., CS and sTp. The category with negative r contains clouds which are responsible for a strong solar radiation attenuation. This category embeds low, medium, and high opaque clouds. These last three types of clouds explain well the diurnal cycle of GSR.

Fig. 10
figure 10

Pearson’s correlation (Karl, 1895) coefficient between the five diurnal cycles of GSR and the five types of cloud cover. CS clear sky, uLow ultra-low and low clouds, Med medium cloud, H_vHop high and very high opaque clouds, sTp semi-transparent clouds

Full size image

3.3 Occurrence percentages of the five types of GSR diurnal cycles throughout the year

We have just described the five diurnal cycles of solar radiation in Douala and the associated cloud types. Each day corresponds to one of the 5 cycles. It is now important to look for the occurrence percentage of each type of diurnal cycle during the seasonal cycle to determine its importance.

The occurrence percentages of the five types of GSR during the seasonal cycle at a monthly time step are presented in Fig. 11.

Fig. 11
figure 11

Monthly occurrence percentage of the five types of diurnal cycles of GSR

Full size image

The DJF season is dominated by illuminated days (C1) with percentage over 60% and up to 80%. This explains the high levels of GSR and temperatures, and the long sunshine duration recorded in Douala during this season (Figs. 3 and 4), as well as the low rainfall amounts (Fig. 5).

This high solar potential period is followed by the March–April–May (MAM) period and preceded by the October–November (ON) one. In both periods, C2 and C3 represent 30% of days, i.e., characterized by semi illuminated days, probably due to the phenomenon of convection. During June–September (JJAS), there is a dominance of dark days, i.e., C4 and C5, more than 25%. During the whole day, the sky is covered with Low and H_vHOp clouds in different phases. Conversely, the illuminated days (class 1) become infrequent, in July and August particularly.

4 Conclusion

For a sustainable and efficient exploitation of solar energy in Douala (Cameroon) in order to satisfy the growing needs of electrical energy and anticipate the problem of global warming, an analysis of solar radiation evolution throughout the annual and diurnal cycles has been carried out.

Based on HelioClim-v3 solar radiation estimates for 2005–2015 and in situ meteorological measurements, four seasons were detected which translate into different levels of GSR which are sunshine duration, precipitations, temperature, and cloudiness. The main bright and dry season occurs in December–January–February with GSR ~ 5000 Wh/m2/day, whereas the main dark and rainy season occurs in June–July–August–September with GSR ~ 3000 Wh/m2/day (and an amount of rainfall greater than 2 m). March–April–May and October–November are intermediate semi-illuminated seasons.

The solar radiation seasonal cycle has also been simulated from the seasonal cycle of in situ temperatures using Annandal et al. (2002) mathematical model. GSR seasonal cycles estimated from satellite retrievals and in situ temperatures show a good agreement (R2 = 0.96).

For a better comprehension of the seasonal cycle of the GSR, the diurnal cycles have also been analyzed. Using hierarchical clustering analysis (HCA), five types of diurnal cycles were detected: those corresponding to illuminated days (C1), moderately days (C2), semi-illuminated days (i.e., slightly dark in the afternoons C3 or either in the mornings C4), and dark days (C5). These five cycles represent 42%, 14%, 18%, 16%, and 10%, respectively of the 4017 days analyzed. Cloud type percentages were computed to understand the sky conditions associated with the five types of diurnal cycles. The illuminated days C1 are mainly associated with clear sky conditions or sky covered by transparent and semi-transparent clouds (cirrus and alto stratus). The very dark days C5 are associated with a large cloudiness composed of low and high opaque clouds mainly without any marked peak in the occurrence of these clouds during the diurnal cycle. Accordingly, the DJF bright season is associated with a high percentage of very bright days (C1, 70%) and the JJAS dark season with a high percentage of obscure days (C4 and C5, 50%).

Following these results, we can conclude that Douala has a solar resource large enough to encourage deeper investigations in solar energy potential exploitations since it appears as a promising resource to meet energy needs. For a complete solar climatology in Douala, it is important, for a further research, to quantify the effects of cloudiness and aerosols on the solar irradiation balance and analyze the response of the climate in the whole region.