Data

The long-term pattern of precipitation in this region is not well known due to a lack of long-term observations even at spatial scales of a few tens of kilometers (Grabs and Pokhrel 1992; Fujita et al 1997; Shrestha et al 1999). Small-scale studies are necessary because the complex terrain can introduce significant differences in climate, even across short distances. Most existing rain-gauge networks, especially in complex mountainous areas, are not dense enough to fulfill this requirement. However, the recently introduced Asian Precipitation – Highly Resolved Observational Data Integration Towards the Evaluation of Water Resources (APHRODITE) project (Yatagai et al 2009, 2012) makes available a lengthy daily gridded high-resolution data set that was prepared by collecting rain-gauge observational data from thousands of stations across Asia. This data set (APHRO_V1003R1) is a long-term (1951–2007) continental-scale product that contains a dense network of daily rain-gauge data for Asia, including the Himalaya and mountainous areas in the Middle East. APHRO_V1003R1 data sets for Monsoon Asia, Russia, and the Middle East (on 0.5° × 0.5° and 0.25° × 0.25° latitude/longitude grids) are available at  http://www.chikyu.ac.jp/precip/. For this analysis we used the 0.25° × 0.25° data. We validated these data for the Indian region by comparing the seasonal monsoon rainfall with data sets from the India Meteorological Department, Global Precipitation Analysis, and the Climate Prediction Center's Merged Analysis of Precipitation (Patwardhan 2012). APHRODITE is also being increasingly used for operational climate applications such as extreme drought and flood forecasting in East Asia (Sohn et al 2011). This data set bridges the gap in long-period observed station data for the HKH region.

Stations in the WH are quite sparse compared with the CH and EH (Yatagai et al 2012); hence, if the spatial interpolation applied by the APHRODITE algorithm does not compensate by reproducing orographic precipitation enhancement then precipitation estimates in this subregion will be differentially biased compared with the other 2 regions.

This study used the following data:

PRECIS is an atmospheric and land surface model of limited area and high resolution that is locatable over any part of the globe. Dynamic flow, the atmospheric sulfur cycle, clouds and precipitation, radiative processes, the land surface, and the deep soil are all described, and boundary conditions are required at the limits of the model's domain. The basic aspects explicitly handled by the model are described in Jones et al (2004). The lateral boundary conditions required to drive the regional climate model are taken from the perturbed physics ensembles simulations of HadCM3, the global climate model of the Met Office Hadley Centre in the United Kingdom.

Perturbed physics ensembles

Simulations from a 17-member perturbed physics ensemble produced by using HadCM3 under the QUMP project of the Hadley Centre have been used as lateral boundary conditions for 138-year simulations of the PRECIS regional climate model. The perturbed physics approach was developed in response to the call for better quantification of uncertainties in climate projections (see chapter 14 of the Third Assessment report of IPCC 2001). A significant amount of perturbed physics experimentation has been done with HadCM3 and variants, starting with the work of Murphy et al (2004) and Stainforth et al (2005).

The QUMP simulations comprise 17 versions of the fully coupled version of HadCM3, 1 with the standard parameter setting and 16 in which 29 of the atmosphere component parameters are simultaneously perturbed (Collins et al 2006). The lateral boundary conditions for 3 QUMP simulations (Q0, Q1, and Q14) for the Intergovernmental Panel on Climate Change's A1B emissions scenario (IPCC 2001) were made available by the Hadley Centre. These 3 QUMP runs were carried out at Indian Institute of Tropical Meteorology in Pune, India, for the period 1961–2098 for the area that stretches from longitude 56.77°–103°E and latitude 1.5°–38°N, and were used to generate an ensemble of future climate change scenarios for the HKH region. These simulations were made at 50 × 50 km horizontal resolution. The daily precipitation and surface air temperature from the QUMP simulations were analyzed to develop high-resolution regional climate change scenarios to study the impact of global warming on the HKH region climate. The impacts were studied for 3 time periods: 2011–2040, 2041–2070, and 2071–2098. The likely changes in precipitation and temperature were tested for their significance by applying the standard Student's t-test. The statistical significance of the trend in area-averaged time series was tested by using the nonparametric Mann-Kendall test (Mann 1945).

