(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Long-term spatio-temporal variability and trends in rainfall and temperature extremes and their potential risk to rice production in Bangladesh

['Mohammed Mainuddin', 'Land', 'Water Business Unit', 'Commonwealth Scientific', 'Industrial Research Organisation', 'Csiro', 'Canberra', 'Australian Capital Territory', 'Jorge L. Peña-Arancibia', 'Fazlul Karim']

Date: 2022-04

Understanding the historical and future spatio-temporal changes in climate extremes and their potential risk to rice production is crucial for achieving food security in Bangladesh. This paper presents results from a study on trend analysis for 13 climate metrics that significantly influence rice production. The analysis was conducted using the non-parametric Mann-Kendall test and the Theil-Sen slope estimator methods. The study included data from all available weather stations in Bangladesh and the assessment was done for both the wet (May to October) and dry (November to April) seasons, which cover the growing seasons of the country’s three types of rice: Aus, Aman and Boro. Results show significant decreasing trends for wet season rainfall (>12 mm/season/year in some stations) in the central and north regions. In addition, dry season rainfall is decreasing significantly in many areas, whilst dry season dry spells are increasing throughout Bangladesh. Decrease in rainfall in some of these areas are of concern because of its impacts on rainfed Aus rice and in the sowing/planting of rainfed Aman rice and irrigated dry season Boro rice. The maximum temperatures in the wet season are increasing throughout the country at 0.5°C every ten years, significantly at most of the climate stations. The analysis shows that the number of days with temperature >36°C has significantly increased in 18 stations over the last three decades, which implies a serious risk to Aman rice yield. The current maximum temperatures (both in the wet and dry seasons) are higher than the optimum temperature ranges for rice production, and this will have likely adverse effects on yield in the face of climate change with increasing temperatures. The results herein have practical implications for planning appropriate adaptation policies to ensure food security in the country.

Funding: The study is done under a project funded by the Department of Foreign Affairs and Trade (DFAT). This project is one of the several projects under the Sustainable Development Investment Portfolio (SDIP) Program of DFAT. Any funding related to overseas activities are considered as Australian Aid. All authors worked in this project. So it is not only me though I lead the project.

This study focused on assessing the inter-seasonal changes in rainfall and temperature. Unlike previous studies that evaluated changes in mean climate metrics or extremes, our focus is on climate metrics with potential impacts on rice production, both temporally and spatially. Historical trends in rainfall and temperature and the associated agriculturally relevant metrics are assessed for two cropping seasons: May to October and November to April, representing wet and dry season crops respectively. Dry season covers the cultivation period of Boro rice and wet season covers the cultivation period of Aus and Aman rice. We qualitatively link the trends in the derived climate metrics to discuss known risk to rice production based on the literature, and similarly remark on the likely impacts of climate change. In achieving these objectives, this study seeks to inform the debate on the impacts of likely climate change on rice production in Bangladesh.

Crop production is partly controlled by rainfall and temperature and partly by non-climatic factors, such as areas of irrigated crops, development of new crop varieties, crop prices, access to markets and labour among others. Many studies [e.g., 38 – 43 ] have assessed the impacts of climate variability on rice production in Bangladesh. It is important to note that previous studies have focused either on the yield of Boro rice, or their assessment was regardless of the country’s representative climatic characteristics [ 44 ]. For example, although Aman rice has the largest acreage (48.4%) with a production of 38% rice grain [ 45 ], studies on impacts of climate variability for Aman yield have been limited [ 43 ].

However, most studies agree that rainfall has increased in the south [ 26 , 30 ], and possibly also in the north [ 28 , 36 ], but decreased in central areas [ 34 ]. It is important to note that some differences between studies may have been because of differences in the extents of the periods and areas studied and the inherent variability in the rainfall data.

Studies of historical temperature trends in Bangladesh reveal continuously increasing temperature across the country ( Table 1 ); the increase in annual temperature is generally true for different metrics, such as the average, maximum and minimum temperatures, and also for different hydro-climatic seasons in a year. In contrast, findings on rainfall trends are mixed, some studies show increasing trends [ 26 ], some show decreasing trends [ 23 , 24 ] and some show no significant trends [ 27 ] in rainfall over the past several decades ( Table 1 ).

