EXPLORING THE CLIMATE CHANGE EFFECTS ON BORO RICE YIELDS OF RAJSHAHI DISTRICT IN BANGLADESH.

Md. Mostafizur Rahman 1 , M. Sayedur Rahman 2 and Sabiha Sultana 3 . 1. Assistant Professor, Environmental Research Group, Department of Statistics, University of Rajshahi, Bangladesh. 2. Professor, Environmental Research Group, Department of Statistics, University of Rajshahi, Bangladesh. 3. Environmental Research Group, Department of Statistics, University of Rajshahi, Bangladesh. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History

Bangladesh is a developing country with high population density and high population growth rate. Rice is the principle food of Bangladesh. Although country considers as fourth rice producer in the world but still it fell food insecurity because of its high population growth rate. The objective of this study was to estimate the relationship between Boro rice y ields and climate variables using aggregate-level time series data for the period of 1987 to 2015. The empirical analysis showed that four climate variables minimu m hu mid ity at vegetating phase, min imu m temperature and consecutive days rainfall average at ripping phase and dummy variables have substantial effects on the Boro rice yield. The result also indicates that average seasonal minimu m temperature, minimu m hu mid ity and average rainfall are statistically significant and positively affect the yield of Boro rice in case of Rajshahi district. Moreover, excessive rainfall may create water logging condition and flooding that also destroys the crop production. Therefore, the concerned authority should take appropriate policies to fight against the climate change impact on rice production to ensure food security for the ever increasing population of the country through imp lementing sustainable agricultural develop ment.

…………………………………………………………………………………………………….... Introduction:-
The people of Bangladesh are popularly referred to as "Macche-BhateBangali" or "fish and rice makes a Bangali". Fish and rice is essential part of the life of Bangladeshi people. Rice is the staple food for over 150 million populations. However, country faces a tremendous challenge for providing food security to the increasing population. Therefore, it is imperative to increase rice production in order to meet the growing demand for food emanating fro m population growth. Although, there have been ups and do wns in the domestic production of food grain. The diverse climat ic phenomena like cyclone, drought, changing rainfall patterns and temperature; there has been a significant lost in food grain production in every year. For example, two rounds of floods and devastating cyclone Sidr in 2007 and cyclone Aila in 2009 caused severe damages in agriculture production, especially the rice production. Climate Change Impacts on Rice Production in Bangladesh is important issue. The rice cropping pattern of Bangladesh has changed areas once occupied by the rainfedaus gradually shifted to boro cultivation. As a result, the contribution fro m each season also changed, aman rice prev iously contributed a major portion of total rice, but 108 boro is now the major contributor to total rice production in the country. Aus, aman, and boro rice were recently reported to account for 7%, 38%, and 55%, respectively, of the total rice production in Bangladesh (Risingbd, 2014). Among these rice only 5% of aman rice and 8% of aus rice are irr igated and boro is fully irrigated rice grown in the dry winter and the hot summer (Mah mood, 1997; Ahmed, 2001). According to BBS (2014), boro is the most important crop in Bangladesh in respect of volume of production. Productivity of boro rice depends on several of climatic parameters such as temperature, rainfall, humidity; hydrological properties of soil such as pH, organic carbon, rice varieties, and major production inputs, such as irrigation and fertilizer management practices. It is very important to have an idea about the contribution of boro rice production of Bangladesh from different statistical viewpoint for different region of Bangladesh. To ensure political stability, ach ieve sustainable development goal and the self-sufficiency in food of Bangladesh we have to pay attention for increasing rice production despite the serious effect of climatic condition. So, the aim o f this paper is to exp lore the climat ic condition of the boro rice yields of the study area based on rainfall, temperature and other climatic condition.

