4.1 Drought Characteristics
The standardized precipitation index (SPI) effectively reflected the changes in drought characteristics in the KRB from 1981–2021. Figure 2 shows the temporal characteristics of drought within KRB during the 1981–2021 period. It is evident from this figure that drought frequently occurred during the study period, as indicated by negative median values of the SPI (e.g., years 1985, 1992, 1994, 1997, 2004–2009, 2012, 2014, 2016, 2018, and 2019). Figure 2 also suggests an increase in severe and extreme drought frequency and severity in recent decades compared to the pre-2000 period.
The highly anomalous SPI values for 1985, 1992, 2005, 2006, 2012, 2016, and 2018 indicate a prolonged dry period or widespread drought across the basin. The dry episode was especially intense in 2016, as seen from the boxplot (Fig. 2). The presence of both extremely dry and wet conditions in 1984, 2009, 2010, and 2013 indicate drought conditions in parts of the basin or during certain periods of a year. In 2010 and 2011, the boxplot indicates extreme drought and extreme wet conditions, but it is a relatively wet year when the entire area is considered as indicated by the positive SPI median.
The medians of SPI for the years 1982, 1984, 1986, 1990, 1991, 1998, 2000, 2010, 2013, and 2020 are positive, indicating wet conditions in the region. The most widespread wet periods occurred in 1982 and 2000. Since 2000, 10 years were dry, and two years were wet.
Figure 2B depicts 3-month SPI (SPI-3) values, which indicate short-period drought and show different patterns than the SPI-12. The comparison of Figs. 2A and 2B suggest that the concurrence of severe and extreme droughts is becoming more frequent and strengthened for a long duration. There are significant differences, too, in the sensitivity of SPI values at these two-time scales.
The heat map in Fig. 3 shows the seasonal drought characteristics in the KRB. Results suggest that drought in the past occurred in all seasons. Winter drought (indicated by 3-month SPI in February) occurred periodically in many of the stations from 1982–2021, but the basin-wide drought was prominent in the years 1985, 1997, 1999, 2001, 2004, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2016, 2017, 2018 and 2021. It is evident from Fig. 3 that basin-wide winter droughts became more frequent and intense after 2000 (e.g., 13 times in 21 years) ) suggesting an increase in the frequency and intensity of winter droughts. Pre-monsoon drought (indicated by 3-month SPI in May) occurred on the basin-wide scale in 1984, 1985, 1989, 1992, 1994 to 1996, 1999, 2003 to 2005, 2008 to 2014, and 2019. For many years though, the drought has occurred on a basin-wide scale, but the intensity largely varies between stations. The result shows that monsoon drought, i.e., 4-month SPI in September, is experienced in KRB periodically in many stations but is experienced in the majority of stations in 1987, 1992, 2001, 2002, 2004, 2005, and 2015. Monsoon drought occurs in fewer periods on a basin scale than other seasonal droughts but shows large variation between stations in a single year. In case of post-monsoon (indicated by 2-month SPI in November), drought occurs in 1988, 1990, 1991, 1993, 1994, 1995, 2000, 2001, 2003, 2007, 2008, 2010, 2011, 2012, 2017, 2018 and 2020 on a regional scale. Post-monsoon drought occurred more frequently but with less intensity than other seasonal droughts in the region.
The annual frequency of extreme, severe, and moderate drought occurrences captured by 12-month SPI in KRB over the period 1982–2021 are presented in Fig. 4. The result shows that the drought frequency increased significantly after 2000, reaching its highest in 2016. An increase in the frequency of drought events after 2000 is readily discernible, with relatively extreme dry episodes occurring in 2005, 2006, 2009, 2010, 2011, and 2016. While extreme drought had a general periodicity (e.g., 1985, 1992, 2005,2006, 2016), results in Fig. 4 indicate a consistent general increase in the frequency of all drought types in the recent decade. Further, severe droughts were found to be more frequent in 1985, 1992, 2005, 2012, 2015, and 2016 with the highest in 2005.
