Based on a Pearson moment correlation coefficient “r” analysis previously conducted on all the years in relation to the 6 engaged climatic factors, the obtained order of decreasing “r” values was: minimum relative humidity > maximum relative humidity > rainfall > minimum temperature > maximum temperature > sunshine hours ( “r” value using both mean and total “r” values for comparison) ( +0.41 and +2.43 > +0.38 and +2.29 > + 0.30 and +1.79 > + 0.14 and +0.87 > -0.37 and -2.19 > -0.38 and -2.25 for the “r” values of each of the respectively listed climatic factors above, obtained from the study area) (Figures 1 and 2; Table 5).
This implies that relative humidity, rainfall, minimum temperature and maximum temperature, were the climatic factors (coded 4ecf) that were observed to be positively correlated (with a positive relationship) with values for clinical episodes of malaria for the period of data gathered. As such, they appeared to be the major defining factors for clues to periods of high intensity of malaria transmission in the study area.
From Table 1, involving part of the predictive model whose data were not retrieved, the order of increasing correlation with the climatic factors is:
2002 (r+ve) < 2004 (r+ve) < 2003 (r+ve) – in terms of all the r+ve correlation comparisons [ 0.589 ± 0.152 and 2.354 (P1) < 0.532 ± 0.177 and 2.128 (P2) < 0.305 ± 0.345 and 1.218 (P3) respectively for mean and total “r” values ], and 2002 (r-ve) < 2004 (r-ve) < 2003 (r-ve) – in terms of the “r-ve” correlation comparison [ 0.220 ± 0.176 and 0.877 (N1) < 0.170 ± 0.176 and 0.679 (N2) < -0.062 ± 0.340 and -0.249 (N3) respectively for the mean and total “r” values in the “r-ve” analysis ]. Thus, the predictive range of correlation coefficient “r” values for years 2002, 2003 and 2004 all enumerated beneath Table 2 indicates only year 2003 to have the highest possibility or likelihood of a negative correlation ( predictive range of r = - 0.062 to +0.305) with the six climatic factors engaged.
The degree of fitting-in (match) was more noticed when all engaged climatic factors (6cf) were used for the model than when only the ascertained and previously determined actual climatic factors (4ecf in this study) that influenced the periods of high intensity transmission of the disease in the study area. The wider the pool of climatic factors engaged in the model, the more the level of fitting-in or accuracy. Hence, the actual periods of high intensity transmission defining climatic factors were used as a check within the simulation and to create a broader picture of understanding of the relationship amongst the varying groups of factors (Table 2).
The statistical computations indicated high intensity of clinical episodes to correlate (correspond) with rise for the climatic factors, and low intensities for lowered levels of most of the climatic factors for years 2002 and 2004, as they both recorded positive ranges of correlation “r” values between clinical episode and climatic factors. This was used to predict periods of the year with high intensity of clinical episodes of malaria as our simulated prediction.
The four climatic factors we found to determine the period of high intensity of transmission which have linkages with the provisions or enabling of a humid (wet) condition in the environment (minimum and maximum relative humidity, rainfall and minimum temperature),. They enhance the provision of suitable conditions for the vector to breed and participate in the process of transmission through infection and re-infection of unsusceptible and susceptible humans recorded.
Similar finding from Mariki et al (2022) opined that malaria transmission depends on climatic conditions that may affect the number and survival of mosquitoes, such as rainfall patterns, temperature and humidity. This study further observed that symptoms were important in the diagnosis of malaria in Kilimanjaro province of Tanzania in the months of rainy session or just after the rainy session, during which they were significant in malaria transmission.
In a related finding from the study of Abdelrazec and Gumel (2017), it was reported that observed peak mosquito abundance for temperature and rainfall values recorded for months of July and August in Peel region of Ontario in Canada. These are in the season of summer in Canadian which is called the rainy season in the hot tropical and sub-tropical climates where malaria is a more common disease and endemic here. A related finding from a study by David et el (2016) revealed that Peak malaria prevalence was recorded in the month of September and followed by June (15.58% and 14.02% respectively) while the value was least for March followed by November (2.21% and 3.69% respectively).
Singh et al (2019) used a dynamic model (VECTR) for malaria transmission that accounts for the influence of population and climatic conditions to investigate malaria transmission dynamics for a highly endemic region in India (State of Odisha) to do numerical simulations for years 2000-2013 and found out that temperature, adult mosquito population and infective biting rates have increased over this period and malaria vector abundance higher during the summer monsoon season. The peak malaria transmission occurred when the monthly mean temperature is in the range of 28-290C and monthly rainfall accumulation in the range of 200-360mm.
According to Stuckey et al (2013), evaluating the effectiveness of malaria control interventions on the basis of their impact on transmission is increasingly important as countries move from malaria control to pre-eradication programs. They added that mathematical modeling can examine relationships between malaria indicators, allowing translation of measured data into measures of transmission, as simulations show nature of statistical correlation, allowing direct comparisons of malaria transmission using data collected across a range of transmission intensities and seasonal patterns.