Bangladesh produces about 34.7 million metric tons of rice annually (4 th largest in the world), with almost 80% of the total cropped area (11.7 million ha) planted with rice [ 15 ]. Rice accounts for 91% of the total food grain production [ 16 ] and 60% of the total agricultural output [ 17 ]. Regardless of technological advancements like improved crop varieties and irrigation methods that have made the country almost food secure [ 18 ], climatic factors still add uncertainty in agricultural productivity [ 18 , 19 ]. Inter-annual and intra-seasonal features of the monsoon predominantly control crop production both during the wet May to October and dry November to April seasons [ 20 , 21 ].Climate variability has been identified as a major cause of changes in crop yield in the country, while increasing temperature and changing rainfall patterns have a deleterious impact on crop production [ 22 ]. In recent decades, expansion of groundwater irrigation for dry season crops, particularly Boro rice, has led to concerns of declining groundwater depth in some parts of the country, especially in the northwest region [ 19 , 23 ]. The groundwater tables are declining due to a combination of factors, which, aside from extraction of groundwater, also depend on climate characteristics like declining rainfall [ 16 , 24 ]. The declining rainfall reduces infiltration to groundwater [ 25 ] and runoff supply to wetlands and seasonally flooded areas [ 24 ], which reduce recharge to the underlying aquifers, resulting in a decline in groundwater tables.

Climate variables like rainfall, temperature, solar radiation, wind speed and humidity have key impacts on agriculture [ 1 – 4 ]. Extreme variability in rainfall and temperature regimes often stress agricultural systems in many regions worldwide [ 5 – 7 ]. For example, agricultural production in India on a year-to-year basis has been strongly linked to monsoonal rainfall characteristics [ 8 – 10 ] and El-Niño Southern Oscillation ENSO [ 11 ]. Bangladesh, a predominantly agricultural country, has a monsoonal climate with an intense wet season associated with almost regular flooding [ 12 , 13 ], and a dry season in which crop production relies on groundwater feed irrigation. Currently, about 79% of irrigation water comes from groundwater [ 14 ].

The Theil-Sen estimator, or Sen’s slope estimator [ 68 ] was used to quantify the slopes of the linear trends in the rainfall and temperature metrics. The Sen’s slope method is widely used in the analysis of climate data time-series and has been used in studies focusing on Bangladesh’s climate [e.g., 32 ]. It is regarded as more robust than the least-squares estimator as it is insensitive to outliers and is more precise in the presence of skewed data [ 69 ]. The slope estimator S is the median over the pairs of points in the time series for all i = 1 to n (4) where x i and x j are the data points with i varying from 1 to n-1 and j varying from i+1 to n.

The non-parametric Mann-Kendall (MK) test [ 65 ] was adopted to examine the existence of trends in rainfall and temperature metrics for the 1980 to 2014 period. The MK test is widely used in hydrology and climatology because of its advantages over parametric methods; for example, it can deal with missing data in the series [ 66 , 67 ], which is characteristic of the climate data used in this study. The MK test does not require any assumption of normality, and it is less sensitive to non-homogeneous time series. The test considers two assumptions (null and test hypotheses) while analysing the existence of a trend: (i) that the data is independent and identically distributed (i.e., no trend exists), and (ii) that the data follows a monotonic trend. The MK statistic is computed as: (1) where x is a climate variable, i varies from 1 to n-1 and j varies from i+1 to n. Each of the data in x i is taken as a reference point, which is compared with the rest of the data points in x j by satisfying the condition: (2) if n ≥ 10, the statistic MK is approximately normally distributed with the mean of 0 and standard deviation of σ 2 . The significance in the trend (Z s ) is calculated as: (3)

Daily rainfall and temperature data were aggregated for the two cropping seasons described above: (i) May to October, which captures the growing season of Aus and Aman rice, and (ii) November to April, which captures the growing season of Boro rice. For each season, several climate metrics linked to agricultural production [ 64 ] were derived from the daily data. The climate metrics, based on rainfall and temperature characteristics, are described in Table 2 with their likely impacts on seasonal crop development and production.

Two cropping seasons (wet and dry) were considered here, these cover the main developmental stages for Aman and Boro and aggregate the four hydro-climatic seasons in wet and dry. Wet season crops (e.g., Aus and Aman rice) largely depend on the characteristics of the monsoon [ 55 – 57 ]. Changes in rainfall can have either positive or negative impacts on crop yield depending on timing, location and duration of the monsoon. The starting of monsoon determines the sowing/transplanting period of Aus and Aman rice. Any delay in transplanting from the optimum period reduces yield and exposes the crop, particularly Aman rice, to a period with lower rainfall, thus requiring supplementary irrigation [ 41 , 56 , 57 ]. Low rainfall and/or a long period with no rainfall during the wet season adversely affects the yield of both Aus and Aman rice [ 58 , 59 ], and the impact is mostly statistically significant [ 41 ]. Also, extreme rainfall and floods, quite often encountered, can drastically reduce Aman rice yield [ 60 , 61 ].

