Twenty first century climatic and hydrological changes over Upper Indus Basin of Himalayan region of Pakistan

Daily meteorological observed station data for the period 1976–2010 of maximum temperature, minimum temperature, and precipitation were acquired from the Pakistan Meteorological Department (PMD). This data set were further utilized for the calibration and validation of a University of British Columbia (UBC) model for 1995–2004 and 1990–1994, respectively, and for bias correction of the climate model's precipitation and temperature time series for the period of 1976–2005.

An observed river flow data for the period of 1990–2004 from Besham Qilla gauging station, which is located 80 kilometers (Km) upstream of the Tarbela reservoir, was acquired from the Water and Power Development Authority (WAPDA), Pakistan. This data are sub-divided into two parts (1995–2004 and 1990–1994), and they are used for calibration and validation of the hydrological model. For the simulation of river flow, the UBC model has the limitation of (not more than) five meteorological stations data and one river flow data. We have eight meteorological stations; therefore, to make a total of five meteorological stations, the data are aggregated by averaging two stations nearest to each other. The averaging criterion is based on the closeness of latitude, longitude, and elevation of the stations. The detail of the meteorological stations is given in table 8.

The output data of climate models for the period 1976–2005, 2006–2035, 2041–2070, and 2071–2100 of CCAM and 2041–2050 and 2071–2080 of RegCM were also used for the extraction of the time series of temperature and precipitation for the corresponding location of meteorological stations. These data were used to compare with observed station data for the purpose of bias correction and then used as an input to the hydrological model to acquire the future river flow projection. The resolution of these data is 50 km and 60 km, respectively. The climate model's output datasets are based on future emission scenarios of the Representative Concentration Pathways (RCP) (van Vuuren et al 2011), RCP4.5 (Clarke et al 2007) with 4.5 W m⁻², and RCP8.5 (Riahi et al 2007) with 8.5 W m⁻² radiative forcing by 2100.

Climate Research Unit (CRU TS3.10, Harris et al 2014) data are used for the general graphical representation of mean seasonal variation, along with the climate model's data. These data includes monthly mean temperature and precipitation for the time period of 1976–2005 over the land at regular global 0.5 degree intervals.

3.2.1. CCAM

3.2.2. RegCM

3.2.3. UBC watershed hydrological model

3.2.4. Snowmelt module in the models

The first phase (phase 1) is based on the CCAM output with RCP4.5 and RCP8.5 lateral boundary conditions (LBC) from the global model (MPI ESM) for the period of 1976–2005 as a baseline and 2006–2035, 2041–2070, and 2071–2100 as future periods. In the second phase (phase 2), we first take the RegCM model output for 2041–2050 and 2071–2080 with RCP8.5 LBC from MPI ESM and second take the 2071–2080 with RCP8.5 and RCP4.5 LBC from Geophysical Fluid Dynamics Laboratory (GFDL). The time slice of RegCM output is decadal (10 years each) and less than that of CCAM (30 years each). However, the RegCM time slices are in the period range of CCAM and are comparable. The main consideration and discussion of this study is based on the results of phase 1. The results of the second phase (phase 2) are only used for a general comparison of the two different models (CCAM and RegCM). To minimize the differences between the actual climate (observed) and model (CCAM) data, the model simulation for 1976–2005 was first compared with observed station data and then the biases were corrected using Best Easy Systematic (BES, Wonnacott and Wonnacott 1972) and Mean Monthly Correction Factor (MMCF) statistical methods. The same bias correction assumption is then applied to model data for future periods of 2006–2035, 2041–2070, and 2071–2100. The same process is repeated for RegCM data. The hydrological model was first calibrated and validated with observed meteorological data and observed river flow data, before applying bias-corrected data as an input for the future hydrological projections. The general methodology of this work is demonstrated graphically in figure 2.

Zoom In Zoom Out Reset image size

3.3.1. Bias correction

The dynamically downscaled data could lead to unrealistic hydrological results without the application of bias correction (Akhtar et al 2008, Bergstrom et al 2001, Graham et al 2007, Haerter et al 2011). Statistical bias correction corrects the biases in the simulated climate time series corresponding to the observed time series. The process of bias correction typically preserves the difference (climate change signal) between baseline (e.g. 1976–2005) and future periods (e.g. 2006–2035, 2041–2070, and 2071–2100). There are several methods to implement bias correction. First, the Best Easy Systematic (BES) (Wonnacott and Wonnacott 1972) estimator is used for both temperature and precipitation. Although the BES method is good for temperature, its performance is low for precipitation. Afterwards, the Mean Monthly Correction Factor (MMCF) method is used for precipitation bias correction.

