3.1 Interannual Variability in the mean Temperature
Figure 2 compares the time series of interannual variation between CRU and CPC over the study period (1981–2020). Despite the differences in the magnitudes of trends, both CPC ( 0.02557 ± 0.0036) and CRU (0.01637 ± 0.00302) both show increasing mean temperature trends from 1981 to 2020 and elucidate significant cooling and warming in the first and last two decades, respectively. The CRU had anomalies between − 0.6 and 0.3 oC, while CPC indicated anomalies between − 0.7 and 0.3 oC. Their temporal trend appears similar, with higher values from the CPC than the CRU. Anomalies increased from − 0.1 to 0.7o C for CRU and CPC. Overall, the highest trends are found over the last two decades for the two data sets (0.6 oC, 0.7 oC), CRU, and CPC, respectively. The lowest trends were experienced in the first two decades (-0.6 oC, -0.7 oC). The results confirm an increase in mean temperature over the study area during the study period.
3.2 Decadal Anomalies
Figure 3 shows the spatial decadal anomalies of the mean temperature from 1981 to 2020, as indicated by the CPC dataset. The anomalies were computed using the entire study period to avoid multi-decadal natural climate fluctuations. The initial two decades experienced a relatively cooler average temperature anomaly between − 0.9 to 0 oC in almost the entire area of the study region, with spatial variation across the country (Fig. 3a, b). The CRU dataset also experienced a similar spatial pattern, with the first two decades experiencing cooler average temperature anomalies (0.6 to 0) shown in (S1). Interestingly, the country experienced a cooler temperature anomaly in the northern region (-0.9 to -0.2) for the initial two decades (Figs. 3a,b) than in the southern sector. Generally, the first two decades give an impression of a cooling pattern in the anomalies for the entire study domain.
Contrarily, the last two decades (2011–2020) depict a spatial warming pattern between (0-0.9 o C) above the mean for both CPC and CRU data sets shown in (Figs. 3c,d) and ( S1 c,d). Though the last decades give a warming impression, there are significant spatial variations. For example, the forested zone on the southwestern and the west-central side experienced relatively cooler anomalies (0 to 0.1o C), especially from (1991 to 2000) for both data sets (Fig. 3c and S1, c). However, from 2011 to 2020, the whole country experienced warm anomalies ranging from (0.2 to 0.9 o C) geographically with fluctuations. Overall, there have been increasing warming temperature patterns with different spatial variations. The last decade (2011–2020) is the warmest (Fig. 3d). The results from the two data sets confirmed the warming consistency across the study area, especially the spatial distinction between the first two and the last two decades.
3.3 Probability distribution of mean temperature and breakpoint change detection
Decadal climatological mean anomalies were conducted to examine which decades were warmer than the other. As per the results, it was observed that the last two decades (2001–2020) were warmer than the previous decades (1981–2000), as indicated in Fig. 3. This prompted the need to investigate if the average temperature over the case study area might have experienced a clear shift in the mean. The change point detection method indicated a change point year of 1997 to be more influential. The probability distribution of mean temperature anomaly during the RP and CP averaged over the entire country is shown in Fig. 4. Figure 4 shows the three-moment statistics for the position, scale, and shape parameters: mean, variance, and skewness, respectively. The data demonstrates a 0.41o C shift in the mean, increasing high temperatures in the CP (1998–2020) compared to the RP (1981–1997). Since there has been a substantial shift in the Probability Density Function (PDF) of the mean temperatures in the past decades, analyzing the influence of land and atmospheric drivers over the study area will help scientists understand the relationship between these variables at the local scale.
3.4 Influence of climatic factors on the observed near-surface air temperature
To evaluate the possible influences of land and atmospheric factors on the changes in NST, the present study used the climatological mean differences of both the annual and seasons between the climatic drivers before and after the breakpoint. Furthermore, the regression technique was used to assess the relationship between atmospheric conditions and NST. The spatial distributions of the climatological mean difference between the current period (1998–2020) and the recent past (1981–1997) in annual DSWRF, HGT, LST, RHUM, TCDC, ULWRF, and VWND during the 1981–2020 period are presented in Fig. 5a-g. The results show a significant positive trajectory between CP and RP in the HGT, LST, RHUM, and TCDC. The positive climatological difference between CP and RP of HGT increases latitudinally from the coastal area through to the northern part of the country from 7.55 to 8.15 (hPa). This indicates the positive difference in the HGT at the upper level leads to an upsurge in near-surface temperature. The phenomena then establish a low-pressure center, causing an increase in near-surface air temperature. A similar pattern could be seen in relative humidity, increasing from 0 to 10 above the mean average from the coastal belt to the northern belt.
