Brunner et al. (2019) investigated the extremeness of the 2018 drought event in Switzerland compared to the 2003 and 2015 events by looking at four variables, i.e. observed precipitation, modelled discharge, soil moisture, and low-flow storage. The latter variables were simulated using the PREVAH hydrological model for 307 medium-sized catchments in Switzerland for the period 1981–2018. This dataset extended the observed discharge dataset in space and time and added the variables soil moisture and low-flow storage. For each of these variables, the two characteristics deficit and deficit duration were considered. The variables were first considered separately in a univariate frequency analysis; pairs of variables were then investigated jointly in a bivariate frequency analysis using a copula model for expressing the dependence between the two variables under consideration. Their findings demonstrated that the use of univariate and bivariate analyses can lead to different severity estimates for individual catchments and divergent conclusions on the extremeness of individual events, which depend on location, variable(s), and the problem under consideration (i.e. the type of return period chosen).

Longobardi et al. (2021) assessed drought features in the Campania region (southern Italy) through an analysis of the spatial and temporal patterns of standardized precipitation index (SPI) time series, based on an in situ measurement database which covers a centennial period from 1918 to 2019. SPI time series were reconstructed for different accumulation timescales (from 3 to 48 months), and the modified Mann–Kendall and Sen tests were applied to identify SPI changes over time. SPI time series were mostly affected by a negative trend, significant for a very large area of the region. Concerning the spatial pattern, negative trends appeared to occur along a northwest–southeast transect, whereas positive trends were observed along a west–east transect at the middle latitude of the region. Those two regions are quite different in terms of orography, which likely influences both the average precipitation spatial distribution and the relevant temporal variability. Also, the accumulation timescale was proven to affect the spatial patterns of the drought characteristics.

Yue et al. (2022) explored the effects of large reservoirs on drought resistance in the Hun River basin (HRB). The standardized runoff index (SRI) was adopted to evaluate the evolution of hydrological drought from 1967 to 2019. Meanwhile, the joint probability of drought duration and severity, identified by the run theory, was computed using the copula function with the highest goodness of fit in order to calculate the return period of hydrological drought events. The authors also evaluated the delay between meteorological and hydrological droughts by calculating the Pearson correlation coefficients between a 1-month SRI and a multi-timescale standardized precipitation index (SPI). Finally, based on the cumulative precipitation deficit thresholds for triggering hydrological drought, the impact of large reservoirs on drought resistance of the basin was revealed. Among the various results, they found that the operation of large reservoirs strengthened the drought resistance in the lower reaches, while it is slightly weakened in the upper reaches of large reservoirs.

Ansari and Grossi (2022) investigated the spatial and temporal evolution of wet–dry events collectively, their characteristics and transition (wet to dry and dry to wet) in the Upper Jhelum basin (UJB)-in South Asia, dominated by two precipitation patterns: westerlies in the north-facing slopes and monsoon in the south-facing slopes of the Himalaya mountain range, crossing the basin in the middle. The standardized precipitation evapotranspiration index (SPEI) at the monthly timescale was applied to detect and characterize wet and dry events for the period 1981–2014. The analysis of the temporal variations of the SPEI showed a strong change in basin climatic features associated with El Niño–Southern Oscillation (ENSO) at the end of 1997, with the prevalence of wet and dry events before and after 1997, respectively. The results of the spatial analysis show a high susceptibility of the monsoon-dominated region towards wet events, whereas the westerlies-dominated region was found to be the hotspot of dry events with higher duration, severity, and intensity. Moreover, the region surrounding the Himalaya divide line and the monsoon-dominated part of the basin were found to be the hotspots of rapid wet–dry transition events.

