Explore the vast range of titles available.

Climate change and human activities are already affecting water resources in the Great Lakes region. A further projected increase in high-intensity precipitation events may cause flooding and added stress on urban infrastructure. In addition to climate, land use may change to accommodate a potential growing population of climate refugees and/or demand for farmland. Future climate and land use are anticipated to exacerbate stress experienced by surface water and groundwater domains. Due to substantial interaction between surface water and groundwater, storage of water was of interest due to its ability to incorporate the states of both systems and potentially buffer watershed responses to precipitation. Thus, quantification of hydrologic components such as storage is imperative for water researchers and policy makers to adequately prepare for future scenarios. The study aimed to analyze future climatic and human effects on water resources in Western New York (WNY) by developing an integrated surface water–groundwater GSFLOW model. Outputs from the EC Earth–RCA4 climate model under two emission scenarios were used in the research to represent an average-case climate scenario, RCP 4.5, and a worst-case climate scenario, RCP 8.5. Land use was adjusted by assuming an increase or a decrease in impervious surfaces. The GSFLOW simulations were organized into times slices representing historical, mid-century, and late century model runs. Model outputs of future time slices were compared to historical to analyze changes experienced by both domains, and statistical significance was determined using two-sample t tests. In addition, lag correlation analyses ascertained buffering capabilities of the groundwater system. Surface water and groundwater discharging into the Great Lakes as well as groundwater storage outputs were selected to represent the surface water, groundwater, and linked surface water–groundwater domains. Simulating future climate produced a comfortable response of the watershed under RCP 4.5 but significant declines in the surface water and groundwater systems under RCP 8.5. Findings suggested the majority of future results were not statistically significant relative to historical outputs under RCP 4.5, but changes were able to highlight hydrologic system dynamics. Neither the surface water nor groundwater system displayed any delayed response to precipitation. However, the groundwater system was less reactive to fluctuations in precipitation, where the surface water domain taking major losses to initially help sustain the groundwater domain. Gaining streams could be turning into losing streams. RCP 8.5 was characterized with both systems taking statistically significant and worrisome losses in mid-century and late century. Substantial losses in the groundwater domain highlighted even surface water was not able to buffer losses in groundwater under declining cumulative precipitation and increasing evapotranspiration. After simulating climate and land use changes, increasing or decreasing impervious surfaces in WNY had complex effects on water resources, such as spatially-varying groundwater head fluctuations. Under the city of Buffalo, increasing impervious surfaces resulted in a decline in groundwater head, which likely was due to less pervious recharge areas in the suburbs. On the other hand, decreasing impervious surfaces had no clear basin-wide or spatial effect. Statistical analyses showed the lack of clear responses of hydrologic components was not due to inadequate variation of impervious surfaces but rather due to complex effects of land use. This work shows even water-rich areas such as WNY are susceptible to extreme climate and watershed changes under RCP 8.5, although the surface water domain may be able to help alleviate the strain on the system under RCP 4.5. These results imply the Great Lakes region must be sufficiently prepared to adapt under variable water supplies with complex effects of future climate and land use. By extension, other surface water- and groundwater-rich areas are likely to have similar futures as WNY and should evaluate the current and future state of their water resources.

This paper presents the results of the 2022 Groundwater Time Series Modelling Challenge, where 15 teams from different institutes applied various data-driven models to simulate hydraulic-head time series at four monitoring wells. Three of the wells were located in Europe and one was located in the USA in different hydrogeological settings in temperate, continental, or subarctic climates. Participants were provided with approximately 15 years of measured heads at (almost) regular time intervals and daily measurements of weather data starting some 10 years prior to the first head measurements and extending around 5 years after the last head measurement. The participants were asked to simulate the measured heads (the calibration period), to provide a prediction for around 5 years after the last measurement (the validation period for which weather data were provided but not head measurements), and to include an uncertainty estimate. Three different groups of models were identified among the submissions: lumped-parameter models (three teams), machine learning models (four teams), and deep learning models (eight teams). Lumped-parameter models apply relatively simple response functions with few parameters, while the artificial intelligence models used models of varying complexity, generally with more parameters and more input, including input engineered from the provided data (e.g. multi-day averages). The models were evaluated on their performance in simulating the heads in the calibration period and in predicting the heads in the validation period. Different metrics were used to assess performance, including metrics for average relative fit, average absolute fit, fit of extreme (high or low) heads, and the coverage of the uncertainty interval. For all wells, reasonable performance was obtained by at least one team from each of the three groups. However, the performance was not consistent across submissions within each group, which implies that the application of each method to individual sites requires significant effort and experience. In particular, estimates of the uncertainty interval varied widely between teams, although some teams submitted confidence intervals rather than prediction intervals. There was not one team, let alone one method, that performed best for all wells and all performance metrics. Four of the main takeaways from the model comparison are as follows: (1) lumped-parameter models generally performed as well as artificial intelligence models, which means they capture the fundamental behaviour of the system with only a few parameters. (2) Artificial intelligence models were able to simulate extremes beyond the observed conditions, which is contrary to some persistent beliefs about these methods. (3) No overfitting was observed in any of the models, including in the models with many parameters, as performance in the validation period was generally only a bit lower than in the calibration period, which is evidence of appropriate application of the different models. (4) The presented simulations are the combined results of the applied method and the choices made by the modeller(s), which was especially visible in the performance range of the deep learning methods; underperformance does not necessarily reflect deficiencies of any of the models. In conclusion, the challenge was a successful initiative to compare different models and learn from each other. Future challenges are needed to investigate, for example, the performance of models in more variable climatic settings to simulate head series with significant gaps or to estimate the effect of drought periods.