4.2 Drought implication on water resources
4.2.1 River discharge
Effect of drought on surface water availability was analyzed by examining the river discharge variations in three major tributaries of KRB (Bheri in the east, Karnali in the middle, and Seti in the west) and the main outlet of the entire Karnali River system (i.e., the Chisapani station; refer to Fig. 1 for station location). Flow duration curves are used to present any differences in discharge having variable exceedance probability (Fig. 5). It is evident from Fig. 5 that droughts significantly impacted river discharge in KRB. For example, at the Asaraghat station (Fig. 5a), river discharge was substantially lower than the long-term average in most drought years (e.g., 1997, 2004, 2006, and 2009) and throughout the years; river discharge exceeded the long-term mean only during the high flow season. Similar results can be observed at the other three stations (Fig. 5b-c), with river flows impacted the most in 2004 and 2009, characterized by extreme and widespread basin-wide droughts. Seti river discharges in 1992, 1997, 2004, 2005, and 2008 show the lower values that coincide with the drought year of the region, and the lowest discharge in 2004 may be due to severe widespread drought in 2004. In Bheri, river discharges in 1993, 2002, 2004, 2005, 2007, and 2012 had lower values as the major drought events occurred in the region. In Chisapani, an outlet of the entire KRB, discharges in 1992, 1997, 2004, 2006, and 2009 reduced to lower values, indicating that droughts profoundly influenced river discharge in KRB.
4.2.2 Groundwater
Next, we analyzed the impacts of drought on groundwater storage. Figure 6 presents the spatial patterns of the linear trend in GRACE-based TWS anomaly in the study domain and its surrounding regions during the GRACE data availability period of 2002–2016. Results indicate a decline in TWS during the analysis period within and outside of the KRB; note that results in Fig. 6 are presented for an extended domain because of the relatively large footprint of GRACE data (Pokhrel et al., 2015). We further examined the changes in groundwater storage and linkages to drought and analyzed the temporal variations and trend in groundwater storage (Fig. 7) averaged over a region encompassing the KRB (white box in Fig. 6). Based on our results, groundwater declined substantially (-1.21 cm/year) and persistently during 2002–2016. Since the other TWS components ( such as river water, soil water, and snow water; see Methods and Data section) remained relatively stable or even had increasing trends (Fig. 8), the majority of the decline in TWS was contributed by the decline in groundwater alone. The inter-comparison study with drought year in KRB and groundwater storage anomaly could be attributed to the drought phenomenon associated with groundwater depletion in western Nepal. We should also note that since there is no significant depletion of groundwater resources due to pumping in KRB–as opposed to that in neighboring northern Indian regions (Fig. 6), the persistent decline in groundwater could be attributed to increasing/intensifying droughts.
Analysis reveals the more significant groundwater deficits due to drought in 2005, 2006, 2007, 2009, 2012, 2013, 2014, and 2016. During the widespread drought periods of 2004–2007, 2012, and 2015–2016 there was a significant reduction in groundwater storage and very low recharge during the monsoon season. During the extreme western winter drought of 2008–2009, the groundwater level in the winter months was historically low. Since the most widespread drought in the region occurred in 2016, the groundwater plummeted in 2016.
4.3 Drought implication on agriculture production
Decreases in climatic yield of wheat, a major winter crop in KRB, were observed when winter drought episodes occurred (Fig. 8). Results suggest that, during the winter drought years (i.e., 1985, 1999, 2004, 2007, 2009, 2010, 2016, 2017, 2018, and 2021), the climatic yield of wheat reduced compared to preceding years. Extreme yield loss occurred in 2008, 2009, 2010, and 2021, the years with the most widespread winter drought in the region. Another abrupt change in the climate yield of wheat can be visualized in 2016, 2017, and 2018 which were years with another predominant drought.
The correlation coefficients between the climatic yield of wheat and SPI time series at 1–5 months for November to March (growing seasons of wheat in the KRB) for selected stations are shown in Fig. 9. Here, a positive correlation indicates the implication of drought conditions on the yield of wheat. High correlations were observed during the wheat-growing period (December-March) with 1 to 5-month lags. This indicates that the whole cropping season is at risk of drought, and this drought sensitivity to wheat production may cause considerable yield loss.
4.4 Drought implication on vegetation
Figure 10 shows the annual average VCI over vegetated areas across the KRB and the annual median value from 2003 to 2021. VCI values were categorized into three groups to define the severity of drought; i.e., below 36 severe drought; 36-50 moderate droughts, and above 50 no drought. Figure 10A shows that in 2004, 2005, 2006, 2008, 2009, 2010, 2011, 2012, 2013, 2016, and 2017 there were large vegetation deficit areas. All the years align with a meteorological drought. The most widespread and severe vegetation deficit areas (Fig. 10B) showing the implication of drought in vegetation were in 2004, 2008, 2010, 2011, and 2012.