Monteleone et al. (2020) proposed a new composite index for agricultural drought monitoring, namely the probabilistic precipitation vegetation index (PPVI). This new index accounts for both meteorological and agricultural drought conditions by combining in a probabilistic framework the SPI and the vegetation health index (VHI). In addition, they developed and implemented a set of rules to objectively identify drought events. Both the index and the set of rules have been applied to a case study in Haiti. The performance of the PPVI has been evaluated by means of a receiver-operating characteristic (ROC) curve and compared to that of the SPI and VHI considered separately. Comparisons between observed and modelled drought events showed that the new index outperforms SPI and VHI in terms of properly identified and characterized drought events, thus revealing the potential for an effective implementation within drought early-warning systems. The main advantages of the PPVI are that (1) it requires few input data (i.e. only precipitation and the VHI), (2) it is a remote-sensing product and therefore easily transferable and scalable over the entire globe, and (3) it is suitable for a timely implementation of drought mitigation strategies due to the relatively short latency time (less than 1 week) of the datasets employed.

Popat and Döll (2021) proposed two drought indices: the soil moisture deficit anomaly index, SMDAI, based on a simplification of the drought severity index, DSI (Cammalleri et al., 2016), and the streamflow deficit anomaly index, QDAI. The first index describes the drought hazard for vegetation, while the second one specifically quantifies the hazard that drought poses for water supply from rivers. Both indices are computed and analysed at the global scale, with a spatial resolution of roughly 50 km, for the period 1981–2010, using monthly time series of variables computed by the global water resources and the WaterGAP model. Although the two indices are broadly similar to values of purely anomaly-based indices, the deficit anomaly indices provide more differentiated spatial and temporal patterns that help to distinguish between the degree and nature of the actual drought hazard to vegetation health or water supply. In particular, the QDAI can serve as a tool for informing water suppliers and other stakeholders about the joint drought hazard for water supply for both humans and the river ecosystem. Like all hydrological drought indicators that reflect the streamflow anomaly, the QDAI needs to be interpreted carefully in case of highly intermittent streamflow regimes.

Peres et al. (2020) proposed a statistical methodological framework to assess the quality of 19 EURO-CORDEX RCMs at 0.11^∘ (∼ 12.5 km) grid spatial resolution, mainly focusing on their ability to simulate climate and drought characteristics (duration, accumulated deficit, intensity, and return period) under current climate forcing. The proposed methodology exploits comparisons with high-density and high-quality ground-based observational datasets, and it was applied to the Sicily and Calabria regions (southern Italy), where long historical precipitation and temperature series were recorded by the ground-based monitoring networks operated by the former regional hydrographic offices. Results show that, among the more skillful models able to reproduce precipitation and temperature variability as well as drought characteristics, several are based on the CCLM-Community RCM, particularly in combination with the HadGEM2 global circulation model (GCM). The proposed study reveals insight into RCM performances in small-scale regions, which are often the target of impact studies but have so far received less attention. It also provides some guidance to select the best models before addressing projection changes in the evolution of extreme hydro-meteorological events, such as drought characteristics.

Richardson et al. (2020) compared the performance of a dynamical sub-seasonal forecast system (EPS-WP) and a first-order Markov model in predicting European weather pattern (WP) occurrences over a range of lead times, showing that the dynamic model is always more skillful, although the difference in skill reduces with lead time. The proposed forecast system was applied in UK meteorological drought prediction. After post-processing mean sea-level pressure forecasts from the European Centre for Medium-Range Weather Forecasts Ensemble Prediction System (ECMWF-EPS) into probabilistic WP predictions, precipitation estimates were derived, and dichotomous drought event probabilities were estimated by sampling from the conditional distributions of precipitation given the WPs. Model estimates were compared with the direct precipitation and drought forecasts from the ECMWF-EPS and with a baseline Markov chain WP method by using a range of skill diagnostics. They found that the Markov model was the least skillful, while the dynamical WP model and direct precipitation forecasts have similar accuracy regardless of lead time and season. However, drought forecasts are more reliable for the dynamical WP model.

