The research, led by Zhicheng Zhong and Qiang Zhao of the University of Jinan along with colleagues in China and the United Kingdom, focuses on Shandong Province, a densely populated agricultural heartland on China’s eastern coast. Spanning the years 1959 to 2023, the study constructed two comprehensive drought indices using a statistical technique known as the C-vine copula. The first, CDIPRE, is anchored in precipitation, treating rainfall as the input to the regional water cycle. The second, CDIRUN, is anchored in runoff, treating streamflow as the output of that same cycle. By building both variants, the team could directly test a deceptively simple question: does it matter whether you measure drought from the top of the water cycle or the bottom?

To understand why this question matters, it helps to grasp what a copula actually does. In hydrology, the variables that govern drought—precipitation, runoff, potential evapotranspiration, soil moisture—do not vary independently. A dry month typically follows from large-scale atmospheric patterns that also suppress soil moisture and reduce river flow. Traditional indices often sidestep these dependencies by simply averaging or weighting standardized variables, an approach that can distort the true joint behavior of the water cycle. Copulas, by contrast, are mathematical functions that describe the dependence structure between random variables separately from their individual distributions. This separation allows scientists to model how variables co-vary, including in the extremes, where drought lives. Vine copulas extend this idea to many variables by decomposing a complex multivariate dependence into a cascade of simpler pairwise building blocks, and the C-vine variant organizes those pairs around a central hub variable—in this case, either precipitation or runoff.

The choice of hub turns out to be consequential. The researchers found that both CDIPRE and CDIRUN successfully integrate information about meteorological, hydrological, and agricultural drought, overcoming the central weakness of conventional univariate indices that track only one drought type at a time. But when the two indices were compared head to head, CDIRUN—the runoff-centered index—outperformed its precipitation-based sibling in capturing the characteristics of different drought types and in assessing agricultural drought impacts. This makes physical sense. Runoff integrates everything that happens upstream in the water cycle: rainfall deficits, evaporative losses, soil moisture depletion, and the slow release of water from storage. A deficit in runoff is, in a sense, the net verdict of the entire system, whereas a precipitation deficit is only the opening statement of a drought that may or may not develop into a genuine water crisis.

The study went beyond index construction to map the geography of drought across Shandong. Using run theory, a standard framework that defines drought events as periods during which an index remains below a threshold, the team analyzed drought duration, severity, and frequency across the province. The results revealed a striking spatial pattern: CDIRUN identified longer total drought durations in the north-central parts of Shandong, while other drought characteristics showed spatial distributions that depended on the scale of analysis. This scale dependence is a reminder that drought is not a single phenomenon but a family of processes operating at different temporal and spatial resolutions, and that any monitoring system must be explicit about which scale it is interrogating.

To probe why drought varies so much from place to place, the researchers turned to Geodetector, a statistical method designed to quantify how much of the spatial heterogeneity in a variable can be explained by candidate explanatory factors. The analysis identified the input variables underlying CDIRUN—precipitation, runoff, potential evapotranspiration, and soil moisture—as the dominant drivers of spatial heterogeneity in drought across the province. In other words, the same ingredients that feed the index are the ones that govern where drought bites hardest, which lends the index a degree of internal coherence that purely empirical composites often lack.

Perhaps the most intriguing part of the study reaches beyond Shandong entirely, into the Pacific Ocean. Drought in any given region is not purely a local affair; it is steered by large-scale climate oscillations that redistribute heat and moisture across the planet. The team examined the statistical relationships between their indices and a suite of climate teleconnection indices, and found that during the study period, the North Pacific Index showed the strongest statistical correlation with CDIRUN. Digging deeper, they identified the Northern Oscillation Index combined with the North Pacific Index as the statistically optimal bivariate combination, and a three-way combination adding the Southern Oscillation Index as the optimal multivariate set. The Southern Oscillation Index, of course, is the classic yardstick of the El Niño–Southern Oscillation, the planet’s most influential year-to-year climate fluctuation. The finding suggests that drought in this part of eastern China carries the fingerprint of Pacific-wide circulation patterns, opening a door toward longer-lead drought forecasting.

The practical implications are considerable. Shandong is a major grain producer, and its agricultural economy has long been exposed to the vagaries of a variable monsoon-influenced climate. Groundwater overexploitation in the wider North China Plain has compounded the region’s vulnerability, making accurate drought assessment a matter of food security as much as academic interest. An index like CDIRUN, which responds to the integrated state of the water cycle rather than to rainfall alone, could give water managers earlier and more reliable warnings of when a meteorological dry spell is turning into a genuine hydrological and agricultural crisis. The authors position the index as a tool for drought monitoring, water resource management, and agricultural planning under climate change, and the study’s structure supports that framing.