The consistency of rainfall data was tested using double-mass analysis [ 52 ]. A detailed description and results of this analysis can be found in the ( S1 Text ), which shows that the double-mass curves in most climate stations essentially follow the same linear relationship throughout the study period (r 2 > 0.99), and thus are considered consistent and suitable for the analyses done herein. There was a notable exception, the insular station at Sandwip, which was impacted by cyclonic activity in 1999, had some daily rainfall totals exceeding 400 mm/day [ 53 ]. Besides these quality-control procedures, no other pre-processing was implemented as the rainfall and temperature data in Bangladesh largely show no serial autocorrelation as demonstrated by the recent work of Mullick et al. [ 32 ] for the 34 climate stations used in this study. The absence of serial correlation was confirmed by calculating the autocorrelations up to lag 10 (i.e., ten seasons) and plotting the sample autocorrelation function (ACF) for each metric (see next section) including bands for rejection testing each autocorrelation = 0 under the assumption of white noise. These are reported in the ( S2 Text ). Therefore, in any case, attempts to correct the data can result in degradation of the results of trend tests [ 54 ].

The Bangladesh Meteorological Department (BMD) is the main source of climate data in Bangladesh. This study used daily rainfall and temperature data from 34 climate stations managed by the BMD. The stations are spatially well-distributed across the country, with at least one station in each of the eight divisional units of the country ( Fig 1 ). Out of 34 climate stations, 10 stations have been operational since 1948, 20 stations were progressively added to the BMD’s network before 1980, while four stations were added to the network after 1980. So, considering continuity of the data and reliability of measurements, data from 1980 to 2014 were included in this study. Only the years having with at least 350 days with data were included in the analysis. The length of data in the final data set that were used for analysis, varies from 23 to 34 years between the stations. While longer the time series can capture variability better, data length of more than 20 years is considered sufficient for trend analysis in hydrological variables [ 51 ]. Fig 2 depicts the data period for each of the 34 climate stations with a demarcation for the duration of missing data. The mean, maximum and minimum data periods are 32, 34 and 23 years (at Mongla climate station), respectively over the assessment period.

Bangladesh is predominantly an agricultural country, characterized by rice paddy agriculture-dominated landscapes. About 21% of the national Gross Domestic Product comes from agriculture [ 45 , 48 ]. Rice dominates the major share (60%) of crop production [ 17 ]. There are three main rice crops in Bangladesh, two of them (Aus and Aman) are predominantly rainfed, and Boro rice, grown in the dry season requires irrigation. Aus is grown from March to July, Aman from July to November and Boro from December to May. Boro rice occupies about 59% of net cultivable area in the dry season and it contributes about 55% of the total rice production in Bangladesh [ 49 , 50 ]. Aman rice occupies 69% of the net cultivable area in the wet season and produces 38% of total rice. The contribution of Aus rice to total rice production is relatively small at only about 7% [ 50 ].

The monsoon season, characterised with frequent rains, starts between late May and early June and ends in late September or early October, with July being the peak monsoon month. In some years, it continues until November, especially in the south-eastern region of the country. The north-eastern part of the country receives the highest annual rainfall (up to 4200 mm) and the western part receives the lowest rainfall (about 1500 mm). The rainfall in Bangladesh is seasonal, with 72% of the total annual rainfall occurring during the monsoon period [ 27 , 33 , 35 ].

The climate of Bangladesh is characterized by four hydro-climatic seasons. These include: dry winter from December to February, hot summer from March to May, monsoon from June to September, and autumn in October and November [ 28 , 47 ]. During the winter months, the mean temperature of the country varies from 18°C to 21°C, and during the summer months the mean temperature varies from 27°C to 29°C. Winter is characterized by relatively cool and sunny days followed by cool nights. January is the coldest month in Bangladesh, and the country’s northern districts (e.g., Rangpur; Fig 1 ) experience the lowest temperatures. Winter temperatures are higher in the central areas (e.g., Dhaka) than in the northern areas, with much higher temperatures along the southern coastal areas due to the influence of the Bay of Bengal. A southerly wind is prominent during winter, and very often brings cool temperatures in the north (3–6°C), central and central-south areas (8–10°C). In February, the temperature rises and continues rising during March to May. The temperature in March to May varies from 30 to 35°C [ 31 , 47 ].