According to Woodcock and Engel 2005, among linear methods, the BES method is better to compute because it is a robust technique and is simple to apply. Mathematically the BES estimator is represented as:

$$\begin{eqnarray}&&{\rm BES}=[{\rm Q}1+2\times {\rm Q}2+{\rm Q}3]/4.\end{eqnarray} \tag{ 1 }$$

The first, second, and third quartiles of the bias series between simulated and observed data are represented by Q₁, Q₂, and Q₃. BES is used for future data and is basically computed for the baseline data. The difference of BES estimator and RCM's output data are calculated at the end to avoid bias. The MMCF method was computed for bias correction of already downscaled precipitation data. The MMCF method has been extensively used in many studies (Teutschbein and Seibert 2010, Lafon et al 2012).

The transformed precipitation is given by the relation:

$$\begin{eqnarray}&&{\rm R}*={\rm k}\times {\rm R}.\end{eqnarray} \tag{ 2 }$$

The model's daily output precipitation after the bias correction is represented by R^*, R is the model's daily output precipitation before the application of bias correction technique, and the mean monthly correction factor is represented by k and is calculated by:

$$\begin{eqnarray*}&&\begin{array}{lllllllllllllll} {\rm k}=\left[ {\rm Monthlymeanofbaselineobserveddata} \right./ \\ \quad \;\;\;\left. {\rm Monthlymeanofbaselinesimulateddata} \right]. \\ \end{array}\end{eqnarray*}$$

3.3.2. Calibration and validation of the UBC model

Calibration of the hydrological model is an important step because the model cannot be directly used for a basin wide study without calibration (Schuol and Abbaspour 2006, Haerter et al 2011). In calibration, the observed daily station data of temperature and precipitation are used as an input to the hydrological model and the model river flow is compared with the observed daily river flows. The sensitive parameters of precipitation gradient factors (P0GRADL, P0GRADM and P0GRADL), adjustment to precipitation (P0GRADL and P0RREP), groundwater percolation (P0PERC), fraction of impermeable area in the band (C0IMPA), groundwater percolation (P0PERC), deep zone share fraction (P0DZSH), time constant for upper groundwater runoff (P0UGTK), and time constant for deep groundwater runoff (P0DZTK) are tuned for future river flow projections. The performance of the model was statistically validated against the Nash-Sutcliffe coefficient of efficiency (CE). This method is usually applied to the calibration of a hydrological model (Bardossy 2007, Shi et al 2008, Westerberg et al 2011). The statistical equation for this method is given below:

$$\begin{eqnarray}\begin{array}{lllllllllllllll} {\rm Ceff} & = & 1-\frac{\sum \limits_{i=1}^{n} {{\left( {{Q}_{oi}}-{{Q}_{si}} \right)}^{2}}}{\sum \limits_{i=1}^{n} {{\left( {{Q}_{oi}}-{{{\bar{Q}}}_{o}} \right)}^{2}}} \\ {} & {} & \left( -\infty \leqslant {\rm Ceff}\leqslant 1 \right) \\ \end{array}\end{eqnarray} \tag{ 3 }$$

where,

$$\begin{eqnarray*}&&{{\bar{Q}}_{o}}=\frac{1}{n}\sum \limits_{i=1}^{n} {{Q}_{oi}}\end{eqnarray*}$$

n represents the total timestep numbers, Q_oi represents the observed water flow of ith timestep, whereas Q_si represents the simulated water flow of ith

The other calculations involved in this study to evaluate the simulated change in water river flow (R²) are represented by the following equations.

$$\begin{eqnarray}&&\begin{array}{lllllllllllllll} {{R}^{2}}=1-\frac{\sum \limits_{i=1}^{n} {{\left( {{Q}_{o\,i}}-\left( b\times {\mkern 1mu} {{Q}_{S\,i}}+a \right) \right)}^{2}}}{\sum \limits_{i=1}^{n} {{\left( {{Q}_{O\,i}}-{\mkern 1mu} {{{\bar{Q}}}_{o}} \right)}^{2}}} \\ \quad \left( 0\leqslant {{{\rm R}}^{2}}\leqslant 1 \right). \\ \end{array}\end{eqnarray} \tag{ 4 }$$

