FPFrancesca Pianosi

Francesca Pianosi

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years

Recognitions

Top Cited Scholar (2025)

Francesca Pianosi has a total of 193 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Comment on egusphere-2026-3267

    by Qidong Fang, Mostaquimur Rahman, Thorsten Wagener, Francesca Pianosi

    Deep Learning (DL) models, particularly Long Short-Term Memory (LSTM) networks, have shown similar or even superior performance to process-based models in estimating streamflow particularly at ungauged locations. However, their ability to extrapolate groundwater levels across time and space is less understood, as the number of studies addressing this issue is so far relatively limited. Here, we exploit the unique availability of a large-sample dataset of groundwater level observations across England to contribute to filling this gap. We configured two LSTM model variants: one using static environmental attributes (LSTM_ENV) and one using random integers as unique identifiers of places (LSTM_RND). Both models were trained using data from 636 stations over the period 1971-2014 and tested over 2015-2019 at both the training stations (in-sample test) and at 341 unseen stations (out-of-sample). Our results indicate that the two configurations achieved comparable performance in in-sample test, but their performances significantly diverge at unseen stations. To put the LSTM models’ performance into context, we also compared them to the performance of a process-based surface-groundwater model at 124 unseen stations. We found that both models effectively capture temporal fluctuations but struggle to accurately reproduce the mean and variability of the water table depth. This systematic bias frequently resulted in negative NSE values despite high temporal correlation, suggesting that evaluating LSTM performance using NSE solely can be misleading. We also found that the LSTM_ENV model performs better at stations characterised by higher specific yield and transmissivity, and that it mostly uses meteorological input features (e.g. precipitation) and topographic features (e.g. elevation and height above nearest drainage) to make predictions at unseen stations. These findings highlight the potential of LSTMs for regional groundwater level predictions and the value of interpretability tools for understanding how such models achieve their performance and whether the environmental features used are informative.

  • Open Access

    Comment on egusphere-2026-4470

    by Tamara Leins, Nikolai Späth, Jan Seibert, Francesca Pianosi, Christian Reinhardt-Imjela, Andreas Hartmann

    Subsurface Stormflow (SSF) is an important runoff-generation process, especially in temperate and mountainous regions. It can play a substantial role in catchment-scale flood generation and nutrient or contaminant transport. Due to process heterogeneity and subsurface occurrence, it is difficult to be quantified in the field. Since hydrological models are often calibrated using only discharge data at the catchment outlet, SSF-relevant parameters are often affected by equifinality, leading to large uncertainties in simulated SSF. In this study, we used the widely used bucket-type HBV model to analyse simulated SSF uncertainty across four catchments. We performed Monte Carlo simulations with 1,000,000 parameter sets and selected the top 5 % as behavioural. We subsequently analysed SSF parameter sensitivity in a regional sensitivity analysis and SSF simulation uncertainty in a GLUE-type approach. We examined whether more information on SSF is hidden in specific discharge conditions using a parameter-estimation approach based on percentiles of the flow-duration curve. In addition, we investigated simulated SSF uncertainty by disentangling simulated SSF volume and occurrence uncertainties. We found that there are high uncertainties in simulated SSF volume and occurrence across all catchments, with mostly low sensitivity of SSF-relevant parameters. We further found that SSF parameter identifiability could be improved by percentile parameter estimation, and more so in snow-insensitive catchments than in snow-dominated catchments. However, SSF simulation uncertainty still remained high. In the SSF uncertainty decomposition we could show that SSF occurrence uncertainty contributes a substantial part to total SSF simulation uncertainty. This study therefore emphasises the need to include SSF data into model calibration to ensure a realistic simulation of SSF and thereby achieve better process representation in lumped hydrological models and thus yield more robust prediction models. Our findings suggest that SSF proxy data indicating SSF occurrence in the catchment might already help to reduce SSF simulation uncertainty.

