Climate Change and Extreme Events in WL1a Agro-ecological Zone of Sri Lanka: Implications on Coconut Production

The temporal variations in rainfall, temperature and the associated extreme events were examined with the objective to ascertain their potential effects on coconut production in Ratnapura district of Sri Lanka. Temporal variations were measured using Mann-Kendal test and extreme events were examined using climate indices related to daily rainfall and daily temperature. Daily rainfall and minimum and maximum temperature data of Ratnapura representing Agro Ecological Region, WL1a obtained from the Department of Meteorology of Sri Lanka for the period of 1961 to 2015, were used for the analysis. The base period was considered as 1961 to 1990 as recommended by the World Meteorological Organization. According to the results of the Mann-Kendall tests, minimum temperature had a positive trend and maximum temperature had a negative trend. Among sector-specific climate indices, Standardized Precipitation Evapotranspiration Index in the long-time scale and minimum temperature showed a significant positive trend and the value of the 95th percentile of maximum temperature showed a significant negative trend. The other indices related to drought, extreme rainfall, extreme warm days and maximum annual number of consecutive dry days did not show significant trends during the period 1991 to 2015. Despite the latter, the former indicates an occurrence of drought like situations in the future. Adoption of both short and long-term drought control measures in coconut cultivations if they are to be promoted in Agro Ecological Region of WL1a are recommended as prolonged droughts and high temperatures adversely affect production of coconuts.


INTRODUCTION
Natural and human systems across the globe are sensitive to changes in climate (IPCC, 2014). Average global temperature has increased by 0.93± 0.07 0 C in 2009-2018 when compared to 1850-1900, the preindustrial baseline (IPCC, 2018). The direction, frequency and intensity of climate change however differ by location (Duc et al., 2019). While precipitation in the mid latitude land areas in the Northern hemisphere has increased since 1901, long-term positive or negative trend has been observed in other latitudes (IPCC, 2014). According to IPCC (2014), extreme precipitation events and their frequency of incidences have increased over time. Statistically significant increases in extreme precipitation were observed in Belgium (Ntegeka and Willems, 2008) and Italy (Brunetti et al., 2001), and an increase in frequency of extreme precipitation was observed in Bulgaria (Bocheva et al., 2009). In addition, Osborn et al. (2000) reported an apperent increase in daily precipitation intensity in winter in the United Kingdom. In South Asia, the increase in air temperature is a main distinctive feature (Sivakumar and Stefanski, 2018). The region also experiences frequent warm extremes (Sheikh et al., 2015) and extreme floods (Mirza, 2011). Regional projections of Intergovernmental Panel on Climate Change (IPCC) based on AR4 atmospheric ocean general circulation models (AOGCMs or simply -GCMs) suggest a significant acceleration of warming in Asia compared to that was observed in the 20 th century. Hence, both changes in mean of climate variables and extreme climate events should take into consideration in policy analysis (Katz and Brown, 1992). Marambe et al. (2013), Premalal andPunyawardena (2013), De Costa (2008) and Ministry of Environment Sri Lanka, (2011), who examined climatic conditions in Sri Lanka, revealed an increase in both minimum and maximum ambient temperature in most districts and a decrease of rainy days prolonging the dry spells. The variability of seasonal rainfall of Sri Lanka has increased in recent decades according to Nissanka et al. (2011) and there is a warming trend in temperature with occurrence of more frequent droughts according to Premalal and Punyawardena (2013) and Naveendrakumar et al. (2018). An increase in the number of consecutive dry days and a decrease in the number of consecutive wet days are also observed (Ratnayake and Herath, 2005;Premalal, 2009). A strong upward trend in warm days and warm nights has also been observed in Sri Lanka (Sheikh et al., 2015).
Climate change and climate extremes threaten global agriculture reducing food production (WMO, 2018). Several studies have revealed that climatic changes (Rosensweig and Parry, 1994;IPCC, 2007), particularly extreme rainfall and temperature (Wolfram and Roberts, 2009;Burke and Emerick, 2016), have adversely influenced agricultural production. The adverse effects of climate change on crop agriculture is expected to be larger in developing countries in the tropics and the level of damage depends on the actual climatic scenario (Seo et al., 2005).
Coconut is one of the predominant food crops in agricultural systems in Sri Lanka and its production is largely determined by climatic factors, and the government of Sri Lanka considers expansion of coconut cultivation beyond geographical areas that are devoted to coconuts. The objective of this study was to examine the long-term climate trends in rainfall and air temperature in a nontraditional coconut growing district, namely Ratnapura, in the Agro Ecological Region (AER) of WL1a to ascertain whether expansion of coconut cultivation is feasible to such areas.
The paper proceeds as follows. The next section provides background information on coconut cultivation in Sri Lanka in light of climate change. The following section describes the data and methods used in the analysis. The estimates of the trends in climatic variables and trends in extreme events are presented next. The last section concludes the study with policy recommendations.