Where,

$$\begin{eqnarray*}\begin{array}{lllllllllllllll} a & = & \frac{1}{n}\left\{ \sum \limits_{i=1}^{n} {{Q}_{oi}}-{\mkern 1mu} b\sum \limits_{i=1}^{n} {{Q}_{si}} \right\} \\ b & = & \frac{\sum \limits_{i=1}^{n} \left( {{Q}_{oi}} \right)\left( {{Q}_{si}} \right)-\frac{1}{n}\sum \limits_{i=1}^{n} {{Q}_{si}}{\mkern 1mu} \sum \limits_{i=1}^{n} {{Q}_{oi}}}{\sum \limits_{i=1}^{n} {{\left( {{Q}_{si}} \right)}^{2}}-\frac{1}{n}{{\left( \sum \limits_{i=1}^{n} {\mkern 1mu} {{Q}_{si}} \right)}^{2}}} \\ \end{array}\end{eqnarray*}$$

The model perfectly captures the observed river flow pattern if the value of Ceff is equal to one. If the value of R² approaches to one, then the observed river flow is perfectly captured by the model. R² is the correlation coefficient used for comparison of the time and shape of the simulated river flow's hydrograph with the observed river flow. Ceff tells how accurate the estimated hydrographs' volume, time, and shape is to that of the observed hydrograph.

UIB climate varies significantly over time and space due to its complex topography, weaker summer monsoon, and higher rainfall during winter and early spring due to western disturbances. These characteristics make this region distinct from other regions in South Asia.

Figure 3 shows the seasonal mean air temperature of CCAM for 1976–2005, 2006–2035, 2041–2070, and 2071–2100 along with CRU for 1976–2005. The model surface air temperature during 1976–2005 is generally in agreement with CRU for seasonal variations. However, there are some biases that underestimate the temperature, such as the summer simulation to east of 76°E. Furthermore, the southern areas, which are dominated by the summer monsoon, show a smaller difference with some overestimation than the northern parts where the difference is larger and underestimated. The same phenomena are also found when the results of the model were compared with the station observed data. This indicates that the model's performance is passive at the northern parts and shares the common problem of cold biases over UIB, which exists throughout the year (tables 1–2). The possible reason for the cold bias may be: (1) the difference in elevation between the model and actual station values; or, (2) the sudden variation in the topography of this mountainous region, such as the windward and leeward problem. The model's results also show warming during spring as well as in winter for the whole period. The monthly seasonal mean of maximum temperature and difference of temperature from the baseline (1976–2005) to the future (2006–2035, 2041–2070, 2071–2100) is shown in tables 1 and 2. In RCP8.5, the increase in temperature for 2006–2035 is 2.2 °C, for 2041–2070 it is 4.2 °C, and for 2071–2100 it is 5.8 °C. In RCP4.5 the increase in temperature for 2006–2035 is 0.5 °C, for 2041–2070 it is 1.5 °C, and for 2071–2100 it is 2.0 °C. The increase in minimum temperature is high in both scenarios for all future periods, which is an important parameter for the sustainability and growth of agriculture in the region. The temperature increase is higher in the spring (MAM) and winter (DJF) than in the summer (JJA), except in the northern parts where the increase in summer temperature is higher in RCP8.5. A comparison of RCP8.5 and RCP4.5 shows that the average temperature increase is more than double in RCP8.5 than RCP4.5. It is very important to note that in RCP4.5 the temperature increase is relatively slow in 2071–2100 as compared to 2041–2070. The RegCM results also show an increase in future temperature and are in agreement with CCAM results. During 2041–2050 the increase in minimum temperature is 1.7 °C and the maximum temperature is 1.9 °C. For the period 2071–2080, the increase in minimum and maximum temperature is 4.3 °C and 4.4 °C, respectively, whereas the total increase in temperature is 1.8 °C and 4.3 °C for 2041–2050 and 2071–2080 periods, respectively. In RegCM, the average temperature increase is more in RCP8.5 than RCP4.5. The temperature increase is higher in spring (MAM) and winter (DJF) as compared to summer (JJA). Thus, the results provide evidence that there are significant changes in maximum and minimum temperatures in both models with RCPs, which indicate the signals of future climatic change over UIB.