Generally, the climatological mean difference between CP and RP experienced an increasing land surface temperature in the central belt and more pronounced in the northern belt, with differences ranging from 0.50 to 1o C. The positive anomalies of land surface temperature (0.35-1 o C) indicate that they have experienced an upward increase in the current period (1998–2020) compared to recent past years (1981–1997). Generally, the northern belt experienced an increase in LST between (0.70-1 o C), while a relatively cooler LST was experienced at the Volta Lake basin (0.30 o C) and the southwestern belt of the country (0.35 oC). However, this increase was not evenly distributed across the country. For example, there is strong evidence of a reduction in the magnitude of NST in the current period (0.34–0.4 oC) in the southwestern part of the country compared to the recent past period. In contrast, the south-eastern and northwestern parts of the country observed an increase in NST in the current period (0.4 to 0.5 oC) compared to the recent past.
In addition, the results prove that the climatological mean difference of DSWRF, ULWRF, and VWND between CP and RP has reduced throughout the country. The reduction in downward shortwave radiation indicates that less solar radiation could penetrate the atmosphere during the current period (CP). This could be associated with the positive anomalies of total cloud cover experienced. The increase in cloud cover absorbs more of the downward shortwave radiation from the sun. Inversely, the decrease is not spatially distributed across the country, though the southwestern part experienced a reduction of downward shortwave radiation in the current period. There was, however, an increasing pattern observed during the recent past period. A similar tendency was also experienced in the upward longwave radiation and vertical wind velocity.
Moreover, there is also a reduction in the climatological mean of the upward longwave radiation flux (ULWRF) between CP and RP in the study region. The reduction of the ULWRF anomalies has a good relationship with the minimization of the DSWRF. A reduction in solar radiation will lead to a reduction in longwave radiation flux.
3.5 Seasonal variations of atmospheric drivers associated with near-surface air temperature
Figures 6 and 7 are spatial plots of the seasonal climatological mean difference between CP and RP of the atmospheric drivers for FMAM and JJAS seasons over the study domain. As the country warms overall, average temperatures increase throughout the year (see Figs. 2 and 3), but the increases may be higher in certain seasons than others. Therefore, there was a need to examine the seasonal variations of the various atmospheric drivers. The results elucidate significant positive changes in the mean surface temperature (NST) of the FMAM season across the country (Fig. 6a), especially in the northwestern part. The entire country experienced a warming pattern between (0.34–0.58 o C) with a spatial climatological increase from southwestern to northwestern. The climatological difference between the CP and RP indicates a higher warming tendency of the FMAM season, more pronounced in the northern belt. This indicates FMAM season has experienced higher warming in the CP at the northern belt (0.50–0.58 o C) than in the country's southern belt (0.34–0.44 o C).
In contrast, the JJAS season experienced positive changes in the climatological mean differences between CP and RP (Fig. 7a). Conversely, the difference is more pronounced in the southern belt (0.34–0.42 o C) than in the country's northern belt (0.26–0.32 o C). In addition, the climatological mean difference of surface downward shortwave radiation flux shows a significant positive increase in the southern belt between (2 to 12 Wm− 2) with a negative declension in the northern belt (-2 to -12 Wm− 2). The mean climatological difference in geopotential height for both seasons was positive (Figs. 6a and 7a). The two seasons showed a contrasting spatial pattern for the northern and southern belts of the country. HGT decreased spatially towards the south in the FMAM season while amplifying towards the southern belt in the JJAS season. Our results indicate the seasonality of geopotential height within the study area. Compared with other parameters, the land surface temperature experienced a consistent increase in the climatological mean difference between CP and RP in both seasons, coupled with spatial variations. The climatological mean difference of 0.4 to 2.0 oC was experienced in the FMAM season, increasing spatially from the southwestern and spreading towards the northern belt. The Lake Volta basin experienced the lowest annual mean difference of around 0.3 oC. A similar pattern was observed for the JJAS season, which experienced a mean climatological difference between 0.1 and 0.7o C. The differences in the seasons are expected because FMAM is dry with high temperatures, while JJAS is a wet season with lower temperatures.
