Reconstruction of July-September standardized precipitation-evapotranspiration index (SPEI) for the Hindu Kush region of Northern Pakistan

The prolonged drought resulting from global warming is considered an important factor affecting West Asia’s socioeconomic growth, with a significant impact on the dynamic forecasting of water supply and forest ecosystems. In such a scenario, understanding historical long-term drought changes is crucial for accurately forecasting regional drought shifts in the Hindukush region. In this study, a 517-year (1506–2022 C.E.) long tree-ring width chronology of the Himalayan Cedar (Cedrus deodara D. Don) from the eastern Hindukush has been developed. The July-September Standardized Precipitation-Evapotranspiration Index (SPEI) has revealed a positive and significant relationship (r = 0.633, p < 0.001) with tree growth, which leads to SPEI reconstruction from AD 1626 in the Hindu Kush region. Our reconstruction model has explained 40.1% of the climate variance during the instrumental period from C.E. 1965 to 2018. Fourteen wet periods (≥ 3 years) were observed before the instrumental period, specifically in C.E. 1629–1635, 1638–1658, 1666–1674, 1680–1701, 1715–1724, 1770–1776, 1794–1797, 1802–1810, 1822–1846, 1850–1857, 1872–1881, 1883–1890, 1906–1914, and 1921–1937. Similarly, twelve dry summer periods were also observed in the past 339 years, such as C.E. 1659–1665, 1675–1679, 1702–1714, 1725–1769, 1777–1793, 1798–1801, 1811–1821, 1847–1849, 1858–1871, 1891–1905, 1915–1920, and 1938–1963. Nevertheless, AD 1663 was individually the wettest (with a value of 2.13), while AD 1754 was the driest (−0.99) year. The spatial correlation analysis and its comparisons with Karakoram-Himalayan drought and precipitation reconstructions have convincingly confirmed the reliability of our SPEI reconstruction. Consequently, this reconstruction can effectively serve as a proxy for large-scale drought variability in the Hindu Kush region of northern Pakistan. Our findings strongly suggest the considerable dendrochronological potential for further climatological studies in the western Hindu Kush Mountains System.


Introduction
In recent years, there has been a worrying increase in the frequency and intensity of droughts, especially in regions that are already dry or semi-arid.This concerning trend is attributed to the impact of climate change on a global scale [1].Changes in temperature and precipitation patterns are just two examples of how the ecosystem is significantly impacted by climate change on Earth.Consequently, previously dry regions are now more vulnerable to droughts in the water supply and system [2].Climate models predict that the majority of continents will experience increased aridity in the twenty-first century, leading to regions expanding by as much as 23% by the end of the century [3,4].This phenomenon is expected to have a considerable negative effect on forests, including reduced production and a rise in the frequency and intensity of fires [5,6].In this sense, it is crucial to improve our understanding of natural climatic variability across time, which is a difficult task due to a lack of historical data.Our knowledge of historical climatic conditions is limited [7], due to weather stations in the region (Hindu Kush) with short ( 50 years) and missing long term data, which significantly restricts our understanding of climatic variability, such as droughts at spatial and temporal scales.
Tree rings have been proven to be a reliable proxy for reconstructing historical climatic fluctuations over the last few decades.Particularly in the Karakoram-Himalayas region of northern Pakistan, the tree rings of various coniferous species have been used recently to reconstruct various climatic factors, such as precipitation [8][9][10], temperature [11][12][13], soil moisture [14], streamflow [15][16][17][18], and droughts [19].In the Himalayas, there is a high frequency and intensity of droughts [20], whereas the primary source of precipitation in this area is the Asian summer monsoon (ASM) during summer, and mid-latitude north-westerly disturbances during winter and spring [21,22].Droughts and floods in the region are mainly caused by the failure or excessive precipitation from one or both of these sources [23].However, assessing and categorizing droughts is difficult due to their degree of variation, duration, and geographic extent [24].Nonetheless, the latest advancements have led to the establishment of new approaches to assess and monitor these traits [25].A valuable tool for comprehending this phenomenon is the employment of these approaches, which also contain drought indices [6].There are also several drought indices accessible [26].There are several indices commonly used for drought analysis, such as the Standardized Precipitation Index (SPI) and Palmer Drought Severity Index (PDSI), we have opted for the SPEI due to its comprehensive approach [24].The PDSI is a drought indicator that covers various temporal scales and integrates meteorological data, whereas the SPEI largely depends on a set time scale and the water balance equation.Both indices take weather into account, but SPEI has an edge in northern Pakistan as it may shed light on the drought conditions across a range of time periods.Droughts are inherently complex phenomena that occur at multiple scales [23].In order to improve our understanding of climate variability, our efforts to reconstruct drought conditions focus on the July-September SPEI, taking into account both temperature and precipitation.This approach allows for a more nuanced and globally applicable perspective on drought conditions [27][28][29][30].
Previous reconstructions in the Hindu Kush region focused on March-August PDSI [31], but our current methodology aims to provide a more comprehensive analysis by considering a wider range of influencing factors.The decision was based on the SPEI's effectiveness as a versatile drought indicator, using meteorological data to assess global drought situations at multiple scales e.g., [29].It has been successful in certain regions to combine tree rings with the SPEI to examine the community dynamics of plants impacted by drought [2,[32][33][34].Nevertheless, this technique is new in Pakistan and has not been widely applied, particularly in the Hindu Kush region.In this research, our aims are: (1) to determine the critical climatic variables influencing C. deodar trees' radial growth, (2) to assess the influence of climate, particularly drought, on tree growth over various periods, (3) to utilize tree-ring width and SPEI data to reconstruct seasonal drought patterns over the past 400 years in the Hindu Kush region, and (4) to examine drought variations at decadal to multi-decadal scales in the Hindu Kush region of northwest Pakistan.