Previous study
There is a growing body of literature in recent years that has observed the influence of climate change on agricultural productivity. Mahmood (1998) and Mahmood et al., (2003) examined the effects of higher air temperature and at mospheric CO 2 concentration on rice yield they also showed that daily maximu m and minimu m air temperatures, daily precip itation and daily solar radiation, all of wh ich could affect rice yield significantly. Rashid and Islam (2007) identified droughts, floods, salinity and cyclones as the major extreme climatic events to which Bangladeshi agriculture is most vulnerable.
The effects of climate change are inherently region specific, incit ing the need for region -based research on climate change. Ruttan (2002) noted rainfall and sunlight has potential effect on agricu ltural p roductivity but the gross effect was largely region-specific. Karimet al (1996) conducted a series of simu lations using the CERES -Rice and -Wheat models for Aus, Aman and Boro rice, and wheat. Basaket al (2010) concluded that climate change was likely to have predominately adverse impacts on the yield of Boro rice. They found that if climate change was to result in increased temperatures, that this would cause grain sterility du ring the growing season and hence a reduced yield. They also found that while changes to the level of at mospheric carbon dioxide and solar radiation might offset the impact of increased temperatures to some degree, that it would not be sufficient to mit iga te it altogether (Iqbal and Siddique 2014). Mah moodet al (2004) observed that since rain-fed rice constitutes over 50% of total rice production in Bangladesh, production of this crop is extremely vulnerable to volatility in the supply of water. Chowdhury and Khan (2015) examined the impact of climate change on the rice production in Bangladesh using time series data.
Reilly et al. (1996) found that as temperatures move changed the production of rice. Although the effect of climate change is serious but there are limited researches conducted in case of developing countries (e.g. Boubacar out a study to examine the relat ionship between three climate variables (maximu m temperature, minimu m temperature and rainfall) and three different rice crops (Aus, Aman and Boro). Awal and Siddique (2011) estimated the trend of rice production by employing ARIMA mod el but they did not consider climate influence. Hossain and Silva (2013) have conducted an initiat ive to find out the climate change impact on rice and wheat yield in Bangladesh. Rimi et al. (2009) concluded that temperature variat ions had spectacular implications on crop yield. Rah man et al (2018) showed the impact of climatic parameter on Aman rice production in Rangpur district of Bangladesh. Fro m the above study we found that it is recent demand to investigate the effect of climate change for rice production specially boro rice in western part of Bangladesh.

Research Methodolog y and Data Processing
The study was conducted in purposely-selected Rajshahi district in Bangladesh. The study was limited to some climatic factors and non-climatic factors. The selected climatic variab les were maximu m temperature, min imu m temperature, average temperature, total rainfall, consecutive rainfall, average consecutive rainfall, maximu m sunshine, minimu m sunshine, average sunshine, maximu m hu midity, min imu m hu mid ity, ave rage humid ity. Time (YA R) and dummy variable (DM Y) are the non-climat ic variables. Estimation of the model required two sets of data (i) Historical crop yield data from period 1987 to 2015 and (ii) Historical data on a number of agro -climatic variables fro m period 1987 to 2015. The info rmation furn ished in the study based on different database. Data collected fro m DAE (Depart ment of Agriculture Extension) climat ic data will be collected fro m Bangladesh Meteorological Depart ment (BMD). Rice production data will also be collected from the different Yearbook of Agricultural Statistics of Bangladesh (BBS).

Ti me di vision for boro rice
Boro rice (HYV) cultivation period is fro m midd le December to middle April. Boro rice has three growing stage which called phase. In this study we reduced data for Boro rice growing period fro m 16 December to 24 April according to these three phases. Growing period of Boro rice are d ivided into three stages; Stage 1= vegetating phase (16December-18Februarey = 65 days) Stage 2 = reproductive phase (19February-25March = 35 days) and Stage 3 = ripening phase (26March-24April = 30 days) After dividing the Boro rice growing period into three stage we found that stage 1 and stage 2 are in the robi season and stage 3 is in the pre-kharif season. For every phase we process climat ic data indiv idually.

Data processing
Data will enter and processing in the computer by using MS-Excel and MS-access programs. All data were converted into standard units (Metric tons, mm, °C). The model is based on the assumptionthat there exist a quantitative relationship between periodic values of agro-climaticvariables and final crop yields. Data is generally available at the districtlevel and second, since many agricultural policies are imp lemented at district level,the exercise may provide useful inputs for district level planning. Climatic data reduced in different phases are given bellow:

Rainfall:
The daily rainfall data has reduced as total rainfall, average rainfall, consecutive days and consecutive average rainfall according to three phases of crop growth.

Maxi mum temperature:
Fro m the daily maximu m temperature data, data has been reduced as maximu m temperature and average maximu m temperature of d ifferent phases of crop growth . Mi ni mum temperature: Fro m the daily minimu m temperature data, data has been reduced as minimu m temperature and average minimu m temperature of the three crop gro wth phases.

Sunshine:
Sunshine data has been reduced as maximu m, minimu m and average sunshine according to thre e phases of crop growth.

Humi dity:
Hu mid ity data has been reduced as maximu m, min imu m and average humid ity of the crop growth phases. Data analysis performed by statistical package SPSS. In Rajshahi reg ion Boro rice production graphs will be prepared for last decade  and made comparison with different climatic parameters. Then effect of the climatic parameters will be analy zed for Boro rice production.