This paper presents the results of the 2022 Groundwater Time Series Modelling Challenge, where 15 teams from different institutes applied various data-driven models to simulate hydraulic-head time series at four monitoring wells. Three of the wells were located in Europe and one was located in the USA in different hydrogeological settings in temperate, continental, or subarctic climates. Participants were provided with approximately 15 years of measured heads at (almost) regular time intervals and daily measurements of weather data starting some 10 years prior to the first head measurements and extending around 5 years after the last head measurement. The participants were asked to simulate the measured heads (the calibration period), to provide a prediction for around 5 years after the last measurement (the validation period for which weather data were provided but not head measurements), and to include an uncertainty estimate. Three different groups of models were identified among the submissions: lumped-parameter models (three teams), machine learning models (four teams), and deep learning models (eight teams). Lumped-parameter models apply relatively simple response functions with few parameters, while the artificial intelligence models used models of varying complexity, generally with more parameters and more input, including input engineered from the provided data (e.g. multi-day averages).

This paper presents the results of the 2022 Groundwater Time Series Modelling Challenge, where 15 teams from different institutes applied various data-driven models to simulate hydraulic-head time series at four monitoring wells. Three of the wells were located in Europe and one was located in the USA in different hydrogeological settings in temperate, continental, or subarctic climates. Participants were provided with approximately 15 years of measured heads at (almost) regular time intervals and daily measurements of weather data starting some 10 years prior to the first head measurements and extending around 5 years after the last head measurement. The participants were asked to simulate the measured heads (the calibration period), to provide a prediction for around 5 years after the last measurement (the validation period for which weather data were provided but not head measurements), and to include an uncertainty estimate. Three different groups of models were identified among the submissions: lumped-parameter models (three teams), machine learning models (four teams), and deep learning models (eight teams). Lumped-parameter models apply relatively simple response functions with few parameters, while the artificial intelligence models used models of varying complexity, generally with more parameters and more input, including input engineered from the provided data (e.g. multi-day averages). The models were evaluated on their performance in simulating the heads in the calibration period and in predicting the heads in the validation period. Different metrics were used to assess performance, including metrics for average relative fit, average absolute fit, fit of extreme (high or low) heads, and the coverage of the uncertainty interval. For all wells, reasonable performance was obtained by at least one team from each of the three groups. However, the performance was not consistent across submissions within each group, which implies that the application of each method to individual sites requires significanteffort and experience. In particular, estimates of the uncertainty interval varied widely between teams, although some teams submitted confidence intervals rather than prediction intervals. There was not one team, let alone one method, that performed best for all wells and all performance metrics. Four of the main takeaways from the model comparison are as follows: (1) lumped-parameter models generally performed as well as artificial intelligence models, which means they capture the fundamental behaviour of the system with only a few parameters. (2) Artificial intelligence models were able to simulate extremes beyond the observed conditions, which is contrary to some persistent beliefs about these methods. (3) No overfitting was observed in any of the models, including in the models with many parameters, as performance in the validation period was generally only a bit lower than in the calibration period, which is evidence of appropriate application of the different models. (4) The presented simulations are the combined results of the applied method and the choices made by the modeller(s), which was especially visible in the performance range of the deep learning methods; underperformance does not necessarily reflect deficiencies of any of the models. In conclusion, the challenge was a successful initiative to compare different models and learn from each other. Future challenges are needed to investigate, for example, the performance of models in more variable climatic settings to simulate head series with significant gaps or to estimate the effect of drought periods.