Study area and climate
Our sampling site is situated in the Rumbur Valley, which is part of the Hindu Kush mountain range in the northwest region of Pakistan (figure 1).The broadleaved and coniferous forests of Rumbur are renowned for their rich flora and diverse wildlife, making it an exceptional and ecologically visible area [35].This region has a diverse environment, topography, and flora, with dry temperate coniferous forests on the upper slopes of the valleys characterized by P. wallichiana, C. deodara, A. pindrow, and J. excelsa, whereas various oak species at lower elevations, including the critical woody plant Q. incana, are abundant [36].This monsoon shadow region receives little precipitation and moderate temperature and therefore falls under the category of a dry Mediterranean climate.The area has distinct seasons, with warm summers and cold winters.According to Haserodt [37], the snowline in the region is located at an elevation ranging from 4800 to 5000 m asl.While the western and central areas have hot, dry summers and cold, rainy or snowy winters as a result of western disturbances, the southern Hindu Kush receives very little precipitation due to the monsoon and/or thunderstorms of the western frontal system [38].

SPEI data gathering and analysis
The SPEI data was collected form the worldwide gridded SPEI CRU TS 4.06 dataset ((http://climexp.knmi.nl),which has a 0.5°spatial resolution [39].The SPEI measures drought severity by standardizing water balance values using long-term climatic norms.It combines precipitation and evapotranspiration data, with negative values indicating dry conditions and positive values indicating wet conditions.This index offers a comprehensive measure of hydroclimatic variability.To identify the optimal scale for correlation between SPEI and the acquired chronology, we utilized monthly SPEI values for time scales of 3, 6, 9, and 12 months (cumulative drought) from the closest gridded dataset (35°.75 ′N, 71°.75 ′E) spanning the years AD 1965 to 2018.According to the gridded monthly climate datasets, the region receives an average annual precipitation of 75.05 mm.The months of February and March, specifically late winter and early spring, revealed the highest levels of precipitation (figure 2).As for temperatures, the January records the lowest monthly average temperature (−7.8 °C), while July records the highest monthly maximum average of 15.82 °C.

