Observed meteorological drought trends in Bangladesh identified with the Effective Drought Index (EDI)

Highlights

• All seasonal and annual drought severity except pre-monsoon is increasing.

• Moderate to severe drought is increasing more in the Barind tract than in the Teesta floodplain.

• Drought spell typically starts between March and May (±15 days) and ends with monsoonal rainfall.

• EDI and rice production have no strong linear relationship.

Abstract

Countries dependent on small-scale agriculture, such as Bangladesh, can be vulnerable to the effects of climate change and variability. Changes in the occurrence and severity of drought are an important part of this issue and form the subject of this paper. We examined the characteristics of meteorological drought occurrence and severity using the Effective Drought Index (EDI), including the drought events, drought chronology, onset and ending of drought, consecutive drought spells, drought frequency, intensity and severity, using North-Bengal of Bangladesh as a case study. The rainfall and temperature dataset of the Bangladesh Meteorological Department (BMD) for the study region throughout 1979–2018 is utilised. The trends of drought are detected by using the Mann-Kendall test and Sen Slope estimation. We evaluated the performance of EDI using the Standardized Precipitation Index (SPI), historical drought records and rice production. This study finds that seasonal and annual droughts have become more frequent over the period studied in all seasons except the pre-monsoon. In addition, the largest decrease in seasonal EDI is found in the monsoon, both in the Teesta floodplain and Barind tract regions. In the decades prior to the late 2000s, a drought spell typically starts between March and May (±15 days) and ends with the monsoonal rainfall in June/July. In the years since the late 2000s, monsoon and post-monsoon droughts spells have significantly increased. Overall, the peak intensity of droughts are higher in the Barind tract than in the Teesta floodplain, and the frequency and severity of moderate to severe drought are increasing significantly in the Barind tract. The drought frequency has increased by at least 10% in North Bengal of Bangladesh over the periods of 1979–2018. Though EDI is strongly correlated with the SPI index, our analysis shows that, surprisingly, rice production is actually decoupled from meteorological drought (as identified by the EDI and SPI). Hence, this research suggests that there are other significant influences on rice yield beyond meteorological drivers. This could include effects from differing irrigation infrastructure, technology and management strategies in the study regions. Challenges to agricultural production may be exacerbated in coming years, should the identified increasing meteorological drought trends continue.

Data Availability

The data that support the findings of this study are available on request from the corresponding author.

Introduction

Drought is a complex meteorological and sociological phenomenon (Spinoni et al., 2019, Wilhite and Glantz, 1985), often difficult to quantify and diagnose the start and endpoints (Akter and Rahman, 2012, Keka et al., 2012, Lowe et al., 2018, Murad and Islam, 2011, Oesting and Stein, 2018, Paul, 1995). It may occur simultaneously or sequentially (Mo, 2008), associated with multiple variables (Wilhite, 2005), which are interconnected, thus hard to distinguish (Hao and Singh, 2015). It is a creeping hazard that develops slowly and has a prolonged duration, and its occurrence can be very patchy geographically (Keka et al., 2012, Murad and Islam, 2011). It is the most widespread natural disaster, affecting many areas worldwide, while agriculture is the most susceptible sector (Wu et al., 2017). The impact of meteorological drought on agriculture is through the reduction of available agricultural water resources, which causes crop water stress and decreases in yield (Lu et al., 2017). An analysis conducted by Karim et al. (2012) indicates that rice production would decline by 33% in the major rice-growing areas of Bangladesh in 25–45 years due to a 14% increase in irrigation demand (Rahman et al., 2017, Rahman et al., 2012, Rimi et al., 2009). Although higher atmospheric CO₂ levels may have a beneficial impact due to the fertilisation effect (Lal et al., 2005, Samarakoon and Gifford, 1995), high temperatures during the flowering period affect the photosynthetic rate for C₃ and C₄ types of crops by increasing water demand and reducing the grain size and quality (Cruz et al., 2007, Hamim, 2005, Hijioka et al., 2014). The rice yield is projected to decline by 10% for each 1 °C rise in growing season minimum temperature during the dry period (Peng et al., 2004). The positive benefit of CO₂ for C₃ and C₄ types of crops is projected to be played out in the next 25–40 years (Easterling, 2005, Lal et al., 2005). Due to the adverse effects of climate change, suitable climates for plant growth and the number of suitable growing days are projected to decline (Mora et al., 2015).

