Effect of climate change to characteristic of extreme rainfall over Batanghari watershed

Climate change will cause extreme rainfall, both wet and dry. Extreme rainfall will cause catastrophic floods and droughts. Therefore, it is important to analyse extreme rainfall and flood events in the future. Disasters and environmental issues in the Batanghari watershed include flooding, drought, forest and land fires, sedimentation, and water quality degradation. Therefore, the Batanghari watershed is selected as a study area in this research. Research on climate change in the Batanghari watershed is crucial to understanding the trend of extreme conditions now and in the future. Characteristics of extreme rainfall will be analysed using the climate index that was developed by the Expert Team on Climate Change Detection and Indices (ETCCDI). Results show that climate change in the Batanghari watershed is indicated by increasing 5 of 6 extreme rainfall indices: R95p, R99p, Rx1day, Rx5day, and SDII. An increase in extreme rainfall is correlated with an increase in flood events in the Batanghari watershed in similar periods. The extreme rainfall index that most influences flooding in the Batanghari watershed is PRCPTOT and R99p. Flooding is expected to occur more frequently in the future than it does now. The Batanghari Hilir Sub-watershed is more vulnerable to extreme flooding in the future than other sub-watershed.


Introduction
Climate change is indicated by shifts in climate pattern and intensity over a certain period that are usually seen in the average condition for 30 years [1].One of the climate factors that is strongly affected by climate change is precipitation, which can be more extreme, causing drought and flooding [2,3].A study of the effects of climate change on extreme rainfall in Indonesia states that from 1983 to 2012, the trend of extreme rainfall was wetter, and daily rainfall for 30 years has increased by around 0.21 mm/day.In northern Indonesia, the intensity of maximum 1-day precipitation amounts (Rx1day) and extremely wet days (R99p) is also increasing.The result of climate projection has also shown that significantly more consecutive dry days will rise in Indonesia.It means drought will occur more frequently in the future than now [4,5].The islands of Indonesia are at high risk due to climate change, such as Sumatera, Java, and Bali Island.Climate change will affect water availability, flooding, and drought.Hence, the study of climate change scenarios is crucial to devising guidelines for effective adaptation strategies [6,7], and various climate models are used to learn more about climate in the future.
The Batanghari watershed is the second-largest watershed in Indonesia and the largest in Sumatra.It is located in Jambi Province and West Sumatra.Disasters and environmental issues in the Batanghari watershed include flooding, drought, forest and land fires, sedimentation, and water quality degradation.The sedimentation rate and variability of river discharge are higher.Therefore, the Batanghari watershed is selected as the study area in this research.Research about climate change in the Batanghari watershed is crucial to understanding the trend of extreme conditions now and in the future.Characteristics of extreme rainfall will be analysed using the climate index that was developed by the Expert Team on Climate Change Detection and Indices (ETCCDI).This research aims to (1) analyse the characteristics of extreme rainfall, (2) analyse the ability of the RCM model to simulate extreme rainfall events, and (3) investigate extreme rainfall events related to flood events.

Data
This research utilises rainfall data from Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) [8] and 12 models of RCM that have been corrected in a previous study [9].Other than that, flood extent data in paddy fields over 22 years  is also used to investigate the effect of extreme events on flood events obtained from the Ministry of Agriculture, with the assumption that flood events and their extent in non-paddy fields are potentially related to flood events and their extent in paddy fields.Anomaly of each extreme rainfall index was used to calculate trend of historical and projected rainfall [10].The anomaly series were calculated as follows:

Trend of historical and projection extreme rainfall.
IOP Publishing doi:10.1088/1755-1315/1314/1/012010 where Z is the anomaly of index at year i, i X is the index at year i,  ̅ is averaged index and s is standart deviation of index.Trend each extreme rainfall index was evaluated by Mann-Kendall [11,12].The Mann-Kendall test statistic S is given as: where n is the length of the time series xk. ..xn, and sgn(…) is a sign function, xj and xk are values in years j and k, respectively.Sign function is given as: The expected value of S equals zero (E[S] = 0) for series without trend and the variance is computed as: Here q is the number of tied groups and tp is the number of data values in p group.The test statistic Z is then given as: Trend of anomaly extreme index is statistically significant if Z est > Z table (1-/2) [13].

Evaluate extreme rainfall of RCM model.
Performance of RCM models in simulated extreme rainfall were evaluated by graph of Cumulative Distribution Functions (CDFs).Similarity pattern of CDF graph of RCM models and observation in describing the extreme rainfall will decide which model has well or lack performance.Furthermore, the model will evaluate with Mann Whitney U test to know the difference of distribution and number of data RCM models and observation.The model is able to simulate the extreme rain index as observation index if the value asymp sig > 0.05 or Ztable Zest  .