Bangladesh is about 147,500 km 2 and is bounded by India in the west, north and east and the Bay of Bengal in the south ( Fig 1 ). The country is located overlying the delta of three large rivers of south Asia: the Padma (name of the Ganges in Bangladesh), Jamuna (name of the Brahmaputra in Bangladesh) and Meghna (name of the Meghna upstream and after the confluence with the Padma and Jamuna rivers). The topography of the country is relatively flat, except for the hilly north- and southeast, and gently rolling northeast and northwest [ 46 ]. The country is divided into eight administrative regions known as divisions: Barisal, Chittagong, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur and Sylhet ( Fig 1 ).

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of T12 dry for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The spatial and temporal variability and trends in the number of days with temperature <12°C for the dry season, T12 dry , for the 34 climate stations are shown in Fig 11 . The mean T12 dry is 23 days/sea (from 1 days/sea in Cox’s Bazar to 59 days/sea in Srimangal), CV is 59% (from 14% in Srimangal to 198% in Cox’s Bazar), and S is 0.080 days/sea/yr (from -0.658 days/sea/yr in Rangpur to 0.636days/sea/year in Sandwip). Similar to T36 wet , T12 dry also follows a visible east-west gradient, with generally 0 to 15 days in the east to 35 to 60 days in some locations in the west. Also, the large number of days with T12 dry is quite regular over the years in the west, but highly variable in the southeast. There are 11 climate stations with significant (p<0.1) trends, 3 with declining and 8 (mostly in the coastal areas) with increasing trends ( Table 4 ). The results elsewhere are mixed.

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of T36 wet for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The spatial and temporal variability and trends in the number of days with temperature >36°C for the wet season, T36 wet , for the 34 climate stations are shown in Fig 10 . The mean T36 wet is 5 days/sea (from 0 days/sea in Teknaf to 21 days/sea in Chuadanga), CV is 124% (from 46% in Chuadanga to 316% in Teknaf), and S is 0.128 days/sea/yr (from -0.045 days/sea/yr in Srimangal to 0.643 days/sea/yr in Jessore). There are 17 climate stations with significant (p<0.1) increasing T36 wet trends; the rate of increase is 0.52°C per year ( Table 3 ). T36 wet is higher at the climate stations in the west and northwest areas (>5 days/sea), while it is lower at the stations in the western and coastal areas (<3 days/sea). T36 wet follows a noticeable east-west gradient, with 1 to 3 days in the east to 10 to 21 days in some locations in the west. The CV is lower (<50%) at all climate stations (>100%) in the west and northwest areas than at the stations in the northeast and coastal areas. The large number of days with T36 wet is quite regular over the years in the west, but highly variable in the southeast. Significantly increasing trends (p<0.1) in T36 wet occur mostly at the climate stations in the western area, the rate being 5 days in every ten years (see station number 28, Satkhira), while the declining trends in T36 wet occur mostly in the coastal areas.

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of Tmin dry for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The spatial and temporal variability and trends in mean daily minimum temperature for the dry season, Tmin dry , for the 34 climate stations are depicted in Fig 9 . Tmin dry is 17°C/sea (from 15°C/sea in Srimangal to 19°C/sea in Cox’s Bazar), CV is 3% (from 2% in Khepupara to 5% in Rangamati), and S is 0.014°C/sea/yr (from -0.048°C/sea/yr in Sandwip to 0.055°C/sea/yr in Khulna). The CV is slightly higher (>4%) at the stations located in the northwest area. There are 21 climate stations with significant (p<0.1) trends, 5 with declining and 16 with increasing trends ( Table 4 ). Most of the climate stations with increasing Tmin dry trends are located in the central and northern areas, and the 5 stations with declining trends are located in the coastal areas.

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of Tmin wet for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The mean Tmax dry is 29°C/sea (from 27°C/sea in Rangpur to 30°C/sea in Jessore), CV is 2% (from 2% in Teknaf to 3% in Sitakunda), and S is 0.014°C/sea/yr (from -0.035°C/sea/yr in Satkhira to 0.070°C/sea/yr in Sitakunda). There are 14 climate stations with significant (p<0.1) trends in Tmax dry , 3 with declining and 11 with increasing trends. The climate stations located in the northwest area are slightly cooler (<28°C/sea, numbers 10, 26 and 29) than elsewhere. The CV is low and does not exceed 3% in any climate stations. The climate stations with significant (p<0.1) increasing trends are mostly located in the coastal areas (Tables 3 and 4 ).

