Effect of Indian Ocean Dipole (IOD) Events on the Second Inter-monsoonal Rainfall in the Wet Zone of Sri Lanka

The climatic variations in the Indian Ocean have a strong relationship with the rainfall anomaly of Sri Lanka. Indian Ocean Dipole (IOD) is an ocean-atmospheric coupled phenomenon associated with an east-west gradient in the tropical Indian Ocean Sea Surface Temperature (SST) anomalies. Identifying the impact of IOD phenomenon on the spatial and temporal variation of the rainfall pattern is a useful tool for seasonal climate forecasting. the of on and IOD-negative the the consecutive mean investigations are to elucidate the El Niño Southern Oscillation (ENSO)-induced anomalous variation IOD

collected. The IOD-positive and IOD-negative years were extracted based on the Dipole Mode Index (DMI) over the neutral years. Five rainfall indices, namely, cumulative rainfall, the maximum rainfall received within a day, number of wet days, heavy rainfall events and the maximum consecutive dry days were statistically analysed. The results revealed a significant positive anomaly of mean cumulative rainfall in the SIM season during the IOD-positive years while in IOD-negative years this anomaly was negative (p<0.05). An apparent increase or decrease of number of wet days and heavy rainfall events was observed during IOD-positive or IOD-negative years, respectively. However, the mean maximum consecutive dry days showed a distinct negative anomaly with the positive IOD events and positive anomaly with the negative IOD events. Further investigations are suggested to elucidate the El Niño Southern Oscillation (ENSO)-induced anomalous variation over IOD impacts.

INTRODUCTION
Indian Ocean Dipole (IOD) is an Oceanatmosphere interaction that may exemplify a characteristic internal mode of the Indian Ocean climate system (Saji et al., 1999;Webster et al., 1999). It is associated with the Sea Surface Temperature (SST) anomaly between two poles in the south-eastern and western equatorial Indian Ocean (Ashok et al., 2001;Yamagata et al., 2002). Similar to the dominant mode of oceanatmosphere interaction in the tropical Pacific; El Niño Southern Oscillation (ENSO), IOD also has three phases, namely, neutral, positive (warm) and negative (cold) (Kousky and Higgins, 2007;McPhaden et al., 2006;Philander, 1985). The positive phase of the IOD (associated with a warm SST) leads to enhanced rainfall in the western Indian Ocean and diminished rainfall in the south eastern Indian Ocean; while the opposite is true for the negative phase (Saji et al., 1999).
In the neutral phase, warm water from the upper equatorial Pacific Ocean is transferred to the Indian Ocean, resulting in a warm SST in the islands of Indonesia. Air rises above this area and falls allowing westerly winds to blow along the equator . During the positive IOD phase, westerly winds weaken along the equator allowing warm water to shift towards Africa. Changes in the winds also allow cool water to rise up from the deep ocean in the east. This sets up a temperature difference across the tropical Indian Ocean resulting in higher rainfall in the countries around the Arabian Sea. Westerly winds intensify along the equator during the negative phase, allowing warmer waters to concentrate near Australia. This sets up a temperature difference across the tropical Indian Ocean, with an aboveaverage rainfall over parts of southern Australia and Sumatra (Yuan et al., 2008;Aparna et al., 2012;Pentakota et al., 2012;BOM, Australia, 2020). An IOD situation is a complex phenomenon with multiple variants, i.e. whether they be called true dipoles, pseudo dipoles, IOD Modoki, etc. (Verdon-Kidd, 2018).
Several studies have been carried out in different countries, especially those closer to the Indian Ocean, to find out the impact of IOD on the rainfall pattern and thereby the weather changes such as floods and droughts. Clark et al. (2003) and Black et al. (2003) reported that the rainfall in October to December, along the coastal belt in Kenya and Tanzania, correlates strongly with SST in the Indian Ocean. Ashok et al. (2001) showed that a positive IOD plays an important role in increasing the Indian summer monsoon (ISMR) rainfall. Furthermore, the IOD influences the inter-annual rainfall variability over central Brazil and subtropical La Plata Basin during the austral spring (Chan et al., 2008). Through limited studies, a considerable influence of IOD events has been observed in the rainfall climatology of Sri Lanka (Zubair et al., 2003;Jayawardene et al., 2015). Sri Lanka has four distinct rainfall seasons, namely, the first inter-monsoon (FIM), southwest monsoon (SWM), second inter-monsoon (SIM) and northeast monsoon (NEM) (Punyawardena, 2020). Based on these four rainfall seasons, the country has two cultivation seasons, i.e. Yala season comprising both FIM and SWM seasons (minor cultivating season; March -September) and Maha season comprising both SIM and NEM seasons (major cultivating season; October to February) (Punyawardena, 2020). During two monsoon seasons, rainfall is mainly received by the southwest and northeast monsoonal winds while during inter-monsoon periods, Sri Lanka receives rains by the convectional activity and its enhanced phase with the passage of the Inter Tropical Convergence Zone (ITCZ) on or near to Sri Lanka (Punyawardena, 2008). Land preparation and crop establishment activities are initiated with the onset of the inter-monsoon rains in each growing seasons. Therefore, the positive and negative anomalies of the rainfall pattern in the inter-monsoon seasons would directly affect the crop production during the Yala and Maha seasons. The SIM is the most important rainy season to the country where the island receives a well distributed rainfall within two months (October and November), which is about 30% of the total annual rainfall (Chandrapala, 2007).
Wet zone (WZ) covers the south-western part of the low lands and central highlands of Sri Lanka, which receives over 2,500 mm of average annual rainfall (Punyawardena et al., 2013). Though the major irrigated paddy-growing areas are not located in the WZ, the rainfed paddy, three major planation crops (tea, rubber and coconut), export agricultural crops, and vegetables are mainly established in the WZ depending on the rainfall pattern (Punyawardena et al., 2003). Further, the WZ precipitation causes a considerable impact on the economic progress of Sri Lanka as the main source of energy is hydropower (Jayawardene et al., 2015). Therefore, it is important to ascertain the rainfall variation in the WZ of Sri Lanka under the effect of IOD in order to enhance the skill of weather predictions enabling to make appropriate on-farm decisions well before the onset of the Maha season.