Relation of extreme rainfall to historical and projected flood. Multiple regression is carried out
to analyze the extreme rainfall which affected flooded event in Batanghari watershed.Extreme rainfall index as independent variable and flood extend as dependent variable of regression model that computed as: (6) Where Y is actual flood extent, β0 is intercept of regression model, βi is slope of regression model and Xi is precipitation index, and εi is error of regression model.Extreme index is reliable in estimating flood extent, if regression model is significant at 3 confident level, α=5%, α=10% and α=15% with pvalue<0.05.Then formed regression model will be used to estimate flood extent in future by using 12 RCM models for RCP 4.5 and RCP 8.5 scenario.Independent variable must not be correlated each other (multicollinearity).Correlation between independent variable in multiple regression analysis is undesirable (Daoud 2017).Multicollinearity can be evaluated by Variance Inflation Factor (VIF), VIF < 10 indicates no multicollinearity in the independent variable [14].

Characteristic of Historical and Projected Extreme Rainfall
Studies on the detection of climate change in general use an anomaly of the long-term average of climate conditions.This study shows the trend of six precipitation indices (PRCPTOT, R95p, R99p, Rx1day, Rx5day, and SDII) (Figure 1).Five of six precipitation indices have a distinct upward trend in the period 1986-2015, significantly (pvalue 0.05); only PRCPTOT has an insignificant trend (p value > 0.05) and is debatable.The result indicated that annual total precipitation is insignificantly changing, but consecutive wet days increase year to year, so that extreme rainfall in the Batanghari watershed increases too and causes flood events to rise.A strong relationship between extreme rainfall and flood events is clearly seen in figures 1 and 2. In late 1990, R99p, Rx1day, Rx5day, and SDII in Figure 1 had an increase in anomalies; the flood event anomalies in Figure 2   Historical and projected extreme rainfall in the Batanghari watershed is displayed in figure 3. The average of the projected extreme rainfall index will decrease compared to now in the Batanghari watershed, either in the rainy season or the dry season.Previous study state that in the future, Indonesia is projected to experience drier conditions than now [4].However, the projected maximum value of precipitation indices shows an increase in the rainy season compared to historical values, especially in the Batang Sumai and Tembesi sub-watersheds, while in the Batanghari Hilir subwatershed, almost all indices experienced a decrease in maximum value.In the dry season, the projected maximum value of the extreme index generally decreases, except that Rx5day will increase in all sub-watersheds.It indicates that the Batanghari watershed will be wetter in the rainy season and drier in the dry season.The interesting thing here is that there will be an increase in the maximum value of the future Rx5day for all seasons and all watershed segments.And this needs to be watched out for because it can cause flash flooding.Flash flooding is caused by very intense rainfall.Several studies show that flash floods and landslides occur because of extreme rainfall [16,17].Rx5day is an index that has a higher correlation with flash flood events than other indices [18].

Performace of RCM model in Simulating Extreme Rainfall
Each model has a different performance in simulating daily rainfall.The ability of the RCM model to simulate extreme rainfall was analysed by the CDF plot and compared to the observation base index.Figure 4 shows the CDF for precipitation indices from observation and the RCM model.Figure 4 shows that the model is able to simulate the cumulative distribution of five of six extreme rainfall indices as observation indices, namely PRCPTOT, R95p, and R99p.Rx1day and Rx5day with asymptotic sig, each index is greater than 0.05.While SDII has an asymptotic sig < 0.05, it means SDII is unable to simulate extreme rainfall as observed because SDII has the number of rainy days variable in its calculation.The RCM model has a weakness in simulating the index with the number of rainy days precisely on a monthly time scale.In order that bias correction be conducted to the rainfall of all models to minimise gaps between model and observation.This result implies that the RCM's ability to simulate extreme indices well means that it is assumed to be able to project future floods