Effects of rainfall and temperature on coconut production
Coconut (Cocos nucifera L.) cultivation in Sri Lanka is a key agricultural land use that is second only to rice, which is the staple food. Coconut is grown as a rain fed cultivation and highly sensitive to moisture deficit that can have a lagged impact on production spanning over several years. Coconut palm requires a well distributed rainfall of 1,500 -2,300 mm/year, mean temperature of 27-29 0 C with diurnal variation of 5-7 0 C and 2,000 sunshine hours/year with at least 120 hours/month (Child, 1964;Mahindapala and Pinto, 1988). Variations from these optimum climatic conditions influence physiology of the palm and ultimately the yield in mature coconut plantations.
The consequences of adverse rainfall and temperature in coconut continue for several years due to its long economic life span and prolonged reproductive phase. The nut (drupe) takes 44 months to develop from flower initiation to a mature nut at harvesting stage. Every stage of nut development is sensitive mainly to rainfall (moisture) and temperature in varying degrees. Percentage of button nut (female flowers after fertilization) fall from the inflorescence is significantly sensitive to total rainfall and minimum air temperature (Peiris et al., 1995;Peiris and Thattil, 1998). The optimum temperature for in vitro pollen germination of coconuts is 28 0 C while maximum can go up to 39.7 0 C (Ranasinghe et al., 2015). Critical temperature for successful reproduction of coconuts is 33 0 C (Ranasinghe, 2012). Coconut palm experience moisture stress when it is exposed to temperature above 33 0 C and soil water deficit due to rain free period of longer than consecutive two months (Kasturi Bai et al., 2003). The droughts can cause crop failure due to reduced fertilization, button nut fall, immature nut fall, and palm loss due to death of seedlings, immature and mature palms. Accordingly, extreme rainfall and temperature can cause direct production loss in coconuts.
Traditionally, coconut cultivation is concentrated in the North-Western and Western provinces in Sri Lanka. Increase in demand for land due to population increase, urbanization and general development activities, especially in Western province, has changed the economic dependence of the traditional coconut-based industries. With the increasing demand of land in coconut triangle and increasing demand for coconuts, coconut cultivation was extended to nontraditional coconut growing areas outside the main coconut growing areas. The government of Sri Lanka took many steps to expand coconut cultivation in non-traditional coconut growing areas and Ratnapura is one of the few districts that showed an increase in extent under cultivation of coconuts according to the latest Agricultural Census in 2014 compared to 2002 (Department of Census and Statistics, 2018). However, sustainable production of these cultivated areas will be determined by the major climatic factors, rainfall and temperature.
However, there is a dearth in studies to assess the effects of climate change and extreme events on coconut production in nontraditional coconut growing areas (see Fernando et al. (2007) and Pathiraja et al. (2017) for general assessments). The studies to assess the effects of climate change on Sri Lankan agriculture are mainly confined to rice and tea (Wijeratne et al., 2007;Jayathilaka et al., 2012). It is expected that climate change may cause negative effects on coconut plantations due to increase in temperature and limited availability of water. Further, the coconut plantations act as carbon sinks that absorbs atmospheric CO2 which can mitigate climate change (Ranasinghe, 2012). A few studies have attempted to estimate the impacts of extreme events on coconut productivity (Pathmeswaran et al., 2018) but limited to regression analysis between coconut productivity and number of extreme events. However, none has attempted to explore deviations from the optimum growing conditions or thresholds for coconuts.