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of Tmax dry for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of Tmax wet for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of RCDD dry for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The spatial and temporal variability and trends in the maximum number of consecutive dry days, RCDD dry , for the 34 climate stations are shown in Fig 5 . The mean RCDD dry is 81 days/sea (from 71 days/sea in Sylhet to 100 days/sea in Teknaf), CV is 36% (from 30% in Teknaf to 43% in Dinajpur), and S is 0.919 days/sea/yr (from 0.232 days/sea/yr in Rangamati to 1.690 days/sea/yr in Tangail). The climate stations in the north (numbers 10, 29 and 26 in Fig 5 ) and the southeast areas have generally longer RCDD dry (>81 days/sea) compared to most climate stations in the central and western areas (<79 days/sea). The CV is relatively high at the mentioned stations in the north, but there are no clear patterns in CVs in the other areas. All 34 climate stations have increasing trends in RCDD dry , 18 with significant (p<0.1) trends ( Table 4 ).

The mean RSEA dry is 241 mm/sea (from 120 mm/sea in Dinajpur to 605 mm/sea in Sylhet), CV is 56% (from 35% in Sylhet to 75% in Sayedpur), and S is -5.33 mm/sea/yr (from -9.368 mm/sea/yr in Chittagong to 0.169 mm/sea/yr in Dinajpur). RSEA dry is relatively high (>200 mm/sea) at most of the climate stations located at the east and southeast areas (with the exception of Teknaf, station number 34 that is situated further south) and relatively low (<200 mm/sea) at the climate stations located at the west and northwest areas. The CV does not reveal any discernible spatial pattern. All but one (Dinajpur, station number 10, in the northwest) climate stations exhibit decreasing trends in RSEA dry . There are 27 climate stations with significant (p<0.1) decreasing trends, mostly located in the central, southwest and southeast areas of the country ( Table 4 ).

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of RSEA dry for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The top row depicts the geographical locations of the 34 climate stations showing the mean (in mm per season, left map), coefficient of variation (CV in %, middle map) and Sen-slope (S, in mm per season per year, right map) in circles with graduated colours and significant S values (p<0.1) shown with a smaller black interior dot. In the rest of rows, time-series (blue lines) of RSEA wet for the 34 climate stations (sorted in alphabetical order) and S (green lines) are shown. Summary information in each plot includes the mean, CV, MK statistic and S.

The spatial and temporal variability and trends in the seasonal total rainfall in the wet season, RSEA wet , at the 34 climate stations are illustrated in Fig 3 . The mean RSEA wet is 2214 mm/sea (from 1323 mm/sea in Rajshahi to 4151 mm/sea in Teknaf), CV is 20% (from 12% in Teknaf to 28% in Chandpur), and S is 0.793 mm/sea/yr (from -14.889 mm/sea/yr in Mymensingh to 23.750 mm/sea/yr in Hatiya). RSEA wet is relatively high (>2000 mm/sea) at the climate stations located in the southeast and relatively low (<1900 mm/sea) at the climate stations located in the central and western areas of Bangladesh. The CV is relatively high (>19%) at the climate stations in the northwest and relatively low (<19%) at most of the climate stations in the coastal, southwest and northwest areas. Increasing trends in RSEA wet occur at the coastal and southwestern climate stations, while declining trends occur at the central and northern climate stations. There are 9 climate stations with significant (p<0.1) trends, with 4 declining (at the north, see Fig 3 ) and 5 increasing (at coastal areas, see Fig 3 , Table 3 ).

