Climate Change Impact Assessment for Aji Basin Using Statistical Downscaling and Bias Correction of Climate Model Outputs

For the future projections Global climate models (GCMs) enable development of climate projections and relate greenhouse gas forcing to future potential climate states. When focusing it on smaller scales it exhibit some limitations to overcome this problem, regional climate models (RCMs) and other downscaling methods have been developed. To ensure statistics of the downscaled output matched the corresponding statistics of the observed data, bias correction was used. Quantify future changes of climate extremes were analyzed, based on these downscaled data from two RCMs grid points. Subset of indices and models, results of bias corrected model output and raw for the present day climate were compared with observation, which demonstrated that bias correction is important for RCM outputs. Bias correction directed agreements of extreme climate indices for future climate it does not correct for lag inverse autocorrelation and fraction of wet and dry days. But, it was observed that adjusting both the biases in the mean and variability, relatively simple non-linear correction, leads to better reproduction of observed extreme daily and multi-daily precipitation amounts. Due to climate change temperature and precipitation will increased day by day. keywords: Indian future climate; Precipitation; Temperature; Bias Correction; downscaling.


INtRODUCtION
Now a days most serious challenges faced by mankind is climate change.For the assessment of future variations in the hydrologic cycle, the sensitivity of regional hydrology to variable climate conditions makes climate-change projections essential.Validation of future climate change prediction it is important to investigate observed change in present climate and can be put into context.Information gap that is provided by climate can be linked by Bias correction methods.For climate impact research, and simulation of data statistical downscaling is broadly applicable 2, 4, 5, 10 .During the historical reference period it facilitates the comparison of simulated and observed impacts and a continuous transition into the future.Bias correction helps to adjust the simulated climate data to the more detailed altitudestratified information related with observational data.Conversely, there are several limitations of statistical bias correction like.When applying the bias correction to future periods, Stationary in bias in the historical data with respect to future data is assumed, which introduces additional uncertainty 7 .The quality of the bias-corrected simulation data, is limited by both the observational dataset and the exemplification of physical processes 8 .

Methodology (Study area-Aji river basin)
Saurashtra has over 100 river basins; among these-Bhadar, Aji, Shatrunji and Machchhu.Aji is the most important river of Saurashtra.Average annual rainfall for Rajkot district is 552 mm for the years1961-2007.The Aji river passes through the city of Rajkot.It is situated between latitude 21ºto 22º N and longitude of 70ºto 71º E. Aji river length is 164 km with 2130 km 2 catchment area.Some of the major tributaries of Aji are the Nyari, lalapari, khokaldadi, Banked and the Dondi.The River originating from hills of sardhar near Atkot, to its mouth at the Gulf of kutch in Ranjitpara of Jamnagar district .There are four dams on Aji River.

Climate Input Data
The daily temperature and rainfall data simulated by CGCM 2.3.2 for two grid points falling in/nearby Aji basin were used for the bias correction and were used based on the represented area using theissen polygon method.The weather data for the future scenarios were 2046-64 and 2081-2100.The hydrological and meteorological data like daily rainfall data, temperature (Max.and Min.) were collected from Meteorological Observatory of Main Dry land Agricultural Research Station, JAU, Targhadiya and State Water Data Center; Gandhinagar.The hydro meteorological data for the future scenarios were obtained from the IITM, Pune.The weather data was obtained through 1, 3 .

the trend-preserving bias correction methods
Employing transformation algorithm, correcting systematic error in RCMs grid point simulated climate variables called bias correction methods.Out of six bias correction methods are employed to adjust RCMs grid points simulations a power transformation method was used for the precipitations data and liner scaling and variance scaling were used for the temperature data 9, 6, 7.

Bias Correction of Monthly Mean Data of temperature (Maximum and Minimum)
Adjustment done for observed monthly mean data and long-term differences between the simulated data during the historical period and unchanged daily variability about the monthly mean.

l i n e a r S c a l i n g o f P r e c i p i t a t i o n a n d temperature
Monthly correction based on differences between observed and present-day simulated values this approach works 12 .

Power transformation of Precipitation
A non linear correction in an exponential form a, Pb 11,12,13 can be used to specifically adjust the variance statistics of a precipitation time series because it does not allow differences in the variance to be corrected.... (7)   ... (8)   Thus, scaling parameter depends on b, but not vice versa 12 .