Test-run of predictive analysis as simulated
The test-run was necessary in order to check the correctness of the predictions on periods of high malaria intensity made for years 2002, 2003 and 2004, and the capability (effectiveness) of the model. It was more of a validation of previous predicted results and from the model. Basically, the procedures in the sequence of operations earlier outlined were followed up to the stage of estimation of the range in the obtained results. Thus, comparison of predicted ranges of “r’ values and the actual field values in the case of the test-run could be conducted unlike in the earlier predictions done for years 2002, 2003 and 2004. This is because there were actual figures from the field to compare with and help validate results for consistency.
Predictive correlation ‘r’ value ranges for 1998 and 2001 are given below:
(I) for 1998 are as follows (Table 3):
(+) 0.0985 to (+) 0.1879 (for all 6cf)
(+) 0.4043 to (+) 0.5288 ± 0.2178 (for only 4ecf)
Actual values of correlation coefficient ‘r’ values from the incidence obtained from field studies:
(+) 0.184 ± 0-193 (for all 6cf)
(+) 0.318 ± 0.044 (for 4ecf).
(II) For 2001 are as follows (Table 4):
(-) 0.2975 to (-) 0.1364 (using all 6cf)
(-) 0.1379 to (+) 0-1329 (using only 4ecf).
Actual values of correlation coefficient ‘r’ values for the year 2001 from the incidence obtained from field data:
(-) 0.313 ± 0-235 (using all 6cf)
(-) 0-221 ± 0-242 (using 4ecf).
Note: The incidence for year 2005 was used for the r+ve and r-ve evaluations of incidence in relationship with climatic factors of year 2001 by Person correlation coefficient analysis.
( ) Brackets have deliberately been for easy understanding, clarity positive and negative correlative values with respect to the two seasons of the year.
The year 2001 incidence results showed least correlation coefficient ‘r’ value by inspection of results obtained (on Table 5), next to that of year 2005 (minimum ‘r’ value for all 6 years studied) form the actual values. Hence, it (year 2001) was utilized for the evaluations.
Test run for 1998 (Table 3)
A comparison of the actual correlation coefficient ‘r’ values (on Table5 ) and the predictive correlation ‘r’ values for 1998 of (+0.184 ± 0-.93) on Tables 2 and 3 (using all 6climatic factors - 6cf) is in line with the results from the model.
Firstly, both results from the actual values shown on Table 5 (for 1998) and the predictive one on Table 2, both indicate a positive correlation between malaria incidence and the six climatic factors. This shows exactness in this area of prediction (100%).
Secondly, the actual (+) 0.184 ± 0.193 ‘r’ values (for all 6cf) is well within the predictive (+) 0.0985 to (+) 0.1879 ‘r’ range of values (for all 6cf). The former (actual values) fits in 100% within the latter’s predictive range.
A predictive analysis of the years 2002, ’03 and ’04 (on Table 2) in which data on incidence were not obtained, indicated a most likely tendency towards high intensity of transmission that was related to the seasons, with respect to the climatic factors influencing it and a yearly non-perennial high intensities in transmission for the years 2002 and 2004 with positive ‘r’ value ranges, while year 2003 predictably had the most likely tendency for a non-season related high intensity transmission (range of ‘r’ = -0.062 to +0.305) amongst the three years. This formed the basis for the “developed mathematical predictive model”, which was also test-run as a check for its predictive outcome of which was alright and in line with actual field ‘r’ values.
The capacity of climate based simulations to predict periods of high intensity of malaria transmission could be of benefit to control measures in areas of policy formulation, planning and proper execution of such control measures in such endemic and bio-geographical area; with added cost-benefit attributes in addition to possibly reducing incidence (perhaps to insignificant levels in some or most areas), and associated health and economic burdens..
The predictions from the model on Table 3 indicate a positive correlation and association of malaria incidence with climatic factors (season related high intensity of malaria transmission). This is exactly what was previously obtained from the actual results on Table 2.
In view of the foregoing, the predictive simulation for malaria intensity was able to predict periods of high intensity of transmission correctly. In terms of the precision of the exact value (which is a greater or more demanding task, of- the extent of association (correlation) between malaria incidence and climate, the capability is alright. The predictive range of correlation coefficient ‘r’ values was well within the range of actual ‘r’ values’ obtained from the field studies, which was used to test-run for the exact relationship existing between clinical episode of malaria and climatic factors probed for.
Test run for 2001 (Table 4)
A comparison of the actual correlation coefficient ‘r’ values for 2001 and the predictive one on Tables 2 and 5 of the constructed predictive model indicate the following:
Firstly, the results from the actual values shown on Table 5 (for year 2001) showed a negative correlation between clinical episode and the 6cf combined. This indicates exactness in this area of prediction (100%).
Secondly, the actual (-) 0.313 ± 0.235 ‘r’ values for all 6cf is well within the predictive (-) 0.2697 to (-) 0.1364 ‘r’ value range for all 6cf (from Tables 4 and 5 respectively). When values are taken to one decimal place, it fits in 100% within the predictive range.
These deductions above indicate a negative correlation and association of clinical episode of malaria with the engaged climatic factors (non-seasonal regarding mean correlation with the engaged 6 climatic factors). A cross-check also, showed that this was exactly what had been previously obtained from the actual field data on Table 5 (graphically depicted in Figure 2).