Relationship between Extreme Rainfall and Flood Events in the Batanghari Watershed
Flood events are strongly influenced by meteorological conditions, especially extreme rainfall.This section will discuss the extreme rainfall that affects flooding using five precipitation indices (PRCPTOT, R95p, R99p, Rx1day, and Rx5day).Figure 5 shows the probability of flood events in the Batanghari watershed, where several major floods occurred in December, January, March, and April, exceeding 21%, with the biggest chance being 27.3% in December and followed by 24.24% in January.The pattern of flood events follows the pattern of monthly rainfall in the Batanghari watershed, namely the Equatorial pattern.Equatorial patterns are characterised by two peaks of rainfall per year that occur in December-January and March-April.The watershed area with the highest chance of flood occurrence is the Tembesi sub-watershed, as shown in figure 5, while the watershed area with the largest number of people affected by floods is the Batanghari Hilir sub-watershed, as presented in figure 6.In the case of the Batanghari watershed, although the Tembesi sub-watershed has the highest flood event, the flood-affected areas are smaller than those in the Batanghari Hilir sub-watershed.Losses due to flooding in the 1989-2010 periods are reported in Table 2. Tembesi Sub-watershed is the area that most frequently floods, which is 56 times for the 1989-2010 periods.It damaged 1640 houses and flooded 32.496 ha of paddy fields.Flood events in the Sumai sub-watershed occurred 43 times, injured 27 people, damaged 1526 houses, and flooded 6.723 ha of paddy fields.The number of flood events in Batanghari Hilir Sub-watershed is less than other sub-watersheds, but the impact is greater: 347 persons were injured, 2327 were evacuated, and 313.419 ha of paddy fields were flooded.It was caused by extreme flooding during the 1989-2010 periods in the Batanghari Hilir watershed.Flooded paddy field data from the Ministry of Agriculture and losses data from the Indonesian National Board for Disaster Management show similarities in flood-prone areas, and the most vulnerable area is the Batanghari Hilir Sub-watershed.Multiple regression analysis has been carried out to estimate the extent of flooding in the Batanghari watershed using the extreme rainfall index as an independent variable and flood extent as a dependent variable.Regression models were only developed for the months with the highest chance of flooding, namely December, January, March, and April (Figure 5).Multiple regression models for each month are presented in Table 3. Extreme rainfall indices that have strong influences on the extent of floods are PRCPTOT and R99p.Similar studies reveal that a significant increase in PRCPTOT has an impact on increased flood occurrences over the Malaysia Peninsula [19]; in Yellow River China, the increase in PRCPTOT also affects flooding.
A significant test of multiple regression that described the relationship between flood extent and extreme rainfall indices is reported in Table 3.The significance of regression models is evaluated at three levels: level, α =5%, α =10%, and 15% with a p value 0.05.The relationship between flood extent and extreme rainfall indices was significant for all confidence levels.Rainfall and floods have a low predictability level due to their unstable nature, so a confidence level of =20% is still accepted for modelling their relationship [20,21].The regression model in Table 3 was then used to estimate the projected extension of flood 12 models of RCM for RCP 4.5 and RCP 8.5 scenarios.Figure 7 shows a boxplot of historical flood extent and projected flood extent for each RCP in the Batanghari watershed.Extreme flood events will be more frequent than now, as seen by the flood distribution that shows many outliers for RCP 4.5 and RCP 8.5 compared to historical floods.The Batanghari Hilir Sub-watershed is projected to experience more frequent extreme floods in the future; it shows more outlier floods than other sub-watersheds.So it can be said that Batanghari Hilir Subwatersherd is the most vulnerable than others.In the future, the average flood extent will increase in all sub-watersheds for each month of flood events, except in Sumai and Tembesi, where it will decrease for December.The result in figure 7 is in line with figure 3, an increase in the maximum value of the extreme index in the rainy season, causing an increased flood in the future.

Conclusion
Climate change in the Batanghari watershed is indicated by increasing five of six extreme rainfall indices: R95p, R99p, Rx1day, Rx5day, and SDII.The increase in extreme rainfall correlates to the increase in flood events in the Batanghari watershed in similar periods.CDFs of R95p, R99p, Rx1day, Rx5day, and SDII can be well simulated by the RCM model, so that RCM is used to project future floods.PRCPTOT and R99p are the most influential extreme rainfall indices for flood events.Flooding is expected to occur more frequently in the future than it does now.The Batanghari Hilir Sub-watershed is more vulnerable to extreme flooding in the future than other sub-watersheds
also showed an increase.An increase in extreme rainfall is indicating climate change in the Batanghari watershed, and land use change due to palm oil plantations aggravates the Batanghari watershed condition[15].

Figure 2 .
Figure 2. Trend of extend flood anomaly in Batanghari watershead in 1989 to 2010

Figure 6 .
Figure 6.Flood extent of Batanghari Watershed based on flood data in paddy field 1989-2010.
Apr Mei Jun Jul Ags Sep Oct Nov Dec Prob. of Flood (%) Sumai Tembesi Hilir

Table 3 .
Multiple linear regression equation between extreme rainfall index and flood extend in Batanghari watershed.