Zoom In Zoom Out Reset image size

Figure 4 illustrates seasonal mean precipitation of CCAM for 1976–2005, 2006–2035, 2041–2070, and 2071–2100. It is found that CCAM overestimated the precipitation for all the seasons especially in Autumn (SON). There is less precipitation during the summer monsoon season, which indicates a weaker monsoon. Overall the occurrence of precipitation is highly variable, as compared to the consistent temperature increase in this region. The projected precipitation increase in RCP4.5 and RCP8.5 is almost same as the 10–11% for 2006–2035 and the 13% for 2041–2070. For 2071–2100, although there is still a difference between the two RCPs, it continuously increases, with 12% (slightly lower by comparison of 2041–2070) and 20%, respectively (table 3). Projected precipitation by RegCM also shows increasing trends for future time periods, which shows a 14% and 23% increase for 2041–2050 and 2071–2080, respectively. By 2080, the precipitation difference of RCP4.5 and RCP8.5 in RegCM is not very promising. The results from Kulkarni et al (2013) over UIB also show a 20–40% increase in 2071–2098 for summer monsoon precipitation from the baseline period (1961–1990). There is also an increase in the precipitation of meteorological stations of the northern part (i.e. Garidupatta, Astor, Bunji, Skardu, Gilgit, and Gupis) but a decrease in the southern part stations (Kakul and Balakot), particularly in winter.

Zoom In Zoom Out Reset image size

The CCAM model data for the baseline period (1976–2005) is compared with observed station data, and differences between model and observed data are found. To minimize these differences, the statistical techniques of the BES estimator method for temperature and the MMCF method for precipitation is used. The results before and after bias corrections are shown in figures 14–16, which clearly show that: BES is best for temperature bias correction, the performance BES is low for precipitation, while MMCF performance is very good for precipitation. The same bias correction assumptions are applied to 2006–2035, 2041–2070, and 2071–2100. It is clear from the analysis that bias correction methods remarkably minimized the differences between observed and modeled data, which almost overlaps the observed and bias corrected curve. These biases-corrected time series data are used as inputs to the hydrological model for impact assessment.

For calibration and validation, the observed river flow data are divided into two parts: first, 1995–2004 is used to calibrate UBC model; and second, the data from 1990–1994 are used for validation. The model parameters mentioned in the methodology were adjusted for calibration. The results of calibration for 1995–2004 with mean monthly flow of 2320 cumecs and the observed monthly flow of 2470 cumecs are shown in figure 5. The statistical analysis shows the higher efficiency of UBC of 0.9 (Berenguer et al 2005) for the study region using Nash-Sutcliffe statistics. The calibration river flow results of the model are closer to the observed river flow in summer than winter. The calibrated model was also validated before simulation for future projection of river flow for 1990–1994. For validation, the efficiency statistics are even higher than the calibration period.