Multi ple Regression Model
Regression analysis is a statistical technique for investigating and modeling the relationshipbetween variables (Montgomery and Peck, 1982). In fact, regression analysis may be the most widely used statistical technique . The yield fo recasting model has been used in this study was specified as:  In each step, a variable is considered for addition to or subtraction from the set of explanatory variables based on some pre-specified criterion. Usually, this takes the form of a sequence of F-tests or t-tests, but other techniques are possible, such as adjusted R 2 , Akaike information criterion(AIC),Bayesian information criterion, Mallo ws'sCp, PRESS, or false discovery rate.In a stepwise regression, predictor variables are entered into the regression equation one at a time based upon statistical criteria. At each step in the analysis the predictor variable that contributes the most to the prediction equation in terms of increasing the multi ple correlat ion, R, is entered first. Model accuracy is often measured as the actual standard error (SE), MAPE, or mean error between the predicted value and the actual value in the hold-out sample. This method is particularly valuable when data are collected in different settings (e.g., different times, social vs. solitary situations) or when models are assumed to be generalizable.

Graphical representati on of the data
Multiple linear regression analysis requires some assumptions which are given bellow; 1. the prediction errors are independent over cases; 2. the prediction errors follow a normal distri buti on; 3. the prediction errors have a constant variance (homoscedasticity); 4. all relations among variab les are linear and additive.
All of these criteria for mult iple linear regressions is shown in Figure 1 Figure 1 we found that our regression assumptions is that the residuals (prediction errors) are normally distributed with mean 0.01 and standard deviation is 0.909. For normality our find ing indicates slight violation of this normality plot suggest that normality assumption is not met. Here we have no major cause for concern. The scatter plot shows that the plotted points creates some homogeneous blocks i.e, the associated random variab les have a homogeneity in their variance. So we may conclude that the data set is free of heteroscadasticity. Above individual coefficient plot shows that there is a positive trend between yield and dummy variables i.e, any increment or decrement of other climatic variable will increase or decrease the production respectively.

Association of the vari ables with yield
The association between these variables is reported in figure 2. respectively. This figure also represents the positive trend between yield and S1_humid ity_min variables and there is a positive trend between yield and S3_con_rainfall_avg variables. And finally, positive trend between yield and S3_ min_tem variables indicates any increment or decrement of the climat ic variable S3_ min_tem will increase or decrease the production respectively.

Results and Discussions:-
This section demonstrated the multip le linear regressions with IBM SPSS using stepwise method. SPSS starts with zero predictors and then adds the strongest predictor Dummy. Then it adds the second strongest predictor (S3_ min_tem). And this procedure continues until none of the excluded predictors contributes significantly to the included predictors. Our emp irical result indicates that 4 out of 46 predictors are entered and none of those are removed. The stepwise regression method for predicted variables is shown in Table 1. In order to compare th e performance of the model we use the summary statistics such as R, R-square, Adjusted R-square, standard error etc. these are reported in Table 2. Adding each predictor in our stepwise procedure results in a better predictive accuracy. From Table 2 we found that the Pearson Correlation between actual and predicted Values for model 1, 2, 3 and 4 are 0.875, 0.921, 0.959 and 0.966 respectively which confirm the positive strong correlation between these variables. R-square for the 1 st model is 76.5%, which means that the predictor dummy variable is account for 76.5% of the variance in overall yield. Similarly we can shows that R-square for the 2nd model is 84.9%, wh ich means that the predictor variables (Du mmy and S3_ min_tem) are account for 84.9% of the variance in overall yield wh ich also indicates that S3_ min_tem is account for (84.9%-76.5%) = 8.4% change in variance proportion in this model. R-square for the 3rd model is 92% and for 4th model is 93.4%. Adjusted R squared is a corrected g oodness-of-fit (model accuracy) measure for linear models. It identifies the percentage of variance in the target field that is exp lained by the input or inputs. The models 1, 2, 3 and 4 identifies 75.6%, 83.7%, 91% and 92.2% of variance in the target field respectively that is explained by the inputs. Depending on adjusted R2 among all the models the 4th model fit the data more accurately. The Durb in-Watson statistic is 1.695 which is between 1.5 and 2.5 and therefore the data is not autocorrelated. The estimated parameters of the regression equation of these four models are given in Table 3.
A rule of thumb is that Tolerance < 0.10 indicates mu ltico llinearity. As the value of tolerance for each case is >0.10, so we can conclude that there exists no multico llinearity. The analysis reveals the value of variance inflation factors (VIF) of the explanatory variables in the final models of equations are less than, so it can be said that the data have no evidence of collinearity problems. Variance inflation fact or (VIF) is nothing only a mult icollinearity measurement technique of a regression model. Fro m this table most of the parameters of all models are statistically significant except the constant term. It is also evident that all the regression models are hig hly significant. The table shows that the independent variables are statistically significantly predicted better result as F(4,23) = 81.0614 i.e. the 4 th regression model are good fit to the data. Our unstandardized coefficients and the constant allow p redicting yield precisely :

