Paola et al. (2026) Diagnosing catchment hydrological memory in the Po, Arno, Ticino and Adda basins (north-central Italy) through a multi-scale standardized drought index
Identification
- Journal: Journal of Hydrology Regional Studies
- Year: 2026
- Date: 2026-11-12
- Authors: Arianna Di Paola, Massimiliano Pasqui, Edmondo Di Giuseppe, Sara Quaresima, Leandro Rocchi, Elena Rapisardi, Emanuele Romano, Ramona Magno
- DOI: 10.1016/j.ejrh.2026.104179
Research Groups
- Institute of BioEconomy, National Research Council of Italy (CNR-IBE), Roma, Italy
- Institute of BioEconomy, National Research Council of Italy (CNR-IBE), Firenze, Italy
- Water Research Institute, National Research Council of Italy (CNR-IRSA), ADRM1, Montelibretti, Italy
Short Summary
This study extends the D(SPI) index to diagnose catchment hydrological memory across north-central Italian basins, finding that while effective memory extent is well-constrained (0.4 to 4.5 months), the detailed attenuation pattern of precipitation influence on streamflow is often not uniquely identifiable.
Objective
- To use the D(SPI) index as a diagnostic framework for characterizing catchment hydrological memory in standardized-anomaly space, testing whether alternative attenuation patterns of precipitation memory can be distinguished against standardized streamflow anomalies (SQI1), and characterizing the extent, seasonality, and uncertainty of this memory.
- To evaluate whether the response structures identified through D(SPI) are supported by established rainfall-runoff formulations (classical linear benchmarks).
Study Configuration
- Spatial Scale: Four river basins in north-central Italy: Po (approximately 71,000 km²), Arno (approximately 8,230 km²), Ticino (approximately 6,600 km²), and Adda (approximately 4,500 km²). These basins span Mediterranean pluvial, snow-dominated Alpine, and lake-regulated hydrological regimes.
- Temporal Scale: Monthly resolution, analyzing drought propagation at seasonal-to-interannual timescales (accumulation scales from 1 to 36 months). Data periods: Precipitation (1961–2025); Streamflow (Po: 1961–2025, Arno: 1961–2025, Ticino: 1961–1994, Adda: 1961–2016).
Methodology and Data
- Models used:
- D(SPI) index: An integrated multi-scale drought index, defined as a standardized weighted combination of SPI values (1 to K months). Five weighting families were tested: equal, linearly decreasing, geometrically decreasing, linearly increasing, and geometrically increasing.
- Standardized Precipitation Index (SPI): Computed for accumulation scales from 1 to 36 months.
- Standardized Streamflow Index (SQI1): Computed at a 1-month scale.
- Standardization: Probability distributions (Kernel Density Estimator, Gamma, Pearson type III, Gaussian) fitted independently for each calendar month, with Kernel Density Estimator (KDE) selected for most cases.
- Linear Benchmarks:
- Benchmark A: Unconstrained Unit Hydrograph (linear convolution of raw precipitation with a free kernel, target: raw streamflow).
- Benchmark B: Unconstrained convolution of standardized precipitation anomalies (linear convolution of SPI1 with a free kernel, target: SQI1).
- Benchmark C: Nash-type two-parameter Gamma kernel on standardized anomalies (convolution of SPI1 with a Gamma kernel, target: SQI1).
- Uncertainty Estimation: Paired moving-block bootstrap (block length L = 72 months, B = 1000 replicates) applied to the entire pipeline.
- Data sources:
- Gridded daily precipitation: ARCIS (high-resolution approximately 5 km observational dataset covering north-central Italy, 1961–present, from over 1700 rain gauges).
- Monthly streamflow:
- Po at Pontelagoscuro: ARPAE-SIMC Dext3r portal.
- Arno at S. Giovanni alla Vena: Centro Funzionale of Regione Toscana.
- Ticino at Miorina (Lake Maggiore outlet) and Adda at Olginate (Lake Como outlet): Global Runoff Data Centre (GRDC).
- Basin shapefiles: EStreams dataset.