Zoom In Zoom Out Reset image size

The bias-corrected outputs of the CCAM and RegCM climate models were used as an input to the calibrated hydrological model for the river flow simulation. For CCAM, simulations of 1976–2005 and projections of 2006–2035, 2041–2070, and 2071–2100 were carried out (based on MPI-ESM with RCP8.5 and RCP4.5). The period of 1976–2005 was considered as a baseline for river flow. The results suggest that total river flow is continuously increasing over time in the Upper Indus River. The increasing rate of river flow in both RCPs is enhanced during the first two time slices (2006–2035 and 2041–2070), while it is smaller and not in the same ratio during the last time slice (2071–2100) in RCP8.5 and it even decreases for RCP4.5 when compared with 2041–2070. In the summer for RCP4.5, the increase is 24% during 2006–2035, 32% during 2041–2070, and 26% during 2071–2100. In RCP8.5, the percentage increase is 23% during 2006–2035, 50% during 2041–2070, and 55% during 2072–2100. The results show that percent increase is higher in winter than summer in both scenarios. Maximum river flow occurring in summer shows that the highest river flow in RCP8.5 is of 9720 cumecs for the period of 2006–2035, 11 837 cumecs for the period of 2041–2070, and 12 222 cumecs for the period of 2071–2100. In RCP4.5 the highest river flow in summer is 9783 cumecs for the period of 2006–2035, 10 428 cumecs for the period of 2041–2070, and 9954 cumecs for the period of 2071–2100. The projection for increased river flow was higher in RCP8.5 than RCP4.5, mainly due to a significant increase in temperatures. The details of mean river flow values input from CCAM are listed in tables 4 and 5, and shown in figures 6(a) and (b). In RegCM, the increase in river flow during summer is 36% in 2041–2050 and 56% during 2071–2080. Maximum river flow occurs in summer and shows the highest rate of river flow, that is: 7327 cumecs for the period of 2001–2010, 8602 cumecs for the period of 2041–2050, and 9926 cumecs for the period of 2071–2080. The average river flow was measured as 2209 cumecs for the period of 2001–2010, 3328 cumecs for the period of 2041–2050, and 4050 cumecs for the period of 2071–2080. The details of mean river flow values input from RegCM are listed in table 6 and shown in figures 6(c) and (d). For both models, the minimum river flow occurs in the months of December-February when river flow is in the range of 70 to 700 cumecs. The percentage increase is higher in winter than in summer. In RegCM, the comparisons of RCP4.5 and RCP8.5 for the future inflow projection with time period 2071–2080 are quite close to each other and show almost similar results in the months of September-May. However, RCP8.5 shows more increase than RCP4.5, with a significant difference in the months of April-August.

Zoom In Zoom Out Reset image size

The increase of river flow in winter is mainly due to an increase in precipitation since the contribution from the snow is less because of low temperature (<0 °C) in most parts of UIB. The increasing river flow rate in summer is mainly associated with warming summer temperatures, which causes higher ablation of snow and, in turn, a higher contribution to river flow. The observed river flows from 1990–2000 were also analyzed to find the correlation of river flow with observed minimum, maximum temperatures, and precipitation. Figure 7 shows that maximum temperature and minimum temperature are the main influencing factors and have strong correlation with river flow while precipitation has a weak correlation with river flow, and it even has a negative correlation (−0.38) over the Skardu station. The reason for the negative correlation is that the temperature is mostly below freezing point over this station, and snow and glaciers become deeply frozen and require ripening before melt can occur.

Zoom In Zoom Out Reset image size

The patterns of temporal changes in the river flow for the future periods are similar but different in magnitude. Chaturvedi et al (2014) has stated that an increased glacial retreat is expected in many sub-regions of the Karakoram-Himalayan region. The glaciers of this region are projected to lose a mass of larger than 1200 kgm⁻²a⁻¹. About 13% glacial area in the Western Himalaya region, 10% in Central Himalaya region, and 34% in the Eastern Himalaya region will have melted by 2030, as projected by RCP8.5, or by 2080, as projected by RCP26. Our findings for river flow are in accord with the results of recent study of Immerzeel et al (2013) that conclude consistent increase (300%) of river flow with RCP8.5 from one selected glacier in UIB during the twenty first century. Our results for RCP8.5 show that the total increase of river flow is about 87%, which is less than the findings of Immerzeel et al (2013) and this may be due to their consideration of only one glacier to estimate future river flows. In our study we represent the whole UIB area, where the area covered by glaciers is not more than 12% of the entire basin and the whole basin area would not contribute to the river flow, therefore our results seem reasonable.

The results of the present study are in accordance with the IPCC report AR4 (IPCC 2007) and future projections on the feedback of global climate change over glacierized basins (Rees and Collins 2006), which says that in first half of twenty first century a short-term increase in river flow is expected while in the last half of twenty first century there may be a sharp decrease in river flow. The results of both models indicate a rising trend with a higher increase in the rate of the river flow in the first half of the century. Meanwhile, the increasing rate is comparatively low in the last half of the twenty first century in RCP4.5 for 2071–2100, which may possibly be due to: (1) a persistent decrease in overall basin snow cover area, and (2) a slight lowering in warming and precipitation trends in RCP4.5 during that period, which might imply relatively less snow ablation and which in turn reduces the river flow. However, in the second case, the lowering in temperature trends might also affect the rate of evaporation losses, at least from the high altitude snow-fed catchments of UIB (e.g. Astore), and this reduction in losses would result in a slightly increased discharge. Nevertheless, the hydrological implications of the climatic variability in UIB would have a significant effect on the distribution of seasonal stream-flows because the months from June-August would be highly suspected to flash floods, which will directly affect the hydrological dynamics and water resources of the region. Due to increased river flow and higher accumulation rate of precipitation, future water availability is projected to be favorable in UIB, at least for the twenty first century.

