3.4 Influence of meteorological parameters on ETref
The wavelet transforms coherence (WTC) between meteorological parameters and ETref in Bangladesh are outlined in Fig 7. The small arrows found in Fig. 7a exhibited the direction of correlation between sunshine duration (SD) and annual ETref where the colored bar on the right top side denotes the strength of this correlation. The rightward movement of arrows inside the contour represented a positive correlation between SD and ETref (in-phase) in the time-frequency domains of 0-4 and 4-5 years, during the 1998-1994 and 2006-2009 periods. Black cone shape was observed in the top left side which was called “cone of influence” and means the confidence level and significance edge effects in its margin.
Fig. 7b depicts that the small arrows towards the right side, indicating a non-significant positive correlation (in-phase) between wins speed (WS) and the ETref in the time-frequency band of 1-4 year to onwards during 1990-1994 of the observation periods. Non-significant edge shaped effects in the left side top corner are also detected in the same frequency band before 1995s.
Fig. 7c reveals that the small arrows towards the right side, implying a significant positive correlation (in-phase) between mean temperature and the ETref in the time-frequency band of 3-5 years during the period 2006-2011. The relative humidity (RD) and the ETref are anti-phase correlation as represented by the arrows towards the left side in the circle donated on the left side top corner of the Fig. 7d and inside the cone of influences. Red color contour is characterized by the colored bar on the left side which suggesting a strong negative correlation in the time-frequency band of 0.5-3.5 years and onwards during 1987-1994 of the observation periods. Fig. 7e shows that left side arrows, indicating a negative significant correlation (anti-phase) between rainfall and the ETref in the frequency band of 0-3.5 years to onwards during 1989-1997 of the observed periods.
Table 3 shows the partial correlation results for ETref and meteorological parameters in Bangladesh. The amount of contribution of meteorological factors differs from the temporal changes in the time series to variation ETref in Bangladesh. The largest significant partial correlation matrix value was observed between MT and ETref (r= 0.797 at the annual scale; r= 0.734 at the seasonal scale for winter, p<0.01). For summer, spring, and winter season 0.987, 0.930, and 0.407 values were obtained respectively. Both annual and winter seasonal ETref variation was strongly influenced by mean temperature except for other seasons.
The second highest partial correlation was observed among SD and ETref, (r=0.654 at the annual scale, p,0.01, r=0.407 for post-monsoon, p>0.05, and 0.591 for winter, p<0.01). Except for the pre-monsoon and monsoon seasons, the seasonal contribution of SD was the main factor in ETref variation in Bangladesh. By contrast, the RH is negatively dominant climatological factors that had an impact on ETref variation during the annual and seasonal scales (r=-0.677 at the annual scale, p<0.01, r=-0.376 for pre-monsoon, r=-0.313 for winter, p<0.05). WS had a little contribution to the variation of ETref during the annual and seasonal scales except for winter. However, no significant relationship was detected between rainfall and ETref during the observation period of 1980–2017 in Bangladesh except for winter. So, it can be said that MT for the annual, and winter and SD for annual, post-monsoon and winter were mostly prominent meteorological parameters in the ETref variations of the country. Although WS had a little contribution to the variation of ETref during the annual and seasonal scales except for winter. Based on the outcomes of partial correlation matrix results, overall MT, RH, and SD are significant meteorological factors in the ETref variations on annual and seasonal periods during the study period. WS and rainfall only influenced the variations in ETref during winter. However, MT, RH, and SD play a major influencer in changing ETref whereas WS and rainfall act as a minor role in the variation of ETref in Bangladesh during the period of 1980-2017.
Table 4 outlines the standardized coefficient values of the five meteorological parameters for variations in ETref. The MT, SD, and RH were the greatest dominant climatological factors that influenced the variations in ETref on an annual basis, with standardized coefficient values of 0.550, 0.472, and -0.384 respectively. Seasonally, RH declined in winter when MT boosted and both factors were dominating and the coefficient values were −0.488 and 0.488, respectively. The RH was the most prominent parameter on ETref variation in the pre-monsoon and monsoon, with the elevated standardized coefficient values of -0.418 and -0.331, respectively. Similarly, the SD was the key influencer on ETref changes during the post-monsoon. Surprisingly, in the winter, RH was the greatest forceful parameter, with a high coefficient value of 0.670. At the same time, the MT was the main factor for ETref variation during the winter, with a coefficient value of 0.593.
The decline in ETref on the annual time scale and during the post-monsoon and winter seasons was affected by RH, MT, and SD. Winter ETref showed a significantly declining trend at the seasonal scale. WS exhibited no alteration on the yearly or seasonal scale. On the whole, ETref changes were affected by the coupling effects of considerably rising MT, and SD and declined RH in Bangladesh. WS is a little contributor to the ETref variation in Bangladesh. Rainfall had no effect on ETref variation as a meteorological parameter, but it plays as a predominant role for drought events in western Bangladesh.