Shukla et al. (2020) tested the suitability of the operational root zone soil moisture (RZSM) products from the recently developed NASA Hydrological Forecasting and Analysis System (NHyFAS) for supporting early warning of droughts in southern Africa (SA). The results indicated that the NHyFAS products can effectively support food insecurity early warning in the SA region. In particular, February RZSM forecasts available in early November (4–5 months before the start of harvest and about 1 year before the start of the next lean season) can explain the interannual variability in regional crop yield production with moderate skill (correlation coefficient of 0.49). The February RZSM monitoring product, available in early March (1–2 months before the start of harvest and 8–9 months before the start of the next lean season) can explain the variability in regional crop yield with high skill (correlation coefficient of 0.79). Furthermore, the February RZSM monitoring product can provide “out-of-sample” crop yield forecasts with higher skill than DJF ENSO (38 % reduction in mean error relative to DJF ENSO), whereas the February RZSM forecasting product, available in early November, can provide crop yield forecasts with comparable skill (∼6 % increase in mean error relative to DJF ENSO). Since the datasets and models used to implement the NHyFAS are available globally, a similar seasonal RZSM monitoring and forecasting framework can potentially be developed at a global scale to support food insecurity early warning in other rainfed regions across the globe.

Sutanto et al. (2020) evaluated the skills of both drought hazard and impact forecasts, which were developed using observed and re-forecast hydro-meteorological data and drought impact reports retrieved from the European Drought Impact Report Inventory (EDII). Empirical drought impact functions were developed by using machine learning approaches (logistic regression and random forest) to predict drought impacts with lead times up to 7 months ahead. The observed and forecasted hydrometeorological drought hazards – in terms of the SPI, SPEI, and SRI – were obtained from the EU-funded Enhancing Emergency Management and Response to Extreme Weather and Climate Events (ANYWHERE) Drought Early-Warning Systems (DEWS). Results show that hydrological drought hazard represented by the SRI has higher skill than meteorological drought represented by the SPI and SPEI for the same accumulation period due to the memory represented in initial hydrological conditions and storage in the hydrological system. For German regions, impact functions developed using random forests indicate a higher discriminative ability to forecast drought impacts than logistic regression. Moreover, skill is higher for cases with higher spatial resolution and less lumped impacted sectors.

Felsche and Ludwig (2021) applied artificial neural networks to predict the drought occurrence in Munich and Lisbon with a lead time of 1 month. The approach takes into account a list of 28 atmospheric and soil variables as input parameters from a single-model initial-condition large ensemble (CRCM5-LE). The data were produced with the Canadian Regional Climate Model (CRCM5) driven by 50 members of the Canadian Earth System Model (CanESM2). Drought occurrence is defined as an SPI-1 less than −1 at the given site. The best-performing machine learning algorithms manage to obtain a correct classification of drought or no drought for a lead time of 1 month for around 55 %–57 % of the events of each class for both domains. Variables like the North Atlantic Oscillation index and air pressure 1 month before the event prove to be essential for the prediction.

Wang et al. (2020) explored the link between drought indices and drought impacts by using correlation and random forest methods. In particular, the study identified which indices link best to a comprehensive database of reported drought impacts for prefectural-level cities in Liaoning Province (north-eastern China) and proposed a drought vulnerability assessment method to study the contribution of various socio-economic factors to drought vulnerability. The results show that the standardized precipitation evapotranspiration index with a 6-month accumulation (SPEI-6) had a strong correlation with all categories of drought impacts. Among the impact datasets, “drought-suffering area” and “drought impact area” had a strong relationship with all drought indices in Liaoning Province, while “population and number of livestock with difficulty in accessing drinking water” had weak correlations with the indices. Overall, population and crop-cultivated area were strongly associated with drought vulnerability, suggesting these factors are good proxy variables of drought vulnerability. However, the complexities of these relationships with drought vulnerability require further investigation.

Hasan et al. (2021) investigated the low-flow characteristics and the river storage draft rates in Selangor (Malaysia), based on the long-term streamflow data from seven stations. In particular, the mass curve was used to quantify the minimum storage draft rate required to maintain the 50 % mean annual flow for the 10-year recurrence interval of low flow. The low-flow conditions were analysed using the 7 d mean annual minimum, while the drought events were determined using the 90th percentile as a threshold level. The study reveals September to December to be a critical period in river water storage to sustain the water availability during low flow in a 10-year occurrence interval. This finding indicated that hydrological droughts have generally become more critical for the capability of rivers to sustain water demand during low-flow periods.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.