Main Results
- Catchment Memory Characterization: The study found that the effective extent of catchment memory (mean age of precipitation influence) is better constrained by the data than the detailed internal attenuation pattern. Multiple weighting families often achieve statistically indistinguishable peak R² values at different optimal accumulation window lengths (K), but their effective memories (mean age) are much closer.
- Effective Memory Ranges: Effective memory (mean age) varies significantly across basins: 0.4 to 1.5 months in the Arno, 1.2 to 3.4 months in the Ticino, 1.5 to 4.3 months in the Po, and 1.9 to 4.5 months in the Adda.
- Seasonal Variability: Strongest precipitation-streamflow coupling (R² ≳ 0.8) occurs at short-to-medium accumulation scales (K = 2–6 months) during autumn in all basins. Summer months (July–August) show the weakest coupling (R² rarely exceeding 0.6). The Po and Adda exhibit more stable precipitation-streamflow relationships throughout the year, while the Arno and Ticino show stronger seasonality and periods of weak coupling.
- Benchmark Comparison: Benchmark models (linear Unit Hydrograph, OLS convolution of SPI1, Nash-type Gamma kernel) and D(SPI) achieve broadly comparable levels of agreement with observed streamflow anomalies. All formulations tend to weaken simultaneously in the same basin-season combinations, suggesting that the primary limitation lies in the precipitation-streamflow relationship itself rather than the specific memory representation.
- Dominant Influence: All benchmark formulations and D(SPI) (at their optimal K) agree that precipitation influence is concentrated in the most recent months, with rapid attenuation thereafter. Increasing model complexity beyond parsimonious representations yields limited gains in R² at monthly resolution.
- Drought Underestimation: In the Po, Ticino, and Adda basins, D(SPI) systematically remains less negative than SQI1 during the most severe drought episodes (SQI1 ≲ −2.5), suggesting that additional processes (e.g., high temperatures, increased water withdrawals) become relevant during extreme events and are not fully captured by precipitation history alone.
Contributions
- Introduced the R²(w, K) response surface as a novel diagnostic signature for characterizing catchment hydrological memory, formalizing the optimization of the D(SPI) index in the joint parameter space of weighting families (w) and accumulation window lengths (K).
- Demonstrated that, at monthly resolution, the effective extent of catchment memory is robustly constrained by observations, while the detailed internal attenuation pattern of precipitation influence is often not uniquely identifiable.
- Provided a framework that bridges fixed-timescale drought indices and transfer-function approaches, allowing both the extent and internal temporal structure of catchment memory to be inferred from observations within a standardized-anomaly framework.
- Showed that D(SPI) achieves comparable explanatory power to classical linear rainfall-runoff benchmarks, supporting parsimony as a useful principle for monthly drought diagnostics.
- Revealed reproducible, basin-specific hydrological memory signatures across north-central Italian basins, highlighting differences in effective memory, seasonal stability, and structural ambiguity, which are consistent with physiographic and management characteristics.
- Established a diagnostic tool for interpreting drought propagation mechanisms and catchment memory across heterogeneous catchments, rather than a predictive model.
Funding
This research received no specific external funding. The work was carried out within the broader, non-funded conceptual and operational framework of Drought Central (www.droughtcentral.it) and as part of the activities of the thematic working group “Drought, Scarcity, and Water Crisis” of the Department of Earth System Sciences and Environmental Technologies of the National Research Council of Italy (https://dta.cnr.it/gruppi-di-lavoro/siccita-e-crisi-idriche/).
Citation
@article{Paola2026Diagnosing,
author = {Paola, Arianna Di and Pasqui, Massimiliano and Giuseppe, Edmondo Di and Quaresima, Sara and Rocchi, Leandro and Rapisardi, Elena and Romano, Emanuele and Magno, Ramona},
title = {Diagnosing catchment hydrological memory in the Po, Arno, Ticino and Adda basins (north-central Italy) through a multi-scale standardized drought index},
journal = {Journal of Hydrology Regional Studies},
year = {2026},
doi = {10.1016/j.ejrh.2026.104179},
url = {https://doi.org/10.1016/j.ejrh.2026.104179}
}
Original Source: https://doi.org/10.1016/j.ejrh.2026.104179