Interpretation of the models
Interpretation of the models with standard error, t statistic and significance level of the variables are given bellow: Model 1: With the coefficients, standard errors, t-value, p-value and R 2 value of the table we can interpret the model 1. DM Y is mean ingfulness and so on. This means that respondents who score 1 point higher on meaningfu lness will -on average-score 1.03 points higher on yield. Impo rtantly, all predictors contribute positi vely (rather than negatively) to job satisfaction. This makes sense because they are all positive work aspects. All predictors are highly statistically significant (p = 0.000), which is not surprising considering our large sample size and the stepwise method we used.

Model 2:
The predictors are meaningfulness and so on. This mean that respondents DMY score 1 point higher on mean ingfulness will -on average-score 0.821 po ints higher on yield and respondents S3_min_tem score 1 point higher on meaningfulness will -on average-score 0.113. If DM Y and S3_ min_tem becomes 0 then there exists a positive value 1.142 which is the intercept term of the model. Importantly, all pred ictors contribute positi vely (rather than negatively) to yield. This makes sense because they are all positive work aspects. All predictors are statistically significant (p< 0.05), which is not surprising considering our large samp le size and the stepwise method we used.

Model 3:
The predictors are meaningfulness and so on. This mean that respond ents DMY score 1 point higher on mean ingfulness will -on average-score 0 .881 points higher on yield, S3_ min_tem score 1 point higher on mean ingfulness will -on average-score 0.113 and S1_hu mid ity_min score 1 point h igher on meaningfulness will -on average-score 0.029 . If DM Y and S1_humidity_ min become 0 then there exists a positive value 1.142 which is the intercept term of the model. Importantly, all predictors contribute positivel y (rather than negatively) to yield. This makes sense because they are all positive work aspects. All predictors are highly statistically significant (p = 0.000), which is not surprising considering our large sample size and the stepwise method we used.

Model 4:
The predictors are meaningfulness and so on. This means that if respondent dummy scores 1 unit h igher on mean ingfulness will -on average-score 0.909 points higher on yield, if predictor S3_ min_tem increases by 1 unit, the production increased by 0.103 units, if pred ictor S1_humidity_ min increases by 1 unit then the yield increased by 0.030 units and if the predictor S3_con_rainfall_avg increases by 1 unit then the yield increases by 0.004 units respectively. The trend line for bor o rice production is given in Figure 3. This line pl ot shows that the mean yield of boro rice is increasing day by day except some particul ar year. If DMY, S3_min_ tem, S1_humi dity_ min and S3_con_rainfall_avg become 0 then there exists a neg ati ve value -0.720 which is the intercept term of the model. This happens because of the impact of other variables which were not associated in the model. The predictors are statistically significant (p = 0.000, 0.00, 0.00 0.038), which is not surprising consi dering our large sample size and the stepwise method was used.

Conclusions:-
Bangladesh is one of the vulnerable countries to climate change in the world. The climate change and frequent change in the variability of the climate condition of the country have direct impacts on agriculture and environment in various ways. The objective of this study was to estimate the relationship between Boro rice y ields and climate variables using aggregate-level time series data for the period of 1987 to 2015. The overall findings reveal that four climate variab les min imu m humidity at vegetating phase, min imu m temperature and consecutive days rainfall average at ripping phase and dummy variables have substantial effects on the Boro rice y ield. The result indicates that average seasonal minimu m temperature, minimu m hu mid ity and average rainfall are statistically significant and positively affect the yield of Boro rice. In terms of the R 2 , p-values and F-values for Boro rice models have been found statistically significant. Every crop has an optimu m minimu m and maximu m temperatures, rainfall and humid ity limit for their reproductive and vegetative growth. When temperature exceeded the upper limit or falls below the range or humid ity crossed the upper limit, crop production changes drast ically. Moreover, excessive rainfall may create water logging condition and flooding that also destroys the crop production. Given the high vulnerability of rice yields to climate variat ions in Bangladesh, different adaptation strategies should be adopted to offset the adverse effects of climate change. To minimize the effects of the climat ic variables farmers of the study area adopt some strategies like changing in planting date, digging of ponds and setting up shallow tube wells, selection of short duration species etc. based on indigenous knowledge and resources. Climate change in Bangladesh is a serious concern since it adversely affects agriculture which is an important sector in the country. Therefore, the concerned authority should take appropriate policies to fight against the climate change impact on rice production to ensure food security for the ever increasing population of the country through imp lementing sustainable agricultural development.