Current climate variability

Observed rainfall and temperature variability:

The HKH region's division into 3 subregions (WH, 62–75.5°E and 32.5–38.5°N; CH, 75.5–84.5°E and 25.5–34°N; EH, 84.5–97°E and 22–30°N; see Figure 1) is primarily based on climate and topography. The observed mean rainfall for June through September for 1961–1990 was 86, 546, and 1042 mm, and the coefficient of variations (defined as [standard deviation / mean] × 100) is 18, 16, and 9% for the WH, CH, and EH, respectively; and mean annual temperatures were 9.9, 8.9, and 13.6°C, respectively. To examine the interannual variability of seasonal (June–September) rainfall, a long time series (1951–2007) of seasonal rainfall was prepared for these 3 regions by using APHRODITE data. The total June–September rainfall over grids that lie in these 3 regions was averaged for every year. The mean rainfall was subtracted from each value and divided by the standard deviation to compute standardized time series for 1951–2007. The standardized rainfall series over the WH, CH, and EH are depicted in Figure 2A. All 3 time series are highly random and do not show any significant trend. However, it is well observed that the seasonal rainfall in the CH and EH declined substantially after 2000.

FIGURE 1

The Hindu Kush–Himalayan region considered in this analysis. The region was divided into 3 subregions: WH, 62–75.5°E, 32.5–38.5°N; CH, 75.5–84.5°E, 25.5–34°N; and EH, 84.5–97°E, 22–30°N. (Map by authors)

FIGURE 2

(A) Interannual variability of standardized June through September (JJAS) rainfall and (B) anomalies of annual average temperature (°C), 1951–2007, in the WH, CH, and EH.

The anomaly time series of annual average surface air temperature over the 3 regions are shown in Figure 2B. These time series were prepared by using monthly surface air temperature data from the NCEP/NCAR reanalysis data set for 1948–2010. Temperatures were quite cool from 1970–1990 in all 3 regions; after the 1990s, they increased, although not to a statistically significant degree. Warming has been more rapid in the CH than in the EH and WH.

The spatial patterns of observed APHRODITE seasonal rainfall data and NCEP/NCAR reanalysis annual average temperature data are shown in Figure 3A and B, respectively. These composites were prepared based on data for 1961–1990 for validation of model simulations. In the WH, rainfall ranged from 50–500 mm. The levels were lowest in the northwestern parts of the WH and increased from northwest to southeast. In the CH, rainfall ranged between 500 and 1000 mm, and in some parts exceeded 1000 mm, whereas, in the EH, it was more than 1000 mm. In northern parts of Bangladesh and far northeastern parts of India, rainfall exceeded 3000 mm. Annual average temperatures (Figure 3B) decreased from south to north over the HKH region. The QUMP simulations were validated with these patterns, as shown below, to examine their ability to reproduce the current climate.

FIGURE 3

(A) Seasonal (June–September) rainfall (mm) based on APHRODITE data; (B) annual average temperature (°C) based on NCEP/NCAR reanalysis data for 1961–1990.

Model-simulated rainfall and temperature variability:

The high-resolution regional simulations generated by using PRECIS, with lateral boundary conditions from QUMP simulations, were studied in detail to evaluate the model's effectiveness in representing the regional climatological features in terms of both annual cycles and spatial patterns. This was the first time that QUMP simulations were used for the HKH region.

Regarding annual cycles, the monthly mean temperature and rainfall simulated by the 3 scenarios based for 1961–1990 for the 3 regions as well as the observed rainfall (from APHRODITE) and temperature (from the NCEP reanalysis) are depicted in Figure 4. In all 3 regions, all 3 QUMP runs simulated the shape of the annual cycle of surface air temperature reasonably well; however, they exhibited a cold bias in winter. The temperatures for the monsoon season were well simulated. The mean temperatures were not corrected for the lapse rate effect of the differences in mean elevations in the various data sets used in the analysis.

FIGURE 4

Annual cycles of surface air temperature and rainfall in the WH, CH, and EH simulated by 3 QUMP experiments, Q0 (red), Q1 (green), and Q14 (blue), their multimodel ensemble (MME; dashed magenta line), and the APHRODITE rainfall values and NCEP reanalysis temperature values (black, observed) for 1961–1990.

The shape of the annual rainfall cycle was well captured by all the simulations for the WH and CH; however, for the EH, the maximum rainfall was simulated 2 months ahead, in May and June. For all 3 regions, a wet bias was exhibited. For the WH, a large wet bias of 20 mm was exhibited for the dry monsoon season (November–March). During the monsoon season (June–September), the rainfall was simulated reasonably well. For the CH, Q1 tallied very well during the monsoon but exhibited a wet bias of 10–20 mm/mo in the remaining months. For the EH, a wet bias of 30–50 mm/mo was exhibited for the premonsoon months. Hence, there was a large uncertainty in the 3 simulations. However, in general, the model did reproduce the shape of the annual cycle of rainfall and surface air temperature reasonably well. The mode of precipitation seasonality in the WH is during the winter months, and the primary hydrological impacts there from precipitation change will most likely be due to changes in winter rather than summer and monsoon precipitation.

