3.1. Spatiotemporal distribution of soil moisture and rainfall
In order to assess the dynamism in relationship and response characteristics of soil moisture (SM) and rainfall (RR), the spatiotemporal distribution of SM and RR over the study area is discussed. Figures 2a and 2b display a climatological 30 years daily averages of ground-based RR gauge measurements and the NASA - POWER satellite-derived SM respectively, spanning 1990–2019 at the 22 synoptic stations in Fig. 1, distributed across the four main climatic zones. The advantage of Fig. 2 is that it displays simultaneously the space and time distributions. It is observed that generally on a spatial scale SM is highest over the southern half where RR is bimodal and lowest over the northern half where RR is mono-modal. On a temporal scale over the southern half, SM peaks from late May - July (~ 150–200th Julian days) immediately following the MAM major RR season (~ 60–150th Julian days), after which another very faint peak (~ 275–330th Julian days) follows the minor SON RR season (~ 250–300th Julian days), and eventually diminishes during the DJF dry season (~ 334–90th Julian days) when RR is lowest countrywide. Over the northern half, SM is generally low, showing a faint peak (~ 250–300th Julian days) following the long-range mono-modal RR season (~ 133–300th Julian days). These concurrent patterns establish the axiom that elevated RR is associated with elevated SM levels. However, there is high variability and significant contrasting SM - RR relationships across the varied ecoregions, creating room for other elements contributing to SM levels such as topography as noted in other studies [22, 23].
Firstly, over the southern half, a high daily mean SM > 0.70 ± 0.07 is mainly associated with coastal stations categorized into Forest climate towards the west-coast and Coastal Savannah-type climate towards the east-coast (see Fig. 1). Despite the high annual cumulative RR ~ 1500 mm over the west-coast relative to the lower RR ~ 700 mm over the east-coast [29], SM levels along the entire coastal stretch are higher. This can be attributed to two obvious reasons viz., elevated RR (relating to the west-coast) and the low-land topographic characteristics < 100 m (see Table 1) causing high seawater intrusion and percolation, and consequential strong weathering, high leaching, and low pH [24, 30]. However, over the northern half, stations show relatively lower daily mean SM between 0.46–0.56 ± 0.06, with a mono-modal RR pattern of an annual total amount ranging between 962.4–1198.9 ± 85.8 mm. The low zonal and seasonal RR/ SM is expected to cause quick responses in SM - RR relationship.
The northern half - mainly the Savannah type region - exhibits a more significant seasonal change in root zone soil moisture has been confirmed from the simulation results by [31] when assessing the SM and RR relationship in order to calibrate the Joint UK Land Environment Simulator (JULES) land surface model across the study area using satellite-based data from TAMSAT-ALERT (Tropical Applications of Meteorology using SATellite data and ground-based observations AgricuLtural dEcision suppoRT). For a detailed statistic of the SM and RR datasets, Fig. 3 shows a boxplot analysis of the SM/ RR pairwise datasets considering zonal and seasonal distribution scales. Figures 3a and 3b show the monthly total RR distribution across the climatic zones and seasons respectively. Firstly, the climatological average monthly total RR is unevenly distributed within and across zones and seasons, following the order 125.2 ± 74.0 mm > 105.6 ± 69.0 mm > 103.4 ± 94.0 mm > 65.8 ± 55.0 mm for the Forest, Transition, Savannah, and Coastal climate zones respectively, with a seasonal variation in mean total RR in the order 153.6 ± 48.0 mm > 129.3 ± 48.0 mm > 112.3 ± 58.0 mm > 21.1 ± 8.0 mm for the JJA, MAM, SON, and DJF climatic seasons respectively. However invariably, contrary to the zonal RR distribution, Fig. 3c shows a monthly maximum zonal SM distribution following the order 0.79 ± 0.09 > 0.73 ± 0.07 > 0.66 ± 0.06 > 0.65 ± 0.06 for the Coastal, Forest, Transition, and Savannah climate zones respectively, while Fig. 3d shows the seasonal monthly maximum SM following the order 0.76 ± 0.06 > 0.69 ± 0.04 > 0.67 ± 0.06 > 0.6 ± 0.04 for the JJA, SON, MAM, and DJF climatic seasons respectively.