The work also sits within a broader scientific reckoning. Recent global analyses have shown that warming is accelerating drought severity worldwide, and reviews of widely used drought indices have catalogued the challenges of assessing drought under a changing climate, where the historical statistical relationships that indices rely on may themselves be shifting. Multivariate approaches built on copula theory have gained traction precisely because they can fuse multiple strands of hydrological information without discarding their dependence structure, and studies across basins from the Yangtze to the Qinghai-Tibet Plateau have explored similar composite frameworks. What distinguishes the new study is its systematic comparison of two anchoring strategies within the same copula framework, applied over a sixty-five-year record in a region where the stakes are high.

There are, as always, caveats and open questions. The analysis is regional, and the relative merits of runoff-centered versus precipitation-centered anchoring may differ in basins with different geology, land use, or degrees of human regulation. Runoff in heavily managed systems is shaped by reservoirs, irrigation withdrawals, and diversions, which can decouple it from the natural water cycle in ways that complicate interpretation. The authors note that data will be made available on request, and the study was supported by the National Natural Science Foundation of China and the Natural Science Foundation of Shandong Province. Still, the central message is likely to travel well beyond one province: drought assessment improves when it follows water through the entire cycle, from the sky to the soil to the stream. As climate change tightens the screws on regional water budgets, the indices we trust may need to look less like rain gauges and more like the water cycle itself.

Subject of Research: Copula-based multivariate drought indices integrating precipitation and runoff within the hydrological cycle

Article Title: Hydrological cycle-based comprehensive drought assessment: a comparative study of precipitation-centered cdipre and runoff-centered CDIRUN indices via c-vine copula

Article References: Zhong, Z., Zhao, Q., Xue, J., Dou, X., Zhao, Y., Wang, J., Liu, J., & Zhao, M. (2026). Hydrological cycle-based comprehensive drought assessment: a comparative study of precipitation-centered cdipre and runoff-centered CDIRUN indices via c-vine copula. Theoretical and Applied Climatology, 157(10), Article 608. https://doi.org/10.1007/s00704-026-06535-x

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06535-x

Keywords: drought, hydrological cycle, vine copula, runoff, precipitation, Shandong Province, climate teleconnections, North Pacific Index, soil moisture, agricultural drought, water resource management, climate change

The stakes are enormous. The UIB, fed by the Hindu Kush, Karakoram and Himalaya ranges, supplies more than 70 percent of the annual flow of the Indus River System and underpins roughly 90 percent of Pakistan’s agriculture and food production. Water from this basin sustains more than 200 million people and the world’s largest continuous irrigation network. Yet the region is among the most climate-sensitive on Earth: recent research has documented a warming and drying regime with precipitation declining at about 30 millimeters per decade, a reconstructed 651-year temperature record showing unprecedented warming in northern Pakistan, and significant roles for snow and glacier melt in buffering soil-moisture and hydrological drought.

Conventional drought monitoring relies on indices such as the Standardized Precipitation Index (SPI), which tracks precipitation deficits; the Standardized Precipitation Evapotranspiration Index (SPEI), which adds the atmospheric water balance; and the Standardized Precipitation Temperature Index (SPTI), which captures the joint effect of precipitation and temperature. Each illuminates only one facet of drought. The problem is that these facets interact in nonlinear, asymmetric ways—precipitation and temperature do not simply add up—and existing multivariate indices such as the Multivariate Standardized Drought Index (MSDI) and the Joint Standardized Drought Index (JSDI) tend to assume a single, symmetric dependence structure shared across all stations and climatic regimes. In a basin where glacier-fed and arid transitional catchments sit side by side, that assumption breaks down.

The new index, developed by Zanib Badar, Ishfaq Ahmad, Touqeer Ahmad and Ibrahim Mufrah Almanjahie and published in Environmental Science and Pollution Research, is built on vine copulas—statistical constructs that decompose a complicated multivariate dependence structure into a cascade of simple pairwise building blocks. The authors first transform each of the three component indices into uniform pseudo-observations using rank-based empirical cumulative distribution functions, a step justified by Sklar’s theorem, which guarantees that any joint distribution can be split into its marginals and a copula describing the dependence alone.

The heart of the method is a canonical vine, or C-vine, structure with SPTI as the root variable, chosen because it showed consistently stronger rank correlation (Kendall’s tau) with both SPI and SPEI across the study’s stations. In the first tree of the vine, unconditional pairwise dependencies between SPTI and the other two indices are modeled; in the second tree, the conditional dependence between SPI and SPEI given SPTI is captured. For each pair, the optimal copula family—drawn from a candidate set of twenty, including Gaussian, Student-t, Clayton, Gumbel, Frank, Joe, BB1, BB7 and their rotated variants—is selected by minimizing the Akaike information criterion. The result is a dependence model that can express asymmetric tail behavior, such as the tendency of precipitation-driven indices to produce joint extremes, which a Gaussian copula simply cannot represent.