  • Open Access

    Comment on egusphere-2026-4470

    by Tamara Leins, Nikolai Späth, Jan Seibert, Francesca Pianosi, Christian Reinhardt-Imjela, Andreas Hartmann

    Subsurface Stormflow (SSF) is an important runoff-generation process, especially in temperate and mountainous regions. It can play a substantial role in catchment-scale flood generation and nutrient or contaminant transport. Due to process heterogeneity and subsurface occurrence, it is difficult to be quantified in the field. Since hydrological models are often calibrated using only discharge data at the catchment outlet, SSF-relevant parameters are often affected by equifinality, leading to large uncertainties in simulated SSF. In this study, we used the widely used bucket-type HBV model to analyse simulated SSF uncertainty across four catchments. We performed Monte Carlo simulations with 1,000,000 parameter sets and selected the top 5 % as behavioural. We subsequently analysed SSF parameter sensitivity in a regional sensitivity analysis and SSF simulation uncertainty in a GLUE-type approach. We examined whether more information on SSF is hidden in specific discharge conditions using a parameter-estimation approach based on percentiles of the flow-duration curve. In addition, we investigated simulated SSF uncertainty by disentangling simulated SSF volume and occurrence uncertainties. We found that there are high uncertainties in simulated SSF volume and occurrence across all catchments, with mostly low sensitivity of SSF-relevant parameters. We further found that SSF parameter identifiability could be improved by percentile parameter estimation, and more so in snow-insensitive catchments than in snow-dominated catchments. However, SSF simulation uncertainty still remained high. In the SSF uncertainty decomposition we could show that SSF occurrence uncertainty contributes a substantial part to total SSF simulation uncertainty. This study therefore emphasises the need to include SSF data into model calibration to ensure a realistic simulation of SSF and thereby achieve better process representation in lumped hydrological models and thus yield more robust prediction models. Our findings suggest that SSF proxy data indicating SSF occurrence in the catchment might already help to reduce SSF simulation uncertainty.

  • Open Access

    Modelling choices can outweigh climate variability in wind resource assessment

    by Saskia Salwey, Hannah Bloomfield, Francesca Pianosi

    Wind power is an important source of renewable energy that is critical for achieving net-zero. To inform planning and policy, wind power modelling tools are frequently used to estimate wind power potential both in the present day and in the future, accounting for changing climate conditions and changes in turbine distribution and efficiency. Despite our reliance on wind power models to inform high cost investments and decisions, the outputs from these models are dependent on a number of uncertain assumptions and choices. In this paper, we use a global sensitivity analysis approach to understand how much uncertainty affects the output of a typical wind power conversion tool, as well as which of the uncertainty sources contribute most to such total uncertainty. As model output, we consider the annual-mean capacity factor, both at the national aggregate level and at individual wind farms across the UK. We find that several of the modelling choices which are regularly generalised in national or large-scale modelling frameworks (e.g. choice of power curve, wind shear exponent and hub-height) have a strong control on the model output. Importantly, they have more influence than climate variability, suggesting that unless carefully setup and parameterized, wind power models may misrepresent the mean annual capacity factor—a commonly used metric for wind farm siting. Our results should act as a precaution for wind power modellers, where we hope to encourage more users to apply sensitivity analysis as shown here, and interpret wind power model output in the context of its uncertainty.

  • Open Access

    Simulating Subsurface Stormflow Across Catchments: Parameter Estimation and Uncertainties

    by Tamara Leins, Nikolai Späth, Jan Seibert, Francesca Pianosi, Christian Reinhardt-Imjela, Andreas Hartmann