Bangladesh is considered the most vulnerable country in the world due to its socio-economic conditions, geographical location and adverse impacts of climate change and climate variability (Akter and Rahman, 2012, Ali et al., 2019, Islam and Nursey-Bray, 2017, Shahin et al., 2014). The country is less resilient to cope with the effects of climate change because of its population density, small size, a fragile economy, developmental inequality, and low adaptive capacities (Naser, 2015). Agriculture is the primary source of livelihood, contributing 14.23% to the GDP and employing about 40.62% of the labour force (Finance Division, 2018). However, climate change is expected to affect agriculture significantly and decrease agricultural GDP by 3.1% each year (Delaporte and Maurel, 2018, World Bank. 2010. Economics of Adaptation to Climate Change. The World Bank, Washington, DC.), which will create pressure on the lives and livelihoods of the smallholding farmers (Khan and Shah, 2011). This will ultimately affect the nation's food security (Misra, 2017; Mohammad Atiqur Rahman et al., 2017), primarily subsistence farming (Habiba et al., 2014). The farming communities are the most vulnerable groups to climate change (Sugden et al., 2014).

Compared to the other parts of the country, the Barind tract and the Teesta floodplain regions of the northern and north-western parts (known as North Bengal) are highly impacted by drought due to high poverty rates, dependency on agriculture, low adaptive capacity and high variability of annual and seasonal rainfall (Habiba et al., 2014, Shahid and Behrawan, 2008). Drought is a recurrent event in these regions (Paul, 1995). Over the years, the severity, frequency and variability of drought have increased in North Bengal (Alamgir et al., 2015, Kamruzzaman et al., 2019a, Miyan, 2015, Mohsenipour et al., 2018, Md Anarul Haque Mondol et al., 2017, Mondol et al., 2016). Several studies indicate that the drought has significant impacts on agricultural productions and the natural environment (Alamgir et al., 2015; Food and Agricultural Organization, 2006; Hossain et al., 2014; Islam, 2009; Karim et al., 1990; Miyan, 2015; Rahman, 2015; Ruane et al., 2013; Sikder and Xiaoying, 2014). Although there have been tremendous improvements in irrigation systems in Bangladesh in recent decades, agricultural activities remain dependent on seasonal rainfall (Akter and Rahman, 2012). A study conducted by Islam (2009) indicate that in recent decades, North Bengal has experienced significant increases in rainfall variability, long seasonal-scale dry spells and numerous instances of below-normal rainfall, significantly hampering the crop growth. Also, variability in temperature has a substantial effect on crop yields (such as rice and wheat) in North Bengal (Amin et al., 2015, Miah et al., 2017).

In addition to the climate variability, the construction of Farraka and Teesta dams in the upstream and the uneven distribution of river water have an impact on downstream water flows which often results in the occurrence of floods during the monsoon period and severe droughts during the dry season (Chowdhury, 2010, Habiba et al., 2011, Ho, 2016, Islam and Sarker, 2017, Mondal and Islam, 2017, Rahman, 2013, Rahman et al., 2017). Therefore, agricultural production in the Barind Tract and the Teesta floodplain areas are vulnerable to water shortages and poor water management. In contrast, a study conducted by Brammer (2016) indicates that drought is not a serious problem for agriculture in Bangladesh compared to about 30 years ago. He pointed that the agricultural productions data do not suggest an increasing impact of drought on agriculture in Bangladesh. For all of these reasons, understanding the dynamics of different types of droughts is very important (Akter and Rahman, 2012, Delaporte and Maurel, 2018, Freitas and Billib, 1997).