Development of chronology and tree ring analysis
The Himalayan cedar, Cedrus deodara D. Don (hereafter C. deodara), is a common and ecologically significant evergreen coniferous tree in the mid-elevation zones and close to the timberline in the Hindu Kush mountain system in northern Pakistan [40].It has a broad ecological amplitude and tolerance [41].No human or natural disturbances were observed in the studied area.At least two increment cores were extracted from healthy trees at breast height during the sampling phase.The robust statistical interpretation was carried out using fifty-nine increment cores obtained from thirty-one trees and prepared through conventional dendrochronological methods [42,43].The core sample preparation involved air-drying, mounting on wooden mounts, polishing with progressively finer sandpapers, cross-dating, and their ring-width measurements with a precision of 0.01 mm using LINTAB-6 (Rinntech, Heidelberg, Germany).For evaluating cross-dated TRW measurements, the quality control program 'COFECHA' was used [44].Furthermore, any age-related and non-climatic disturbances trends were standardized by detrending the C. deodara chronology using the 'ARSTAN program' to establish a negative exponential trend [45].However, in some cases where this trend did not align with the raw data (which was the case for only 14 samples), we opted to use linear regression instead [46].By computing the average of the detrended series, we developed a standard mean chronology for the study area.
The statistical details of the standard chronology (hereafter STD), produced for all the trees, are presented in table 1.To identify the credible starting years of the chronologies, we utilized the subsample signal strength (SSS) threshold, since early data was insufficient.If the SSS is > 0.75 [47], the tree-ring chronology is typically considered for further analysis.Since the criterion of 0.75 is attained at a sample depth of 6 cores, hence we considered the most reliable period to be 1626-2022 CE.To evaluate the accuracy of the chronology, the interseries correlation coefficient (Rbar) and the SSS were calculated for a 50-year duration and then adjusted by 25 years [47].The Rbar, which varies between 0 and 1 (figure 3, lower panel), indicates the level of mutual variance employed to determine the extent of shared variation among the series [48].

SPEI reconstruction and statistical analysis
A Pearson correlation analysis was conducted to identify the climate signal inherent in the tree ring data.The analysis explored the relationship between the STD chronology and climate variables from 1965 to 2018, using gridded data centered at 35°75′N, 71°75′E, aligning with the availability of most station data available from 1965.Climate growth response analysis was conducted using monthly average temperature (T mean ), monthly average maximum temperature (T max ), monthly average minimum temperature (T min ), mean monthly precipitation, PDSI, and SPEI.
To investigate the possible impact of the previous year's climate on tree-ring growth and development, Pearson correlations were assessed using climate data from June of the previous year to September of the current year.These investigations aimed to provide valuable insights into the correlation between climate and tree growth in the study region, revealing significant associations between tree growth and weather conditions.To reconstruct past seasonal drought variability, multiple linear regression was used to transform the STD chronology into predictions of seasonal drought (SPEI).To ensure the development of a reconstruction model with minimal residuals, the most significant SPEI correlations were carefully selected.To account for the lagging effects of climate on tree growth, both current-year growth (at lag t) and the previous year's growth (1-year lag; t + 1) were employed as independent variables [42].
The prediction performance of the regression model was evaluated using independent calibration and validation statistics for two distinct periods: calibration from 1965-1991 C.E. and validation from 1992-2017 C.E. (table 2).Various statistics, such as Pearson coefficient of correlation (r), explained variance (R 2 ), adjusted explained variance (R 2 adj), Durbin-Watson test (DW), F-value, and reduction error were used in the split-period calibration/ verification investigations [42].

Comparison and assess the spatial representativeness of reconstructed SPEI
To check the accuracy of our SPEI reconstruction for the region, additional analysis was conducted to assess the spatial correlations between SPEI reconstruction and the first-order difference data with CRU TS 4.06 June-September SPEI field.The analysis was carried out using KNMI Climate Explorer (http://climexp.knmi.nl).Furthermore, our reconstructed series was compared with earlier drought and precipitation reconstructions specifically for the Hindu Kush-Karakoram region [9,31,43] to examine the regional representation of our series.