    Subsurface Stormflow (SSF) is an important runoff-generation process, especially in temperate and mountainous regions. It can play a substantial role in catchment-scale flood generation and nutrient or contaminant transport. Due to process heterogeneity and subsurface occurrence, it is difficult to be quantified in the field. Since hydrological models are often calibrated using only discharge data at the catchment outlet, SSF-relevant parameters are often affected by equifinality, leading to large uncertainties in simulated SSF. In this study, we used the widely used bucket-type HBV model to analyse simulated SSF uncertainty across four catchments. We performed Monte Carlo simulations with 1,000,000 parameter sets and selected the top 5 % as behavioural. We subsequently analysed SSF parameter sensitivity in a regional sensitivity analysis and SSF simulation uncertainty in a GLUE-type approach. We examined whether more information on SSF is hidden in specific discharge conditions using a parameter-estimation approach based on percentiles of the flow-duration curve. In addition, we investigated simulated SSF uncertainty by disentangling simulated SSF volume and occurrence uncertainties. We found that there are high uncertainties in simulated SSF volume and occurrence across all catchments, with mostly low sensitivity of SSF-relevant parameters. We further found that SSF parameter identifiability could be improved by percentile parameter estimation, and more so in snow-insensitive catchments than in snow-dominated catchments. However, SSF simulation uncertainty still remained high. In the SSF uncertainty decomposition we could show that SSF occurrence uncertainty contributes a substantial part to total SSF simulation uncertainty. This study therefore emphasises the need to include SSF data into model calibration to ensure a realistic simulation of SSF and thereby achieve better process representation in lumped hydrological models and thus yield more robust prediction models. Our findings suggest that SSF proxy data indicating SSF occurrence in the catchment might already help to reduce SSF simulation uncertainty.

  • Open Access

    Comment on egusphere-2026-3267

    by Qidong Fang, Mostaquimur Rahman, Thorsten Wagener, Francesca Pianosi

    Deep Learning (DL) models, particularly Long Short-Term Memory (LSTM) networks, have shown similar or even superior performance to process-based models in estimating streamflow particularly at ungauged locations. However, their ability to extrapolate groundwater levels across time and space is less understood, as the number of studies addressing this issue is so far relatively limited. Here, we exploit the unique availability of a large-sample dataset of groundwater level observations across England to contribute to filling this gap. We configured two LSTM model variants: one using static environmental attributes (LSTM_ENV) and one using random integers as unique identifiers of places (LSTM_RND). Both models were trained using data from 636 stations over the period 1971-2014 and tested over 2015-2019 at both the training stations (in-sample test) and at 341 unseen stations (out-of-sample). Our results indicate that the two configurations achieved comparable performance in in-sample test, but their performances significantly diverge at unseen stations. To put the LSTM models’ performance into context, we also compared them to the performance of a process-based surface-groundwater model at 124 unseen stations. We found that both models effectively capture temporal fluctuations but struggle to accurately reproduce the mean and variability of the water table depth. This systematic bias frequently resulted in negative NSE values despite high temporal correlation, suggesting that evaluating LSTM performance using NSE solely can be misleading. We also found that the LSTM_ENV model performs better at stations characterised by higher specific yield and transmissivity, and that it mostly uses meteorological input features (e.g. precipitation) and topographic features (e.g. elevation and height above nearest drainage) to make predictions at unseen stations. These findings highlight the potential of LSTMs for regional groundwater level predictions and the value of interpretability tools for understanding how such models achieve their performance and whether the environmental features used are informative.

  • Open Access

    Exploring the generalisation ability and interpretability of Long Short-Term Memory (LSTM) networks for large-sample groundwater level predictions