The spatial patterns of seasonal rainfall as simulated by the 3 QUMP runs for the baseline period 1961–1990 were compared with the observed APHRODITE precipitation patterns, as shown in Figure 5A. PRECIS seems to provide an adequate representation of present-day conditions. The seasonal (June–September) rainfall pattern was well simulated for the HKH region, as seen in Figure 5A. All the 3 QUMP runs exhibited a wet bias in the northern parts of the CH. Q0 and Q14 captured the seasonal rainfall amount and patterns quite reasonably, but Q1 exhibited a dry bias for the WH and some parts of the CH, which has also been seen for the Indian landmass (Krishna Kumar et al 2011).

FIGURE 5

(A) Composite seasonal (June–September) monsoon rainfall (mm/d) simulated by 3 QUMP runs compared with the observed APHRODITE values for 1961–1990; (B) composite annual average surface air temperature (°C) simulated by 3 QUMP runs compared with the observed NCEP reanalysis values for 1961–1990.

The composite spatial patterns of annual average temperatures for 1961–1990 are given in Figure 5B. The cooler temperatures in the northern parts and warmer temperatures in southern parts of the HKH region were well simulated by all 3 experiments. All three exhibited a warm bias of 1–2°C in the EH. Cooler temperatures were simulated in the far northern tip of the WH.

Projected changes in rainfall and temperature

Projected percentage changes in seasonal rainfall in 3 QUMP simulations from the baseline (1961–1990) to the near future (2011–2040), the middle of the century (2041–2070), and the end of the century (2071–2098) are depicted in Figure 6A. Key projections were that the monsoon rainfall may decrease over the CH during 2011–2040, whereas there may be a 5–10% increase in rainfall in the WH and EH. In 2041–2070, rainfall may accelerate, but the northern parts of the CH may be deprived of rainfall. In 2071–2098, all 3 experiments project a 20–40% increase in seasonal rainfall for the entire HKH region. Because precipitation seasonality in the WH is during the winter months, the primary hydrological impacts there from precipitation change will most likely be due to changes in winter rather than summer and monsoon precipitation. There was a large discrepancy in the 3 projections of summer monsoon rainfall in the WH.

FIGURE 6

(A) Percentage change (compared with a 1961–1990 baseline) in monsoon precipitation; (B) change in average annual surface air temperature in 3 QUMP experiments in 3 time periods.

The projected changes in the surface air temperature for the same time frames and the QUMP runs are shown in Figure 6B. As time passes, the temperature is projected to rise by 0.5°C to more than 5°C. The maximum warming is projected for the WH. Q14 projects maximum warming for the HKH region. For 2011–2040, the warming is projected to be 0.5–1°C; for 2041–2070, 1–3°C; and for 2071–2098, 4–5°C, which would be quite threatening to the snow and glaciers throughout the region.

The possible percentage changes in standard deviation of summer monsoon rainfall under the 3 QUMP simulations for the 3 time periods are shown in Figure 7A. Monsoon rainfall was projected by all 3 QUMP simulations to be much more variable (a 20–40% rise in standard deviation) in the CH and EH toward the end of the century than it was in 1961–1990; a smaller change in variability (not more than 10%) was projected for the WH.

FIGURE 7

(A) Percentage changes in standard deviation of monsoon precipitation in 3 QUMP experiments for 3 time periods with respect to 1961–1990; (B) ensemble average of percentage change in seasonal rainfall; (C) ensemble average of change in annual average surface air temperature (°C) for 3 time periods with respect to 1961–1990.

There is a large uncertainty in simulation by the 3 QUMP runs, and the best solution is to consider an ensemble average of the 3 runs (ie the average of projected changes given by the 3 QUMP simulations at every grid point). The ensemble projected changes in seasonal rainfall and annual average temperatures for the 3 time periods are also shown in Figure 7B and C, respectively. For 2011–2040, rainfall was projected to decrease over the CH and EH; during 2041–2070 there may be a 5–20% change, whereas, at the end of the century, there may be an increase of 20–40% in the CH and EH over 1961–1990 rainfall levels. Similarly, the annual average temperature is projected to rise in 2011–2040 by 1–2°C, in 2041–2070 by 1–3°C, and in 2071–2098 by 3–5°C for the HKH region. The rise in mean annual temperature may be greater in WH than in CH and EH, as seen in Figure 7. This may be a serious threat to the glaciers and snow cover in the HKH region. The ensemble mean values of the 3 QUMP runs are given in Table 1. All runs exhibited a wet bias for all 3 regions, but the mean annual temperatures were reasonably well simulated by the ensemble mean.

TABLE 1

Ensemble mean summer monsoon rainfall (mm) and mean annual surface air temperature (°C) for 1961–1990, 2011–2040, 2041–2070, and 2071–2098 compared with observed values in the WH, CH, and EH. (Table extended on next page.)

TABLE 1

Extended. (First page of Table 1 on previous page.)