As seen, the SM - RR pairwise datasets show coherency in seasonal variation, and non-coherency in zonal variation, suggestive of a rather complicated SM - RR interplay. This apparent discrepancy can be resolved by considering that, a team of factors including prevailing climate [32], ecological conditions [33], topography [22, 23], and geography [34], as well as soil types and slopes, are contributing to the SM - RR interplay over the region. For example, [35] showed from observational experiments conducted over contrasting ecosystems in the loess hilly region of China that, highly forested areas with deep-rooted plants caused pronounced soil desiccation, while shrublands retained high soil water holding capacity, coupled with a strong regression slope between SM and RR. The study thus recommended shrubland types for land reclaim and vegetation restoration projects as they are more likely to protect the land from soil erosion. [36], also showed that soil water holding capacity decreased along inclined semi-arid regions of Inner Mongolia from the upper to lower areas after a rain event over the area, and depended inversely on the initial soil water.
However, the trend is shown in Figs. 3b and 3d suggests a strong positive linkage in the SM - RR relationship over the region, while Figs. 3a and 3c bring to bear the geographical and topographical effect in the SM - RR interplay. For example, the Coastal zone with the lowest RR over the region (Fig. 3a) would be expected to have low SM, while the Forest and Savannah zones have the highest SM, which is not entirely the case. Contrarily, the Forest zone is only second to the Coastal Zone with the Savannah zone showing the lowest zonal SM, accounting for the relative dryness towards the Sahel region shown in the Soil Moisture and Ocean Salinity (SMOS) and Aqua satellites [37].
3.2. Soil moisture and rainfall relationship considering rainfall onset
Figure 4 is a 30 years weekly averaged climatological time series for SM and RR over the four climatic zones, showing the mean rainfall onset dates as 19th, 12th, 13th, and 16th weeks across the Savannah, Transitional, Forest, and Coastal zones respectively. The onsets dates were determined based on the AGRHYMET’s definition as the first rainy 10 days with at least a cumulative rainfall value of 25 mm, followed by less than 2–3 weeks dry spell [38]. It is worth mentioning that, these dates are long-term averages of station RR datasets over the zones and change by ± 23 days for each year. The objective here is to assess how onset dates feature in the SM - RR relationship, and thus attempt to establish a general theoretical reference for application in agriculture as soil moisture is a key consideration factor in fixing planting dates [39].
Firstly, it is interesting to note from Fig. 4 that, SM follows closely the ITCZ - driven RR pattern for all the zones, namely, a bimodal over the southern half Coastal and Forest zones (Figs. 4c and 4d), intermediary over the Transitional zone (Fig. 4b), and mono-modal over the northern Savannah zone (Fig. 4a). However, the difference in SM and RR intensity changes during the onset week for each zone is remarkable. Table 2 summarizes the percentage changes in RR and SM during the onset week for each zone extracted from Fig. 4. The changes were determined with reference to the commencement of the year. The results show that, with the exception of the Coastal zone where an 86% rise in RR is commensurate with a 14% rise in SM, the Savannah and Forest zones show significantly low SM responses to RR, while the Transition zone shows a slight negative SM response of 0.01–99% increment in RR by the start of the season.
Table 2
Percentage change in RR and SM from the first week to onset week of the year for each climate zone
Zone
|
Mean onset
|
Percent change in RR during onset (%)
|
Percent change in SM during onset (%)
|
Savannah
|
19
|
98
|
0.04
|
Transition
|
12
|
99
|
-0.01
|
Forest
|
13
|
85
|
0.04
|
Coastal
|
16
|
86
|
14.00
|
The slightly negative and low responses for the stations in the Transitional, Savannah, and Forest zones respectively, may be attributed to SM deficits as a result of the high atmospheric evaporative power during the dry Harmattan season which commensurate with the boreal winter which climatologically extends from late November through to February [40–42].