Variance Scaling of temperature
Power transformation is an effective method to correct both the mean and the variance, but is limited to precipitation time series.Another approach to correct both the mean and the variance of temperature time series stepwise was presented by 11,12,13 .Means of the RCM-simulated time series are adjusted by linear scaling (Eq.( 3) and ( 4

ReSUltS AND DISCUSSIONS Precipitation for Control Scenario (1981-2000)
The Figure 2shows that the monthly mean of RCMs simulated precipitation higher than that of observed during monsoon months indicating over estimation by CGCM 2.3.2RCM.The raw RCMs precipitation had a positive (Over Estimated) bias from June to August.however, for the rest of the months, the RCM simulation was matched closely with observation.In fact, after bias correction, the RCM simulated precipitation was exactly matched with observation.It indicates that the power transformation method of the bias correction of RCM simulated precipitation is good for correcting the mean precipitation.precipitation on monthly window by comparing with actual observation during the control period.Figure 3 shows that the RCM simulated precipitation has less variability than actual observation during June to November months.The variability of the daily rainfall simulated during the rest of the months was well matched with that of observation.

Precipitation for Future Scenarios (2046-2064 and 2081-2100)
It was observed that RCMs has predicted a higher precipitation during April, May, June and July months for both scenarios.In fact, for the rest of the months, the corrected and un corrected monthly daily mean precipitation was found nearly equal.After bias correction, the precipitation amounts were reduced

Minimum temperaturefor Control Scenario (1978-2000)
The Figure 6 shows that the monthly mean of RCMs simulated precipitation higher than that of observed during monsoon months indicating over estimation by CGCM 2.3.2RCM.The raw RCMs simulated are still overestimated (Positive bias) during January to June months.however, for the rest of the months, the RCM simulation was matched closely with observation.In fact, after bias correction, the RCM simulated minimum temperature was exactly matched with observation.When the analysis was performed using raw data without bias correction, RCMs showed a large amount of disagreements (January-July) for both scenarios.

Minimum temperaturefor Overall Scenarios (1961-2000, 2046-2064, 2081-2100)
Overall we showed that the Temperature is increased day by day due to global warming.
And its directly affected to the climate change impact(Fig.8).

Maximum temperaturefor Control Scenario (1978-2000)
Figure 9 shows that the RCMs simulated are still overestimated (positive bias) during January-April and underestimated (negative bias) during May, June, July months.It can be seen that the RCMs simulated the higher maximum temperature during February, March and September while lower during May-July, and November-December.During the rest of the months, the uncorrected maximum temperature did agree with the bias corrected.The maximum positive bias was found during March while that of highest negative was found during June month during 2046-64.(fig.10(a)).
While in 2081-2100 it can be seen that the RCMs simulated the maximum temperature higher during January-April and August-September while lower during rest of the months.(Fig. 10(a, b)).Teutschbein, Claudia, and Jan Seibert."Bias correction of regional climate model simulations for hydrological climate change impact studies: Review and evaluation of different methods", Journal of hydrology, 2012.
First, b is identified by matching the coefficient of variation (CV) of the corrected daily RCM precipitation (P b ) with the CV of observed daily precipitation (P obs ) for each month m: find b m such that ...(4) ...(5) ...(6) Using Brent's method 8 it is done with a root-finding algorithm.long-term monthly mean of observed precipitation matched with the monthly mean of the intermediary series P *1 contr (d) by using standard linear scaling parameter: )).The mean-corrected control(T -1 control (d)) and scenario runs (T -1 scen (d)) are shifted on a monthly basis to a zero mean.standard deviations (of the shifted time seriesT* 2 contr (d) and T* 2 scen (d))are scaled based on the ratio of observed r and controlrun.

Figure 3
Figure 3 depicts the comparison of the coefficient-a and exponent-b for correcting the biases in the mean and CV of RCM simulated daily

Fig. 2 :
Fig. 2: the comparison of monthly mean of observed, raw and bias corrected RCM simulated daily precipitation during the control period 1981-2000 for the Aji basin

Fig. 4
Fig. 4 (a,b): Comparison of monthly mean of bias corrected and uncorrected daily precipitation during future scenarios 2046-2064 and 2081-2100 for the Aji River

Fig. 11
Shows that the highest increase in the maximum temperature in the future can be during the December to March.This can affects the cereal crops sown during the winter season.

Fig. 10 (Fig. 11 :
Fig.10(a,b): Comparison of monthly mean of bias corrected and uncorrected daily maximum temperature during future scenarios 2046-2064 and 2081-2100 for the Aji River