Discussion

Variability and trends in rainfall metrics linked to rice production The seasonal total rainfall for the wet season, RSEA wet , is highly variable spatially, with an approximate east-west gradient, and is more variable in time in the northwest and central areas (Fig 3, Table 3). RSEA wet is increasing, generally significantly, in the coastal area (>20 mm/sea/yr in some stations, e.g. Hatiya and Sandwip) and the southwest (e.g. Khepupara) area, but decreasing significantly in some central (such as Mymensigh, Madaripur) and northwest areas (>12 mm/sea/yr in some stations, e.g. Rangpur and Rajshahi). These results agree with those of Mullick et al. [32] who reported generally increasing monsoon and post-monsoon rainfall trends during 1966 to 2015 in the southwest and coastal areas, and mixed trends in the northwest and central areas. Mojid et al. [23, their Fig 2], for a similar study period as in this paper (1985–2015) but at annual scale, reported that the overall average annual rainfall of six climate stations in the northwest region decreased significantly, and at a similar rate (~9.2 mm/year) of the seasonal results reported here. In addition, the results of Nishat and Mukherjee [70] and Shahid [33] on the maximum seasonal rainfall during the pre-monsoon and monsoon seasons with prominent increasing trend in the coastal regions are similar to the results reported herein. However, Rahman et al. [21, their Fig 7], for a longer study period (1954–2013), found similar patterns on an annual scale except for increasing trends in the northwest areas, while Mullick et al. [32] did not find significant rainfall trends during the monsoon in northwest Bangladesh for the last 50 years (1966–2015). The increasing trends in the annual total rainfall observed by Rahman et al. [21, their Fig 7] in the northwest areas that are in contrast to the results of the other investigators might be due to data limitation in their study since some of the currently available climate stations in that region were installed during 1980s or later. Inter-decadal variations in the Indian Monsoon have been reported in Krishnamurthy and Goswami [71], suggesting epochs of 30 years with below and above mean rainfall. Krishnamurthy and Goswami [71] showed that selected Indian stations reported below mean rainfall during 1959 to 1988 and above mean rainfall conditions during 1989 to 1998 (their analyses ended in 1998). These observations lead to the general agreement that the seasonal total rainfall for the wet season has decreased significantly in some central and northwest areas of Bangladesh that are major rice-growing areas. The decreasing rainfall, often causing droughts, drastically reduces yields of the rainfed Aus and Aman rice in these areas leading to increased requirement of supplemental irrigation with eventual stress on available water resources. To cope with the wet season droughts, use of supplemental irrigation is gradually increasing in these areas. However, the supplemental irrigation may be less important or not required at all if wet season rainfall increases in the future as the climate change rainfall projections of Hasan et al. [72], Kamruzzaman et al. [73] and Karim et al. [74] indicate that the annual total rainfall will increase in the drought-prone northern region, and regionally with an increase in the Indian Monsoon rainfall [75]. The seasonal total rainfall for the dry season, RSEA dry , follows a similar spatial pattern as RSEA wet (i.e., an approximate east-west gradient), but is generally more variable in time in the southeast (Fig 4, Table 4). RSEA dry is decreasing throughout Bangladesh, significantly in many areas, more so in the central and southeast areas (>9 mm/sea/yr). These results agree with those of Mullick et al. [32], who also reported decreasing trends of RSEA dry at most of the climate stations across the country during winter (December to February). Decreasing trends of RSEA dry during winter and pre-monsoon seasons were also reported by Basak et al. [76] and Rahman et al. [21]. Although the rainfall projections of Karim et al. [74] suggest a moderate increase (~10 mm/year/season) during the dry season in northwest Bangladesh the observed past trend exposes the opposite. The dry season is the cultivation period of Boro rice, the irrigation requirement of which varies from season to season due to available rainfall in the dry season [55, 77]. So, if the current decreasing trend of dry season rainfall continues, it will increase the pressure on already stressed water resources of the country. In spite of decreasing RSEA dry irrigated Boro rice acreage continuously increased all over Bangladesh in the past 3–4 decades with associated increased groundwater abstraction. This practice caused decline in groundwater level in the recent past decades with eventual scarcity of dry season irrigation water, predominantly in most areas of the northwest region [50]. To cope with this, the government has put embargo on installing new irrigation wells and is vigorously advocating crop diversification by replacing high water-demanding rice with low water-demanding crops. Farmers are also increasingly adopting crop diversification practices since they very often obtain more economic profit than they obtain from irrigated Boro cultivation [50]. Whilst most climate stations do not reveal significant trends for RCDD wet , RCWD wet and RCWD dry (see S3 Text, Figs P-R), RCDD dry is increasing across the country (Fig 5, Table 4), quite dramatically (>1 days/sea/year) in some central and northern areas where dry spells are already long (>70 days). This observation demonstrates that the seasonal RCDD dry index exposes more specific and accurate climatic condition than would be obtained from its annual resolution. The increasing trends in RCDD dry at many climate stations (e.g. Rangpur, Bogra, Rajshahi, Dhaka and Faridpur) have occurred in relatively recent dry seasons; the time-series in Fig 5 show that at several stations (numbers 6, 7, 9, 10, 11, 15, 26, 27, 33 in Fig 5) RCDD dry is higher on average from the year 2000 onwards than the previous years. In addition to the decreased seasonal total rainfall in the dry season, the RCDD dry also contributed to increased irrigation requirement for Boro rice cultivation.