METHODOLOGY
Daily rainfall data were collected from 16 meteorological observation stations scattered throughout the WZ of Sri Lanka, representing nine different Agro-ecological Regions (AERs), for a period of 44 years from 1976 to 2019 ( Figure 1). Data were sourced from the Natural Resources Management Centre of the Department of Agriculture, Department of Meteorology, and the Tea Research Institute of Sri Lanka.
Several Ocean-atmospheric indices have been used to investigate the IOD events such as SST, Sea Sub-surface Temperature, Sea Surface Heights, surface zonal winds and velocity potential (χ) fields (Ashok et al., 2001;Zubair et al., 2003;Rao and Yamagata, 2004;Chan et al., 2008;Xin-Yu et al., 2015). For the present study, SST defined by the National Oceanic and Atmospheric Administration (NOAA) of the United States Climate Prediction Centre (CPC) was considered (Reynolds et al., 2007). Intensity of the IOD was represented by the anomalous SST gradient between the western equatorial Indian Ocean (50E-70E and 10S-10N) and the south eastern equatorial Indian Ocean (90E-110E and 10S-0N) ( Figure 2).  The largest count of consecutive days that received rainfall <1mm during the season This gradient is named as Dipole Mode Index (DMI). When the DMI is positive, the phenomenon is referred to as the IOD-positive and while a negative DMI is referred to as the IOD-negative (Saji et al., 1999;Webster et al., 1999). The periodically updated east and west SST anomalies, published by Physical Science Laboratory, NOAA (DMI: Standard PSL) was considered for the present study (https://psl.noaa.gov/gcos_wgsp/ Timeseries/DMI/). Different IOD related studies shows a substantial variation in selecting the IOD events depending the spatial and temporal difference (Verdon-Kidd, 2018;Chan et al., 2008;Hong, 2008). For the present study, the operationally used value of above +0.4 °C was considered as the positive IOD events and below −0.4 °C was considered as the negative IOD event (http://www.bom.gov.au/climate/enso/#tabs=In dian-Ocean). The DMI range of -0.4 °C to +0.4 °C is the IOD-neutral period. Based on this threshold values, IOD-positive, IOD-negative and neutral years were identified for the four rainfall seasons of Sri Lanka.
Five rainfall indices were considered to find the effect of IOD events on the rainfall variability of the country (Table 1). One-way ANOVA, Dunnett's test and Cross Tabulation-Chi-square test were used (p=0.05) to investigate the statistical significance, using Minitab statistical software.