METHODS OF ANALYSIS AND DATA
As indicated in the previous sections, changes in pattern of climate is studied at global and regional level across space and time (IPCC, 2014) by observing trends and variability in precipitation (Portmann et al., 2009;Karpouzos et al., 2010), temperature, and extreme events (Tan et al., 2017). In this study, trends in precipitation, temperature and extreme events and variability of precipitation and temperature were analysed first. Then, they were compared with the threshold values for coconut.

Trend analysis in climatic variables
Trend analysis is an empirical method that can be used to compute and explain change of a variable over a period of time using available data. Chandler and Scott (2011) considers 'trend' as a long-term temporal variation of statistical properties of a process, where 'long-term' depends on the application. Parametric and non-parametric tests are performed in trend detection and analysis. However, outliers and missing data, which are frequently encountered in meteorological data, adversely affect normality and homogeneity of variance. Non-parametric tests have the advantage over parametric tests for data series that are non-normally distributed and have outliers and missing data.
The time series nature of data may violate some of the assumptions and may give rise to the problems of autocorrelation and heteroscedasticity. In general, this is an undesirable property for many statistical estimations. In meteorological time series data, autocorrelation in observed variables and unobserved error term is possible. Presence of autocorrelation makes the standard errors to be incorrect and therefore the usual hypothesis testing and confidence intervals also become incorrect.
Considering the above limitations, Mann-Kendall (MK) trend test (Mann, 1945;Kendall, 1962) was used to assess long term upward or downward trends in observed climatic variables. This test is widely applied to statistically detect the long-term trend in meteorological time series (Brunetti et al., 2001;De Silva and Sonnadara, 2014;Duc et al., 2019). This is a non-parametric and rankbased test, hence assumptions for distributional properties are not required. The test also allows autocorrelation (Hamed and Rao, 1998). The test calculates the Kendall's Tau statistic and test the null hypothesis that there is no trend existing in the data series under investigation. When the trend parameter is significant, the rate of change can be calculated using the Sen's slope estimator (Campa et al., 2002). In this study, trend analysis was done using R software package.

Extreme event analysis
The climate indices with respect to extreme events related to agriculture sector developed by the World Meteorological Organization Commission for Climatology (CCI) Expert team on Sector -specific Climate Indices (ET-SCI) were used for the analysis. Table 1 provides the list of indices. Data were analysed using ClimPACT2 software package (Alexander and Herold, 2016). In order to compare with the requirement of rainfall and temperature for coconuts, two user defined indices were used, i.e. i) The lowest optimum temperature for pollen tube growth of coconut is 26.02 0 C hence, 26 0 C was selected as the user defined index (TNge26) for temperature and ii) the average daily rainfall for better growth and production of coconuts is 5 mm and hence the annual number of days when rainfall is more than or equal 5 mm (R5mm) was selected as the user defined index for rainfall and included in the selected indices in Table 1.

Standardized Precipitation Evapotranspiration
Index (SPEI) and Standardized Precipitation Index (SPI) at different monthly levels were used in the analysis as composite measures of drought in addition to the indices based purely on rainfall and temperature. ClimPact2 software recommended by the WMO was used for the extreme event analysis.

Variability analysis
Coefficient of variation (CV) was used to compute the variability in rainfall and temperature. Higher value of CV indicates higher variability and vice versa. The decadal variability of rainfall and temperature during the period of 1961 to 2010 namely, 1961-1970, 1971-1980, 1981-1990, 1991-2000 and 2001-2010

Threshold levels for coconut production
In general, well distributed annual rainfall of 1,300-2,300 mm, and annual temperature of 27 o C are required for the optimum growth of coconuts. Temperature beyond 33 0 C causes moisture stress in general, despite that there are varietal differences. Critical temperature for fruit set under both heat (atmospheric drought) and water stress (soil drought) is 33 o C for 'Sri Lanka Tall' cultivar (Ranasinghe, 2012) which is commonly grown in Rathnapura district. Beyond this, the fruit set is reduced. Optimum (Topt) and maximum (Tmax) temperatures for pollen germination and pollen tube growth differ with the variety (Table 2).

Data and Data Sources
Ratnapura area which falls within the AER of WL1a was chosen as the study area as it is one of the non-traditional coconut growing areas where extent under coconut cultivation has been increased in the recent years. Rainfall and temperature were explored in this study, as they are the main climate variables that determine production in tropical climate under rain-fed cultivation.