Variability and trends in temperature metrics linked to rice production Both the maximum and minimum temperatures have increased significantly (p<0.1) across several climate stations over the period from 1980 to 2014. Tmax wet is spatially variable (Fig 6), approximately following an east-west gradient. The west, on average, is slightly cooler (~2°C) than the east. Tmax wet is increasing significantly (Table 3) throughout the country by 0.5°C in every ten years. If the trend continues, the increased Tmax wet will drastically affect Aman rice production by deteriorating grain formation. Tmax dry follows a similar pattern as Tmax wet but is cooler in the northern areas (~2°C). It is significantly increasing in the coastal areas by 0.5°C in every ten years. It is decreasing in the north and west, significantly at some locations (Table 3) thus revealing the opposing trends and more specific climate information for crop growth in the two seasons that would not be obtained from the analysis of data on annual resolution. Tmin wet follows an east-west gradient, with temperatures in the east being slightly cooler (~1°C) than in the west. The minimum temperature shows seasonality; Tmin wet is generally increasing throughout the country, sometimes by 0.3°C in every ten years. Although, Tmin dry follows a similar pattern as Tmin wet but is cooler in the northern areas (~3°C) than in the other areas. It is significantly decreasing in some coastal areas (Table 4) by 0.5°C in every ten years and increasing in the north, significantly at some locations in the northwest (Table 4), by 0.3°C in every ten years. Rahman et al. [2] also reported similar results, but with higher incremental rate of the minimum temperature, while Wassmann et al. [78] reported a gradual increase in temperature in most regions of the country. Similar to our findings, Rimi et al. [79] also reported an insignificant increasing trend of the annual maximum and minimum temperatures for the southern region (e.g., Satkhira district) over the period from 1950 to 2006. Basak et al. [76] estimated an increasing trend of 0.012°C per year for the yearly average maximum temperature and 0.015°C per year for the yearly average minimum temperature. In contrast, our analysis of the extensive data set provides higher increasing trend (0.024°C per year) for the yearly average maximum temperature but lower increasing trend (0.012°C per year) for the yearly average minimum temperature than the estimates of Basak et al. [76]. Further rise in average day temperature of 1.0°C by 2030 and 1.4°C by 2050 has been predicted [80]. All the climate stations in this study reveal an increasing trend in temperature during May to October. Nishat and Mukherjee [70] also reported an increasing trend in the mean seasonal temperature within the range of 0.40–0.65°C during the period from 1967 to 2007. We observed 0.52°C per year increasing rate of the extreme high temperature (>36°C), T36 wet , for the wet season (Table 3). But Nishat and Mukherjee [70] reported 0.87°C increase in the maximum temperature during pre-monsoon (March-May) and 0.42°C during post-monsoon (September-November) months. Based on the seasonal maximum values, these temperatures are in agreement with our results in terms of their timing of occurrence. In our analysis, the monthly low temperature, T12 dry , shows a decreasing trend in December (0.022°C) and January (0.031°C) but increasing trend (0.002 to 0.04°C) in all other months of the year. Partially contrasting timing and magnitudes of occurrence of the dry season minimum temperature were also reported; for example, Nishat and Mukherjee [70] reported an increase in the minimum temperature during December to February (0.45°C) and June to August (0.52°C). While the historical trend shows both the increase and decrease in temperature, climate change projections indicate that both seasonal and annual temperatures will increase in the future across Bangladesh [72, 81, 82]. For example, Agrawala et al. [83] predicted 1.1°C, 1.6°C and 2.7°C increase in winter/dry season temperature for the year 2030, 2050 and 2070, respectively. Based on 11 GCMs, Fahad et al. [81] predicted mean temperature rise of 3.2°C to 5.8°C for Bangladesh at the end of the twenty-first century. Similarly, Hasan et al. [72] and Caesar et al. [82] predicted 2.0°C to 4.5°C and 2.6°C to 4.8°C rise in mean temperature, respectively by the end of the century. Thus, although the magnitudes may vary the results of all these studies lead to the general consensus that both the mean seasonal and annual temperatures will increase to a large extent in the future.