Many indices have been used globally for drought characterisation over the past few decades based on the effectiveness, data availability and climatic characteristics (Bandyopadhyay and Saha, 2016). The strength and weakness of currently used drought indices can be found in Svoboda et al. (2016). The Standardised Precipitation Index (SPI) is the most popular meteorological drought index (Byun et al., 2010). However, upon analysing the limitations of the current drought indices, the Effective Drought Index (EDI) has been proposed as a useful tool to distinguish and characterise droughts (Byun and Wilhite, 1996, Byun and Wilhite, 1999). A detailed analysis, comparison of drought indices and advantages of EDI over other indices can be found elsewhere in Byun and Wilhite (1999), Byun and Kim (2010) and Deo et al. (2017). EDI can be used to monitor droughts daily, weekly, monthly, or seasonal basis or for any other specific period and can be applied worldwide (Byun and Wilhite, 1999, Deo et al., 2017, Smakhtin and Hughes, 2007). This index's main strength is the ability to detect the onset and end of the drought and drought conditions earlier than any other indices (Jain et al., 2015, Zarei et al., 2017b). Most of the current indices use normalised statistical metrics to analyse the deficit periods (Mishra and Singh, 2011, Mishra and Singh, 2010), while their multi-scale drought characterisation is ranging from the month (smallest time scale) to longer periods (Deo et al., 2017). These indices are not considered the daily consecutive or accumulated stress of drought (Byun and Wilhite, 1999). Drought severity is calculated mostly from the climatological mean of water deficiency for some predefined duration without considering the diminishing of water resources over time. Studies have confirmed that EDI is more efficient than the SPI in assessing both short term (e.g. Daily, weekly and monthly) and long-term (e.g. Seasonal and annual) droughts (Byun and Kim, 2010, Dogan et al., 2012). SPI measurement is based on the probabilistic distribution, frequency of which depends on the time period and need aggregation of the rainfall (McKee et al., 1993), is not suitable for drought ranking (González and Valdés, 2006). Byun and Kim (2010) compared the EDI with the SPI, and their analysis indicates that EDI detects short term/long term drought that cannot be detected by the long term/short term SPIs, respectively. In addition, they pointed out that the short-term SPIs do not detect a short-term rainfall, create many values for the same period, consider the same weight for both long past rainfall and recent precipitation, may overestimate the relatively low rainfall shortage. Moreover, complexity with determining the appropriate base period for drought analysis is often difficult to find out the start of drought using the most current indices (Byun and Wilhite, 1999, Deo et al., 2017, Jain et al., 2015). In contrast to these limitations, EDI considers any rainfall amount and rainfall day, one single value for the day or period, long memory of rainfall (365 days) and the consecutive weight of rainfall; confirming that EDI is superior to SPI in measuring the drought severity (Byun and Kim, 2010). SPI has limitations for continuous monitoring of drought status (Deo et al., 2017). Moreover, Morid et al. (2006), cited in Deo et al. (2017), indicates that EDI is better for detecting the start and end of drought than the percent of normal, SPI, China Z index and the Z score. EDI can be used due to its self-defined time step, which is free from setting up time step problem (Byun and Wilhite, 1999, Deo et al., 2017, Jain et al., 2015). In addition, considering the multiple data requirements, many indices are not feasible for some regions (e.g. PDSI). Thus, it can be concluded that EDI may be most effective in characterising drought in some regions.

Though there have been numerous studies on the impacts of climate change in Bangladesh, there are few studies on drought characterisation, which are either model-based or based on farmers perception without quantifying agricultural impacts using agricultural production data (Akter and Rahman, 2012, Dash et al., 2012, Habiba et al., 2014, Jabber M.A., 1990 Causes and effects of drought/aridity in Bangladesh using remote sensing technology. In: Proceedings of ESCAP workshop on remote sensing technology in application to desertification/vegetation type mapping, Tehran, August 1990., Jabber M.A., Chaudhury M.U., Huda M.H.Q., 1982 Causes and effects of increasing aridity in Northwest Bangladesh. In: Proceedings of first thematic conference on remote sensing of arid and semi-arid lands, Cairo, Egypt, January 1982., Kamruzzaman et al., 2018, Karim et al., 1990, Mazid et al., 2005, Mondol et al., 2016, Rafiuddin et al., 2011, Rahman et al., 2018, Saleh et al., 2000, Shahid and Behrawan, 2008). Moreover, so far, no comprehensive study was done to analyse the chronology of drought and accumulated drought severity in Bangladesh. Again, almost no study has been conducted to analyse the onset and end of the drought and its peak intensity. Most of the existing work lack the study of the relationship between drought and agricultural production in Bangladesh. EDI has been used recently in Bangladesh (Kamruzzaman et al., 2019b). However, so far, no comprehensive study has been undertaken to evaluate and assess seasonal or growing period droughts in the important agro-ecological zone of the Barind tract and the Teesta floodplain area in North-Bengal. Thus, it is difficult to draw conclusions about past events and plan for future impacts. Such work is necessary to detect and monitor future drought spells and their intensity to support policymaking. To fulfil this knowledge gap, the present paper aims to look for evidence of secular trends in the frequency or severity of meteorological drought and quantify any changes in agricultural output that might be linked to the secular trends in drought occurrence. Though this research is based on two specific agro-ecological areas, the methods we used and insights into the effectiveness of EDI for drought characterisation and the impact of meteorological drought on crop production are applicable to similar regions in Bangladesh and elsewhere.