From the fitted vine, the authors derive a Vine Degree of Influence for each component: the sum of the absolute Kendall’s tau values of every pair-copula in which that variable participates, across both trees. Normalizing these scores yields station-specific weights for combining SPI, SPEI and SPTI into the final VMDI. Because the weights incorporate conditional dependence, not just pairwise association, the index preserves the full nonlinear structure of the drought system while remaining a simple, interpretable weighted sum that water managers can compute and decompose. Validation against a fully nonlinear benchmark derived from the Rosenblatt transform showed Spearman correlations of 0.856 to 0.903 across stations, meaning the linear index retains roughly 73 to 83 percent of the multivariate signal—a deliberate trade-off that buys operational transparency.

The framework was tested at six meteorological stations—Astore, Bunji, Chilas, Gupis, Gilgit and Skardu—spanning elevations from about 1,200 to over 2,300 meters, using precipitation and temperature records from the Pakistan Meteorological Department and the Water and Power Development Authority for 1971 to 2017 at a one-month accumulation scale. Before merging, the authors confirmed the three indices carry complementary rather than redundant information: SPI and SPTI correlate strongly (Pearson r of 0.94 to 0.98 at most stations), but SPEI is nearly orthogonal to them at the glacier-influenced stations of Gilgit, Bunji and Gupis, signaling temperature-driven drought signals invisible to precipitation-based measures.

The performance results are striking. For precipitation-driven drought signals, VMDI achieved near-perfect agreement with SPI and SPTI—Pearson correlations above 0.91 and up to 0.99—while cutting mean absolute error by more than 50 percent relative to JSDI and MSDI, improving agreement by 15 to 30 percent, boosting skill scores by 30 to 40 percent and F1 detection scores by 33 to 37 percent. Its lower-tail dependence coefficients, which measure how well the index tracks joint drought extremes, exceeded 0.70 at all stations, reaching 0.89 at Gupis and 0.83 at Astore. In threshold-free ROC analysis, VMDI’s area under the curve ranged from 0.932 to 0.993 for SPI and SPTI, with the Nash–Sutcliffe efficiency peaking at 0.970 for SPTI at Astore. The Diebold–Mariano test confirmed that VMDI’s error reductions over the benchmarks were statistically significant in most comparisons, and it posted the lowest mean absolute error in 17 of 18 station-reference combinations.

The story is more nuanced for temperature-driven, evapotranspiration-related drought. At Gilgit and Bunji, where temperature anomalies are decoupled from precipitation, the empirical JSDI outperformed VMDI against SPEI by roughly 27 to 38 percent, with better mutual information, lower conditional entropy and stronger lower-tail dependence. The authors argue this is not a flaw but a feature of the vine weighting: those stations receive the lowest SPEI weights (about 0.227 to 0.229), appropriately down-weighting a signal that behaves independently there, while glacier-fed Astore and Skardu receive higher SPEI weights (0.274 and 0.283) reflecting tighter temperature–precipitation coupling through snowmelt. Robustness checks reinforced confidence: a leave-one-year-out analysis refitting the vine 47 times per station found the dominant copula family stable in at least 91 percent of replicates at four of six stations, and switching between near-tied families changed VMDI values by less than 0.05 standardized units on average. Cross-correlation and lag analyses confirmed the index introduces no artificial time offset, with dependence peaking at zero lag everywhere.

The practical upshot is a two-index strategy for complex mountain basins: VMDI as the primary tool where precipitation deficits dominate, and JSDI as a complement where heat-driven evaporative extremes prevail. Because VMDI requires only three widely computed indices and fixed station-specific weights, the authors say it can be slotted into existing early-warning systems without recalibration, and it remained robust across drought definitions from the 10th to the 25th percentile. Future work, they suggest, will extend the framework to longer accumulation scales of three, six and twelve months, fold in streamflow and soil moisture to track hydrological drought propagation, and interpolate station weights across terrain using elevation, aridity and glacial cover—steps that could help the world’s glacier-fed water towers anticipate drought with the sophistication the hazard demands.

Subject of Research: A vine-copula-based multivariate drought index for monitoring drought in high-elevation catchments of the Upper Indus Basin

Article Title: A novel multivariate vine-copula–based drought index with application to high-elevation catchments

Article References: Badar, Z., Ahmad, I., Ahmad, T., & Almanjahie, I. M. (2026). A novel multivariate vine-copula–based drought index with application to high-elevation catchments. Environmental Science and Pollution Research. https://doi.org/10.1007/s11356-026-38256-z

Image Credits: AI Generated

DOI: 10.1007/s11356-026-38256-z

Keywords: drought, vine copula, Upper Indus Basin, multivariate drought index, SPI, SPEI, SPTI, C-vine copula, high-elevation catchments, drought monitoring, Pakistan, early warning systems