Climate and TRW analysis
Figure 4 illustrates the correlation between the STD chronology and monthly climate data from AD 1965 to 2018.We observed a significant positive correlation (p < 0.05) with SPEI variables from June of the prior year to September of the current year.Moreover, a significantly strong positive correlation (p < 0.001) with SPEI was evident from July to September of both the prior and current years.The PDSI from June-September of the previous year and from February-September of the current year exhibited a substantial relationship (p < 0.05) with TRW chronology (figure 4(a)).Precipitation was relatively weakly correlated with TRW compared to SPEI, except for September (r = 0.29) and November (r = 0.22) of the previous year and February (r = 0.25) to March (r = 0.24), and September (r = 0.33) of the current year.We did not observe a positive and significant relationship between temperatures and TRW chronology, except for September (r = −0.29) of the prior year, and February (r = −0.26),March (r = −0.24),and September (r = −0.32) of the current year, which were negatively and significantly associated (figure 4(b)).Additionally, a strong and significant association in the firstdifferenced data and SPEI from July to September (figure 6, right panel; b) was also recorded.Given that correlations with temperature were notably smaller and non-significant compared to those with drought (PDSI and SPEI), It is suggests that drought (SPEI) is the main limiting factor for tree growth and development in this region.

Calibration, verification, and SPEI reconstruction
The TRW chronology was selected for a robust reconstruction due to its strong and positive correlation with the July-September SPEI.To ensure the stability of the regression model, a split-period calibration and verification analysis method was used.resulted in positive RE values.These findings indicate that the verification period was significantly impacted by these results.Based on the significant statistical findings and analysis above, the SPEI for July to September was reconstructed through a multiple linear regression model with the TRW chronology serving as the predictor, as presented below: • JAS refers to the July to September • TRt refers to tree-ring width (TR) for the current year (t).
• TR t−1 designate to the tree-ring width (TR) for the previous year (t-1).
The agreement between the reconstructed SPEI changes and observed data from AD 1965 to 2018 is depicted in figure 5(a), with consistent variances throughout the period.Furthermore, figure 5(b) displays the July-September SPEI reconstruction in TRW covering the period AD 1626 to 2018.The wet trend shown in the early seventeenth, early nineteenth, late twentieth, and early twenty-first centuries is shown in figure 5(b) as the reconstruction's most noteworthy feature.After applying an 11-year low-pass filter to the reconstructed SPEI series, the average drought is 0.38.Over the last 393 years of SPEI records, we observed that 43 years were extremely wet and 41 years were extremely dry periods.This classification is based on the 1σ criterion assuming a deviation of 0.50 from the long-term mean of 0.38.Before the advent of instrumental records, between the years of AD 1626 and 1964, the wettest year was 1643, with an SPEI of 2.13, and the driest was 1754, with an SPEI of −0.99.The longest (3 years) wetter and drier periods over the previous 393 years occurred, respectively, in 1822-1846 CE (0.83 ± 0.38) and 1702-1714 (0.03 ± 0.19) (table 3).

Spatial representativeness
Further spatial correlation analyses show that the reconstructed and observed SPEI had a significant (p < 0.01) and positive correlation with the current SPEI pattern from CRU TS 4.06 and the first-order difference data in northern Pakistan, as shown in figures 6(a) and (b).The regions with the most notable correlation were located in the transitional zone between northeast and northwestern Hindukush, as well as the Himalayan region.This correlation was particularly strong and spanned a wide geographical area.However, the correlation analysis highlights the Hindu Kush, Karakorum, central-eastern Himalayan, and Pamir (Tianshan) ranges as having particularly strong correlations.This strong correlation indicates that the Rumbur reconstruction effectively captures drought changes in the region and is in agreement with the CRU TS 4.06 dataset.

Climate-growth relationship
The study conducted in the Hindukush region has shown a strong connection between tree growth, represented by STD chronology, and two important climatic factors: drought, measured by the SPEI and PDSI.The results indicate a significant correlation between STD chronology and drought, especially when measured by SPEI, suggesting that moisture stress has a significant impact on the growth pattern in the study area.However, the consistently weak correlation with precipitation alone emphasizes the inadequacy of using precipitation as a sole indicator of drought conditions, necessitating the incorporation of comprehensive indices like PDSI and SPEI for more reliable hydroclimatic analyses.The correlation was not significant and an inverse relationship between the growth response and temperatures for most months of the study period.This suggests that moisture stress, rather than temperature changes, is the main factor affecting the growth of C. deodara in the specific region.This  may be due to water scarcity and soil moisture depletion in the Hindukush region lead to prolonged drought conditions, which severely impact the sensitive ecological balance of the native flora e.g., Rossi et al [49], especially C. deodara trees.Limited adaptive strategies and drought-induced stress responses further weaken the trees, making them vulnerable to diseases, pests, and mortality.However, positive moisture input during July-September, as indicated by SPEI, can significantly impact tree growth as these months are critical for their development.The correlation between the standard chronology of C. deodara and drought during July-September, which extends from June of the previous year to September of the current year (figure 4), is consistent with previous studies on other conifer species in the Hindu Kush-Himalaya region of Pakistan [27,32,36].Studies on the Tibetan Plateau (TP) in China [50][51][52][53] and the central Himalaya [54,55] have also reported that the forest growth of conifers is sensitive to drought.