    by Qidong Fang, Mostaquimur Rahman, Thorsten Wagener, Francesca Pianosi

    Deep Learning (DL) models, particularly Long Short-Term Memory (LSTM) networks, have shown similar or even superior performance to process-based models in estimating streamflow particularly at ungauged locations. However, their ability to extrapolate groundwater levels across time and space is less understood, as the number of studies addressing this issue is so far relatively limited. Here, we exploit the unique availability of a large-sample dataset of groundwater level observations across England to contribute to filling this gap. We configured two LSTM model variants: one using static environmental attributes (LSTM_ENV) and one using random integers as unique identifiers of places (LSTM_RND). Both models were trained using data from 636 stations over the period 1971-2014 and tested over 2015-2019 at both the training stations (in-sample test) and at 341 unseen stations (out-of-sample). Our results indicate that the two configurations achieved comparable performance in in-sample test, but their performances significantly diverge at unseen stations. To put the LSTM models’ performance into context, we also compared them to the performance of a process-based surface-groundwater model at 124 unseen stations. We found that both models effectively capture temporal fluctuations but struggle to accurately reproduce the mean and variability of the water table depth. This systematic bias frequently resulted in negative NSE values despite high temporal correlation, suggesting that evaluating LSTM performance using NSE solely can be misleading. We also found that the LSTM_ENV model performs better at stations characterised by higher specific yield and transmissivity, and that it mostly uses meteorological input features (e.g. precipitation) and topographic features (e.g. elevation and height above nearest drainage) to make predictions at unseen stations. These findings highlight the potential of LSTMs for regional groundwater level predictions and the value of interpretability tools for understanding how such models achieve their performance and whether the environmental features used are informative.

  • Open Access

    Evaluation of a socio-hydrological water resource model for drought management in groundwater-rich areas

    by Doris E. Wendt, Gemma Coxon, Saskia Salwey, Francesca Pianosi

    Groundwater is a drought resilient source of water supply for many water users globally. Managing these highly-used groundwater stores is complicated by the episodic nature of droughts and by our limited understanding of water systems’ response to extreme events. Models are useful tools to simulate a range of prepared drought interventions, however, we need to ensure robust representation of surface water and groundwater storage, users of both resources, and associated management interventions for drought resilience. A robust modelling approach is therefore essential for decision-making in groundwater management. In this study, we present a Socio-Hydrological Water Resource (SHOWER) model for drought management in groundwater-rich regions. We evaluate SHOWER using a response-based and a data-based model evaluation in Great Britain which considers the modelling uncertainty, dynamic impact of management and modelling setups available. In the response-based evaluation, we first examined the model consistency with our understanding of the system functioning and evaluated the influence of modelled management scenarios in normal and droughts conditions on discharge and groundwater levels. Secondly, in the data-based evaluation we tested the accuracy of heavily influenced discharge and groundwater level simulations in three catchments representative of typical hydrogeological conditions and water management practices in Great Britain. Results from the response-based method show consistent simulations for all model setups. We identified which parameters were influential to model output at which times. Integrated water management interventions have significant impact on flows and groundwater beyond parameter uncertainty and show leverage to reduce droughts by minimising shortages in water demand. The data-based analysis shows that calibration can be focused on either source-specific or combined model outputs using a “best overall” calibration approach that captures groundwater levels and low flows. The source-specific calibrations result in the highest and narrowest KGE ranges for discharge and groundwater (KGE: 0.75–0.84 and 0.62–0.95 respectively) with larger ranges using a “best overall” approach (KGE: 0.55–0.79 and 0.27–0.91). With the modular and open-access structure of SHOWER we aim to provide a useful new tool for groundwater managers to explore management interventions further, increasing drought resilience strategies using a robust modelling approach.

  • Open Access

    Widespread human footprints on groundwater trends across England

    by Qidong Fang, Mostaquimur Rahman, Thorsten Wagener, John Bloomfield, Francesca Pianosi

    Groundwater is a significant part of the global water cycle and an essential water source for domestic, industrial and agricultural use. In England, groundwater supplies ∼30% of public water, exceeding 75% in the most densely populated and water-stressed Thames and Southern regions. We analyse groundwater level trends of 2902 wells across England with 9–189 year records. We show that about half of the stations experience long-term trends or sudden changes. Long-term trends are very spatially heterogeneous. We investigate the links between these trends and various potential drivers and other proxy variables, and find that population density is the primary explanatory factor for increasing trends in densely populated urban areas, and irrigation intensity is the dominant factor for decreasing trends in intensive irrigated areas. Our results demonstrate that the spatial variability of groundwater trends across England is characterised by widespread anthropogenic footprints.