Figure 8 displays how the various atmospheric drivers correlate with the temperature anomalies of the RP (blue dots) and CP (red dots) periods of the country. To understand the relationship between the atmospheric drivers and temperature, the standardized anomalies of atmospheric drivers are regressed against the temperature anomalies of the two periods. The geopotential height, relative humidity, land surface temperature, and sea surface temperatures are significantly correlated (0.67,0.59,0.81, and 0.52), respectively, with the upward pattern of temperature anomalies, as shown in Fig. 8. The implication of the results indicates an increase in HGT, RHUM, LST, and SST are associated with an increase in the mean temperature over the study domain. A significant correlation of -0.53 exists between downward shortwave radiation and temperature increase (Fig. 8). DSWRF increased from 1981 to 1997(RP) while the average temperature reduced. Over the current period, the inverse was witnessed, where temperature amplifies steadily with DSWRF decreasing. Also, a decrease in upward longwave radiation over the current period is associated with temperature amplification; a correlation coefficient of -0.43 exists between ULWRF and temperature.
Figures 9 and 10 display the seasonal relationship between atmospheric parameters and the mean temperature for FMAM and JJAS seasons. The correlation coefficient between downward surface radiation and the temperature was negative between the two periods of RP and CP for both seasons. A negative correlation coefficient of -0.23 and − 0.37 existed between DSWRF and NST for the two seasons. It implies the intensity of the downward surface radiation is reduced in the current period for both seasons. There was a strong correlation of 0.74 and 0.72 between geopotential height and near-surface temperature for both seasons. Near-surface temperature and HGT have increased in the current period compared to the recent past period for both FMAM and JJAS seasons. Another weak positive correlation of 0.47 and 0.15 existed between relative humidity and NST for FMAM and JJAS seasons. The correlation between total cloud cover and the near-surface temperature was negative in the dry season (FMAM) (-0.05) but positive (0.36) for the wet season (JJAS). Another weak positive ( 0.13 ) and negative (-0.35) correlation existed for FMAM and JJAS seasons for the upward longwave radiation. Vertical wind velocity correlated negatively with near-surface air temperature for both seasons, with a correlation of -0.08 and − 0.37 for FMAM and JJAS seasons, respectively. There was a strong positive correlation between land surface temperature and near-surface air temperature for both seasons. A correlation of 0.82 and 0.82 existed for both seasons. Also noteworthy are the correlations between sea surface temperature and near-surface air temperature. There existed a positive correlation of 0.40 and 0.33 for both FMAM and JJAS seasons. Most of the correlation coefficients were statistically significant (p < 0.05).
3.6 Changes between the recent past and the current period
Figure 11 compares the interannual changes of the various atmospheric drivers and NST between the recent past (RP) and current period (CP) from 1981 to 2020. The figures show the various drivers' intensity before and after the breakpoint. As shown in the figure, there are numerous variations in the intensity of the variables between the two periods. For example, NST, HGT, LST, RHUM, and SST experienced relatively lower intensity in the past but a higher intensity during the current period. This is confirmed by their various interannual time series (S2). In contrast, other variables like DSWRF, ULWRF, and VWND experienced a higher intensity during the recent past but a decreased intensity during the current period, as depicted by Figures (11 and S2). Turning now into the seasons, Figs. 12 and 13 show the interannual seasonal intensity of the various atmospheric drivers and NST for FMAM and JJAS seasons.
In an almost similar pattern to the annual figures, the two seasons experienced variations in the intensity of the NST and the other atmospheric variables. For example, in both seasons, NST varied in lower intensity for RP with little increase in intensity in 1983 and 1987. However, the CP showed a relatively consistent higher intensity (Figs. 12 and 13). Other atmospheric parameters like HGT, LST, RHUM, SST, and TCDC experienced lower intensity during their recent past but a high increase in intensity during the current period for both seasons. It is worth noticing that DSWRF, ULWRF, and VWND experienced relatively higher intensity in the past but a lower intensity in the current period.