Results of Trend Analysis
The trends in daily rainfall and temperature in AER of WL1a are presented in Table 3. No significant trend was observed in daily rainfall. This is in line with the findings of Jayawardene et al. (2005) who reported that the trends of annual average rainfall in Ratnapura were not significant. Naveendrakuma et al. (2018) also found no significant trend in daily rainfall of Rathnapura for the period of 1961 to 2015 whereas De Costa (2008) reported nonsignificant trend in rainfall at 5% significance level in Rathnapura for the period of 1860 to 2010.
Further, Nissanka et al. (2011) reported that there was no significant trend in annual rainfall in Rathnapura for the period of 1961 to 2010. The present analysis confirms a significant (P<0.05) decrease in maximum temperature and increase in minimum temperature in the long run. Naveendrakuma et al. (2018) also found a decreasing trend in maximum temperature during the months of April and December and increasing trend in minimum temperature. The same pattern was also confirmed in the report of Sri Lanka's Second National Communication on Climate Change (Ministry of Environment, Sri Lanka, 2011). Maximum temperature in AER of WL1a in last sixty years has gone beyond 36.8 0 C which is well above the critical temperature of 33 0 C for fruit set of coconuts. Decline of this maximum temperature would be favourable for coconut production. Trends in extreme events were analysed using ClimPact2 software considering baseline reference period as , and the results are presented in Table 4. Trends in most of the indices were not significant. Minimum daily temperature or night temperature showed statistically significant increasing trend with 0.009 0 C increases per annum. The highest value of minimum temperature observed during 2011 to 2015 period was 26.9 0 C and average minimum temperature during the same period was 23.15 0 C. Therefore, increasing trend of minimum temperature would not be an issue for the production of coconuts in the near future. This is further confirmed by the insignificance of TNge26, user defined index to capture the trend in temperature beyond 26 0 C which is the lowest temperature value for the reproduction of coconuts. Value of the 95 th percentile of maximum temperature is the other index that showed a significant trend. In contrast to the other significant indices, this showed a negative trend, indicating that the value of the 95 th percentile of maximum temperature is at a declining trend. This can be considered as a favourable condition for the growth of coconuts as this indicates the probability of exposure of coconuts into higher extreme temperature, especially during reproductive stages is low. Plant growth is indirectly linked to temperature, and positive deviations from its optimal value will always be damaging. Precipitation close to water requirement of a particular crop will be beneficial for many crops (Wolfram and Roberts, 2009).
When drought indices were considered, only SPEI at 12 and 24 months were significant with an increasing trend ( Figure  1). The increasing trend in drought indices may have negative effects on coconut production in AER of WL1a if this trend is continued in the future.

Results of the variability analysis
Variability was analysed using Coefficient of Variation (CV). Mean, minimum, maximum and CV were compared between decades ( Table 5). Variability of daily mean rainfall during the study period has increased as indicated by increase of CV in recent decades. Variability of daily maximum temperature and minimum temperature has not changed heavily as indicated by CV.

CONCLUSIONS
The results revealed a significant negative trend in maximum temperature, a positive trend in minimum temperatures, and a nonsignificant trend in precipitation for the period of 1991-2015 when compared to 1961-1990 period. The SPEI of long-time scale and minimum daily temperature showed a statistically significant positive trend while very warm day threshold showed a significant negative trend. Trends with respect to all the other indices were not significant. The negative trend in maximum temperature is a beneficial factor for the growth of coconut as maximum temperature is one of the contributing factors that determines pollen viability. The positive trend in minimum temperature would not cause significant losses as minimum temperature levels in AER of WL1a are below the critical levels. However, the increase of SPEI, which is a drought indicator, constrains coconut cultivation under rain-fed condition. Therefore, shortand long-term measures of drought adaptations at farmer levels in AER of WL1a in Rathanpura district will be required to avoid any potential losses.    Plant Physiology Division, CRI are highly esteemed for providing climate data. Support provided by the National Science Foundation under the project of "Analysis of home-garden ecosystems in the context of climate resilience and biodiversity in different agro-ecological regions of Sri Lanka", under the grant number NTRP/ 217/CC&ND/TA-04/p-02/01, is highly appreciated. Comments from the two anonymous referees are acknowledged.