RESULTS AND DISCUSSION
The IOD-positive and IOD-negative years during the study period 1976-2019 within the four rainfall seasons are shown in Table 2. There were only 2 or 3 IOD events during NEM and FIM seasons in both positive and negative phases of the IOD. This clearly indicates that statistically acceptable number of IOD events does not exist during the NEM and FIM seasons to proceed with further analysis. Within an IOD year, IOD events usually started around May or June, peaked in October and then decayed in December (Saji et al., 1999).
Among the two rainfall seasons (SIM and SWM) where IOD was effective, only SIM season showed statistically significant observations, with a peak of IOD events within a year. The SWM season did not show any apparent increasing or decreasing trends of IOD events (data not shown). Therefore, the present study was mainly focused on the effect of IOD events on the SIM season. Previous studies have also revealed a positive anomaly of rainfall around the Arabian Sea during October-November period under the influence of a positive IOD (Black, 2005;Black et al., 2003;Chan et al., 2008). Zubair et al. (2003), using monthly datasets from 1869-2000, also reported an enhancement of Maha season rainfall during the positive IOD phase. Ashok et al. (2001) discovered that the ENSO-Indian Summer Monsoon rainfall (ISMR) correlation is low (high), when the IOD-ISMR correlation is high (low).

Comparison of the Mean Cumulative Rainfall
The spatial distribution of positive and negative mean cumulative rainfall anomalies over the neutral years are illustrated in Figure 3. An anomalous variation was prominent in Upcountry WZ and Mid-country WZ mainly during the IOD-positive years.
The average of all 16 locations considered in the study showed a significant increase in mean cumulative rainfall during IOD-positive years while a significant decrease in IOD-negative years (p<0.05) ( Table 3). Among all locations studied, the locations in the WU3 AER showed a significant positive anomaly.
| 291 Table 2: IOD-positive and IOD-negative years within four rainfall seasons during the study period   IOD-positive Years  IOD-negative Years   FIM  SWM  SIM  NEM  FIM  SWM  SIM  NEM   2010 1983 1982 1998 1983 1980 1980 1978 2017 1994 1994 2019 1992 1981 1984 1983 1997 1997 1989 1996 1985 2012  The amount of water needed for crops can be supplied by rainfall, irrigation, or by a combination of both (Brouwer and Heibloem, 1986). Rainfed cultivation is prominent in the WZ as it receives a sufficient amount of rainwater during both cultivation seasons. The quantum of rainwater received is an important factor to be considered when planning a cropping season to select the crop variety and the extent of cultivation. The present study revealed that the quantum of rainfall received within the SIM period in the WZ is in the range of 430 -867 mm. During a positive IOD period, it has increased from 525 -1,120 mm and decreased to a range of 370 -825 mm during IOD-negative years (Table 3). Hence, if a positive IOD phase is predicted in a given year, excess soil moisture conditions in uplands and water-logging and/or flooding in low-lands can be expected during the SIM season.

Comparison of the Mean Maximum Rainfall per day
The mean values of the maximum rainfall received per day did not show a significant positive or negative anomaly during the IOD-positive period (Table 4). However, the IOD-negative years showed below average values for every selected location except at Gampaha-Henarathgoda.
According to the study, the mean maximum rainfall received per day during a neutral year ranged from 60 to 130 mm. This range decreased to 51 -124 mm during an IOD-negative phase ( Table 4). Receipt of the maximum rainfall per day in a given location depends on several factors such as atmospheric pressure, wind speed and sheer, convectional activity, the position of the ITCZ, occurrence of low level atmospheric disturbances, and cyclonic storms resulting in high variability of daily rainfall during the SIM season (Punyawardena, 2008). Therefore, the impact of IOD can be masked by these factors.