There are several limitations of this study that need to be addressed. Firstly, it focuses on small spatial scales of the UIB basin, where the river flow is contributed mostly from snow and glacier melt. A more robust modeling approach is required to handle the glacier and snow melting and accumulation. Secondly, this paper is associated with bias correction and hydrologic model calibration. The bias correction and calibration approaches implicitly assume that it will retain these in future scenarios and this has added an additional uncertainty to the future projection of climate models, despite the fact that the bias correction was performed with recent observed data. Lastly, the uncertainties from numerical models used in this study that were studied through only two scenarios (RCP4.5 and RCP8.5), two climate models (CCAM and RegCM), and one hydrological model. It needs to be improved with finer resolution and multi-model (climate and hydrological) ensemble results to explore the uncertainties in more detail.

Uncertainties in the climatic and hydrological projection are one of the main issues that need to be addressed for better understanding of the results. Box-whiskers plots (figures 8–10) represent maximum temperature, precipitation and river flow for the observed (1976–2005), base line period (1976–2005), and future (2006–2035, 2041–2070, 2071–2100) time slices for both the emission scenarios (RCP4.5 and RCP8.5). The middle lines in the box-whiskers show the median values and the lower and upper boundaries show percentiles. The range of possible lower and upper values is denoted by the dashed line, and the circles show the outside values from the range. It is clear that ranges of RCP8.5 are higher than that of RCP4.5 and has more variation in future periods than the baseline. Uncertainty of RCP4.5 is lesser than RCP8.5 for temperature and river flow. PDFs (figures 11–13) represent the probabilistic distribution of maximum temperature, precipitation and river flow. All of the PDFs of maximum temperature exhibit a clear bimodal distribution, which is indicative of the fact that the maximum temperature reaches its maxima at two distinct regions. For the observed time series, the range is between 10–13 and 30–33, as shown in figure 11. The distribution of simulated temperature in both scenarios and for all-time slices has shifted more towards negative (left) and positive (right) phase as compared to the observed temperature and baseline period, respectively. For example, the simulated baseline temperature has shifted to the left as compared with the observation and the temperature ranges from −10 to 30 °C, approximately. Similarly, the probability density of the mode values is higher than the observed temperature. The shift in distribution is more obvious in RCP8.5, clear differences in the mode values in different time slices of RCP8.5 also differentiate it from RCP4.5. Although uncertainties exist in the scenarios, their use makes it feasible to study the hydrological changes with climate projections between the lower and upper limit of the future climate. Figure 12 represents the PDFs of precipitation for observed, baseline period and future time slices for both emission scenarios. It reveals that most of the observed precipitation ranges between 0–2 mm day⁻¹ and is consistent for all data sets. Unlike maximum temperature, the distribution of the observed precipitation is unimodal. There is some difference between observed and simulated precipitation, the density of the mode is almost 100% for observed precipitation while it ranges between 40% and 60% for simulated data sets. Figure 13 represents the distribution of inflows for the baseline period and the future time slices under two emission scenarios. The PDFs show that the uncertainties increases with time. The mean river flow for the base line period is 2769 cumecs (table 7) and the probability is higher from 0–1000 cumecs, while for the future periods it decreases and the range also shifts a little to the left. For RCP4.5, the different time periods have different distribution with different mode values and ranges. For the period 2071–2100, the distribution shows that the probabilities of higher values decrease as compared to 2041–2070. For RCP8.5, the probability of higher values for the time slices of 2071–2100 increases when compared to the other time slices in the range of 10 000–15 000 cumecs but decreases in the range of 5000–10 000 cumecs. Table 7 represents the average values of meteorological variables and river flow for the two scenarios, with their corresponding 95% margin of errors. In the case of river flow, it shows that there is comparatively more uncertainty in RCP8.5. For meteorological variables, table 7 shows that RCP8.5 probability is more uncertain than RCP45 because of higher error. But RCP8.5 is regarded as the upper threshold of future climate change, so it will limit the variation magnitude of temperature and precipitation for the river flow simulation.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size