The Harmattan season is characterized by low RR and intense atmospheric dryness along with strong dusty north-easterly winds developed from a continental scale pressure difference between the sub-tropical subsidence zone and the ITCZ, looming southward and lingering along the Gulf of Guinea [42].
Notably, during this period is the low atmospheric moisture convective activity and the consequential rainfall suppression which causes intense soil moisture stress due to high atmospheric evaporative power, strong dry winds, and low relative humidity [43]. Such environmental harsh conditions result in extremely low SM over the sub-region. The implication of this is that a large amount of RR is needed to satisfy the SM requirements to sufficiently support plant growth over the northern regions, and hence the amount of SM indicative of the start of rainfall season for planting time of various crops across the zones cannot coincide. For example, [39] found a strong correlation between planting dates in April/ May and planting area with SM and RR over regions in northern Cambodia, and farmers thus resorted to earlier planting dates during wetter conditions.
3.3. Statistical assessment of the relationship between soil moisture and rainfall
Based on the observation data presented in Fig. 4, the relation between soil moisture (SM) and rainfall (RR) is expected to differ for each station. In order to quantify this relationship, Fig. 5 shows the results of station-by-station Pearson’s correlation coefficients (r). Generally, from Fig. 5a, there is an all-positive correlation ranging between 0.38–0.82 ± 0.10, which peaks for the Savannah and Coastal stations at significant p-value < 0.05, while falling for the Transition and Forest stations. Following the example of [11] in the assessment of SM - RR relationship on a global scale using satellite data, the categorization of very strong positive (VSP = 0.8-1.0), strong positive (SP = 0.60–0.79), moderate positive (MP = 0.40–0.59), and weak positive (WP = 0.20–0.39) is used. One station shows a very strong positive SM – RR relationship, 14 stations show a strong positive SM-RR relationship, 6 stations show a moderate positive SM-RR relationship, while one more station is weak positive.
The variations in correlation coefficients are connected with the diverse and contrasting ecosystem conditions where stations are situated. In this regard, it is worth mentioning that, the results agree with the report by [11] using NASA’s global SM and RR datasets, showing strong correlation coefficients over low vegetated regions and weaker correlation over highly forested regions. The weak correlations associated predominantly with the Forest climate stations can be attributed to the physical phenomenon of hydraulic redistribution where deep-rooted forested plants draw water from deep soil layers to surface root zone regions (100 cm) as a biological adaptation for dry season survival. [44] and [45] have described this condition for deeply rooted forested plants in the Congo Basin and Amazon respectively. Again, from Fig. 6, the very weak to moderate correlations are associated with the forested stations closer to the Volta River basin. Sehler [11] revealed that the geophysical process of river displacement - whereby rainfall over upper stream causes the lower stream end to expand, not necessarily due to RR, creates a condition of inverse SM - RR interplay, for areas nearby river systems. Conway [46] has illustrated this phenomenon in Eastern Africa, where heavy downpours over the Nile River are transported northward to Egypt and Sudan, causing elevated soil moisture not necessitated by rains in that area, thus yielding weak to negative SM - RR correlations.
Furthermore, from Fig. 6, the highest Pearson’s correlation coefficient of 0.81 is associated with the Savannah climate, while all other stations show strong positive correlations > 0.61. The land cover type, geography, and topographic features are expected to allow for such high positive correlations as reported by [11] for other regions with similar climates and land cover. In practice, [41] realized this from soil moisture measurements at eddy covariance stations in Sudanian Savannah ecosystems across the West Africa terrain. They reported high variability and quick response of SM to RR, mainly due to the variability in vapour pressure deficit over the region. Overall, whereas one Savannah stations show a very strong positive SM - RR relationship, and one Forest station shows a week positive, the larger number of stations across the study area shows moderate to strong positive correlation.