Comparison with other reconstructed records
To gain a better understanding of the regional patterns of drought variation during late summer and early autumn, the reconstructed SPEI for July to September in the Hindu Kush region was compared to other treering-based PDSI and precipitation reconstructions (figure 7).These comparisons provided important insights into the similarities and differences in SPEI changes between regions during this time period.The findings of these studies are significant in understanding the geographic similarity of drought variance.Our SPEI reconstruction is highly consistent with the reconstructed March-August PDSI in the source region of the Hindu Kush mountains [31] (figure 7 .Our reconstructed SPEI records are positively and significantly (p < 0.01) correlated with Hindu Kush (r = 0.72), Karakoram (r = 0.13 and r = 0.17) [8,9] at the inter-annual scale (figure 7).
After applying an 11-year low-pass filter to the series, correlations are also significant (p < 0.01) (figure 7).After the 11-year low-pass filter was applied, the prominent dry intervals in the C. E. 1659-1665, 1675-1679, 1702-1714, 1725-1769, 1777-1793, 1798-1801, 1810-1821, 1847-1849, 1858-1871, 1891-1905, 1915-1920, 1938-1963, and   1794-1797, 1802-1810, 1822-1846, 1850-1857, 1872-1881, 1883-1890, 1906-1914 and 1921-1937 are consistent with the PDSI of Hindu Kush [31] and precipitation from the Karakoram region [8,9] (figure 7).The results of our research are consistent across different areas of study, indicating a strong connection between changes in drought patterns in these regions.This reinforces the accuracy and reliability of our SPEI reconstruction.Some inconsistency between our reconstruction and the precipitation reconstruction by Treydte et al [8] may be due to temperature/precipitation effects on the SPEI, which prevent it from having a perfect connection with precipitation.These discrepancies, we observed between our SPEI reconstruction and the precipitation reconstruction from Karakoram region could be due to several factors.Possible reasons may include regional and microclimate variability of both regions (Hindukush and Karakoram regions) e.g., [12,56], character of species associated with an ecological and physiological appearance in deciding climate association with tree growth [57], differences in data sources, discrepancies in the reconstruction methods used [58], varying spatial or temporal resolutions, and inherent uncertainties associated with each reconstruction approach [52,59].Additionally, by taking into account the various ways that radial development responds to environmental perturbations, it is possible to better understand discrepancies that exist between reconstructed ring widths within particular periods [60].The interpretability of paleoclimatic reconstructions might be negatively impacted by these diverse reactions because they could produce discordant patterns that provide signals from the two proxies that differ from one another.Another possible explanation for this variation could result from several factors, including variations in species, geographic location, and reconstruction indices [51,61].
Our study reveals the existence of two significant mega-drought events in Asian history, namely the Strange Parallels Drought (1756-1768) and East India Drought (1790-1793), which have been accurately captured through our SPEI reconstruction [61,62].It's important to note that the wet period from 1992 to 2016 closely matches the results of Yadav et al [63] and Ahmad et al [31], demonstrating the accuracy of our SPEI reconstruction.The study suggests that dry and wet fluctuations were significant in certain regions of Hindu Kush region, northern Pakistan, which were effectively captured by the reconstructed SPEI.The spatial correlation analysis of the first-differenced data and reconstructed SPEI revealed a positive and significant correlation (p < 0.05) with the global SPEI data (July-September, figure 6) sets, indicating high representativeness for the Hindukush region and the northwestern part of Pakistan.Our analysis of the SPEI shows the effects of climate change on tree growth and hydroclimate patterns.We have linked SPEI with growth behavior to detect periods of water stress.This information helps unravel the intricate interactions between climate and trees.It also has practical applications in decision-making, land-use planning, and water resource strategies, assisting in adaptive measures to mitigate the impacts of changing climate on regional ecosystems e. g., [64,65].