Risk of rainfall variability and trends to rice production The success of Aus and Aman rice production largely depends on the monsoon, whereas Boro rice being largely irrigated is much more sensitive to changes in temperature than to changes in rainfall [62]. Depending on timing and magnitude, the changes in rainfall can have either positive or negative impacts, which are specifically significant to Aus and Aman rice production [41]. Changes in rainfall also affect recharge rates into aquifers that, can affect irrigated agriculture, particularly Boro rice, in the northwest and central regions of Bangladesh. These regions are the main crop production zones and considered the food bowl of the country [25, 84]. The northwest region, despite having only 31% of net cultivable area of the country, produces 37% of Aman rice, 35% of Boro rice, and 60% of wheat, maize and potato of the total production of the country [50]. In our analysis, most climate stations reveal decreasing rainfall trends (four significantly, Table 3, Fig 3) during the wet season in the northern parts of the country. This can have obvious adverse effects on crop production, specifically Aman rice production [85] that covers ~69% of the country’s total cultivated land area for rice production [86]. The decrease in rainfall in the wet season (as shown by RSEA wet ) increased in-seasonal drought for Aman rice over the past decades [87]. Due to sub-humid and dry conditions, droughts occurred quite frequently in the northwest region during June/July to October/November [87, 88] and about 5.5 Mha of land have been classified as drought-affected [89]. The decreased rainfall in the wet season increased the need of supplementary irrigation for Aman rice in the drought-prone areas. Consequently, supplementary irrigation is increasingly being practiced in those areas over the past years. Without supplemental irrigation, rice production may fluctuate greatly on a year-to-year basis [41, 86] with the resulting adverse effects on food security in the country, particularly in dry years. This is likely to be exacerbated by climate change [90] since the future climate change in Bangladesh is anticipated to increase the variability in the production of all three cultivated rice types: Aus, Aman and Boro [18]. In addition, the wet season rainfall recharges groundwater, which is extracted and used in the dry season for irrigation. There is strong evidence [23, 24, 91, 92] that groundwater levels are declining in the northwest region, which is the most intensely groundwater-irrigated area in Bangladesh [14]. The reduction in rainfall in the future is likely to further exacerbate the situations as there will be less recharge to the aquifers from rainfall in the wet season [25] with the eventual adverse effects on the productivity of irrigated crops (e.g., Boro rice, wheat, maize) grown in the dry season. The effects of decreasing rainfall in the dry season, as shown by RSEA dry , on groundwater are also of concern, particularly in the northwest region [93]. Boro rice is grown in the dry season using groundwater irrigation in the northwest and central regions. The decrease in dry season rainfall will increase irrigation requirements of the crops, thus exerting further pressure on groundwater for irrigation, particularly in the northwest region where irrigation is almost entirely dependent on groundwater. To cope with this situation, rice breeders are trying to develop short-duration varieties to reduce water requirement; but until now the success is limited [94]. As described before that until now restriction on installation of additional irrigation wells and advocating crop diversification have the government’s policy to keep both groundwater usage and food production sustainable. However, this policy might not be adequate or even appropriate in future since irrigation requirements are also projected to increase due to climate change [55, 95]. The increasing irrigation amount and associated costs will greatly affect the livelihood of the farmers in groundwater-irrigated regions of Bangladesh. The declining groundwater table due to high extraction will also intensify water scarcity and lead to insufficient supply of freshwater in dry periods for the growing population in the urban areas [23]. In the face of climate change, declining groundwater levels, variable rainfall and irrigation will represent a challenge for Bangladesh and South Asia’s food security [96, 97]. However, if climatic variability [71] and projected increases in the Indian Monsoon rainfall [75] are considered, the negative effects of declining rainfalls (particularly in the northwest region) may be reverted in the medium to long term [74]. On the other hand, floods due to extreme rainfall may become more prevalent in the region [98]. The poldered (large areas enclosed and protected from seawater flooding by man-made earth embankments) coastal areas (south and southwest regions) of Bangladesh are already prone to flooding and water logging due to high rainfall and limited drainage opportunities since the water levels in the surrounding rivers of the polders are high [99, 100]. Thus, water management remains a critical problem within the polders and appropriate water management method(s) is(are) yet to be developed. Aman rice is the main crop in the region grown in the wet season and famers usually cultivate low-yielding local varieties of Aman rice [101]. The Government of Bangladesh intends to increase the productivity of Aman rice in this region by introducing high yielding varieties to secure rice grain self-sufficiency and improve livelihood of the local people [99, 101]. The increasing trend in wet season rainfall (RSEA wet ) in this region may not only seriously hamper this effort but also make the current cropping practice difficult with eventual deleterious impact on the livelihood of millions of people. The number of consecutive wet days, RCWD wet , is the lowest (102 day) in the western region and highest (149 days) in the northeast region, the average wet days being 119 days. Seventy-nine per cent (79%) of the wet days occur within May to October. RCWD wet increases annually by 0.19 days and seasonally (May-October) by 0.12 days. These results demonstrate that rainfall has become increasingly variable with uneven temporal and spatial distributions. The erratic rainfall pattern often produces extreme events like floods and droughts, which have adverse effects on crop production, especially on rice production [85].

[END]

[1] Url: https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0000009

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/