Comparison of the Number of Wet Days
There was an association between IOD-phases and the number of wet days in all selected locations (p=0.05) except in five locations, namely, Talawakele, Pussallawa-West Hall, Bombuwela, Labuduwa and Galle for IOD-positive years (Table  5). On average, 55.6% of the number of wet days were observed during the neutral years (range: 46.2% -63.6%). This has increased to 62.9% with the positive impact of IOD phase (range: 56.4% -72.5%) and decreased to 50.8% during an IODnegative period (range: 37.7% -67.6%).
Distribution of rainfall received during the season is also an important information to find out whether the crop water requirement is being met by rainfall alone. The reduction of rainfall during the IOD-negative phase may not affect the crop performance in the WZ because of the prevailing relatively low evapotranspiration regime of the region. Monthly evapotranspiration rate ranges from 45.6 mm to 67.0 mm during SIM in the WZ (DOA, 2018). The present study also revealed that more than 15 consecutive days of receiving rainfall of < 1 mm was hardly observed even during the IOD-negative phase in the SIM season (Data not shown).

Comparison of heavy rainfall events
There was an association between IOD-phases and number of heavy rainfall events in all considered locations (p=0.05), except at Pussallawa-West Hall, Ratnapura, Bombuwela, Labuduwa and Galle for IOD-positive years (Table 6). On average, 10.7% of the heavy rainfall events were observed during the neutral years (range: 8.6% -11.8%). This has increased to 15% with the positive impact of IOD phase (range: 11.9% -19%) and decreased to 8.5% during an IOD-negative period (range: 5.7%-11.5%).   Black et al. (2003) has reported that extreme short rains are associated with increasing IOD events in East Africa. Heavy rainfall events can be considered as receiving intense rains within a short period of time. Receiving >25 mm of rainfall per hour is considered as an "erosive rain" (Punyawardena, 2008). High intense rains result in soil erosion especially in the sloping terrain in the WZ (i.e., Mid and Up country WZ). Therefore, if an IOD event is predicted, agro-met advisories may emphasize on the importance of adopting soil | 295 conservation measures in landscapes that are prone to soil erosion and land degradation.

Comparison of the Mean Maximum Consecutive Dry Days
An apparent reduction of the mean value of the maximum consecutive dry days was observed with the positive IOD events in all selected locations. Colombo and Ratmalana in WL3 AER showed a significant decrease of the maximum consecutive dry days (p<0.05). During the IODnegative period, 14 out of 16 locations showed an increasing trend in the maximum consecutive dry days, where a significant increase (p<0.05) was reported from Ratnapura (Table 7).
In the IOD-neutral phase, the maximum consecutive dry days ranged from 7 to 11, which decreased to 4-7 days during the IOD-positive phase, and increased to 6-12 days during the IOD negative phase (Table 7). In general, the maximum consecutive dry days is not a meaningful indicator for the WZ as rainfall has exceeded the evapotranspiration demand (DOA, 2018).
The IOD cannot be viewed in isolation from the ENSO. Instead, it is assumed that in some years, a strong ENSO forcing can create a bias in the Indian Ocean-coupled system (Ashok et al., 2001;Black et al., 2003;Zubair et al. in 2003;Chan et al., 2008). As for the SIM season considered, five El Nino years and one La Nina year were observed during the IOD-positive period and one La Nina event was recorded during the IOD-negative period (Abeysekera et al., 2017).
A positive anomaly of rainfall has been recorded during the El Nino years in the WZ of Sri Lanka for the SIM season while a below normal rainfall was recorded during the La Nina period (Hapuarachchi and Jayawardena, 2015;Abeysekera et al., 2019). The locations and periods used in the previous studies (Abeysekera et al., 2017;2019) were the same as in the present study. Therefore, further investigations are needed to understand the interaction effects of IOD and ENSO on the all four rainfall season of Sri Lanka.

CONCLUSION
The present study identified the impact of the IOD events on the rainfall climatology of second intermonsoon seasonal rainfall in the Wet zone of Sri Lanka. A robust statistical analysis revealed an apparent increase in cumulative seasonal rainfall, mean number of wet days and heavy rainfall events with the positive IOD events. The opposite of this was observed during IOD negative events. No discernible changes were observed for the mean maximum rainfall received per day within the season. The mean maximum consecutive dry days showed a distinct negative anomaly with the positive IOD events and positive anomaly with the negative IOD events. The influence of the IOD phenomenon on the Wet zone during the SIM season was evident in this study suggesting its usefulness as a tool for seasonal climate forecasting and agro-met advisories.