Conclusions
A long-term tree-ring chronology of Cedrus deodara from the Hindu Kush region of northern Pakistan was developed, spanning 517 years (AD 1506 to 2022).Our study demonstrated a strong positive correlation between the growth of this conifer species and SPEI data from July to September (r = 0.633, p < 0.001), indicating that drought had a significant impact on tree growth in the Hindu Kush region.By utilizing this correlation, we were able to reconstruct July to September SPEI from AD 1626 to 2018 using a robust reconstruction model.In northern Pakistan, the Hindu Kush mountain range has experienced changes in its SPEI over several decades to centuries.Our reconstruction has uncovered noteworthy SPEI records that shed light on these alterations.However, the reconstructed July-September SPEI showed that 43 of the last 393 years had the wettest, while 41 had the driest conditions.Notably, this reconstruction accurately depicted obvious drought, and comparisons with other drought and precipitation reconstruction series further verified its dependability and validity.Nevertheless, it is crucial to create additional tree-ring chronologies in the Hindu Kush mountains of northern Pakistan to acquire a better understanding of the patterns in climate change brought on by global warming and to comprehend how tree growth responds to climate variation.

Figure 1 .
Figure 1.Sampling site in the Rumbur valley, Hindu Kush region of northwestern Pakistan.

Figure 2 .
Figure 2. Monthly climatic variability in CRU TS 4.06 data, including monthly average temperatures and monthly mean precipitation.

Figure 4 .
Figure 4. Correlation coefficients between the STD chronology and CRU data, including precipitation, PDSI, SPEI (a) and temperatures (T mean , T max , and T min ) (b).The correlations were computed from the previous year June to the current year September for the period of 1965-2018 C.E.The horizontal dotted solid lines represent significance at the level of p < 0.05 and * indicate significance at the level of p < 0.01 confidence limit.

Figure 5 .
Figure 5.The actual and estimated SPEI from July to September in Hindu Kush over the common calibration period (1965 to 2018 C. E.) (a).Reconstructed SPEI with an 11-year low pass fit filter (during.1626-2018 C.E).The central horizontal black dotted line represents the long-term mean SPEI (0.38 °C), whereas the horizontal dotted blue lines depicted the standard deviation (0.38 ±0.50 °C) (b).

Figure 6 .
Figure 6.Spatial correlation fields for reconstructed June-September SPEI data with gridded SPEI CRU TS 4.06 dataset (1965-2018 C. E.) (left panel; a); the right panel first-differenced data (right panel; b), correlation analysis was conducted using the KNMI Climate Explorer (http://climexp.knmi.nl/).The black circle is the sampling site in the Rumbur valley.

Figure 7 .
Figure 7. Comparisons between the SPEI reconstruction from July-September (a), the Hindu Kush PDSI reconstruction [31] (b), and the Karakoram precipitation reconstruction [8, 9] (c) and (d).The yellow-shaded region indicates variations in SPEI between this research and earlier reconstructions, whereas the grey and green-shaded areas highlight similarities of dry and wet periods, respectively.The r shows the correlation comparisons of SPEI reconstruction with Hindu Kush [31] drought and precipitation reconstructions form Karakoram [8, 9].* significant at p < 0.01; Raw stands for the un-filtered data, 11-yr LPF stand for 11-year lowpass filtered series.

Table 1 .
Site information and summary statistics of C. deodara TRW chronology.

Table 2 .
Calibration and verification statistics of the Rumbur (Hindu Kush) July-September SPEI reconstruction.

Table 3 .
After 11-year low pass filter wet and dry periods listing for 3-year during AD 1626-1964.Wet and dry periods given based on below and above the long-term mean (0.38).
Long term wet and dry periods listed before the instrumental data (1965-2018 C.E.).