Sequential estimation of aquifer parameters to assess groundwater pumping in data scarce regions

Accurate estimation of groundwater abstraction remains challenging in data-scarce regions where pumping records are rarely available and groundwater models are commonly calibrated using only limited parameter information (e.g., specific yield), while others are adopted from literature. This practice can propagate bias and uncertainty into model predictions. We hypothesize that groundwater abstraction can be estimated more reliably when aquifer and pumping parameters are identified sequentially, rather than simultaneously, through iterative conditioning of the parameter space.To evaluate this concept, we applied a simplified three-parameter groundwater model and generated 20 synthetic groundwater time series, each with unique pumping inputs. When all parameters were estimated simultaneously, strong parameter correlations produced large uncertainties in pumping estimates, with errors ranging from –26% to +185%. To overcome this, we implemented a sequential GLUE (Generalized Likelihood Uncertainty Estimation) framework, performing 100,000 Monte Carlo simulations per well per iteration. In each iteration, parameters that showed clear convergence—indicated by narrowing behavioural ranges and reduced coefficients of variation—were fixed before proceeding to the next iteration. This sequential reduction of the feasible parameter space substantially improved parameter identifiability and reduced pumping-estimation uncertainty, yielding abstraction estimates within ±10% of the prescribed synthetic values.The framework was subsequently applied to 75 observed groundwater time series from field wells (2,600 observations), demonstrating that the sequential approach improves recovery of aquifer parameters and produces realistic estimates of pumping even where no pumping data exist. The results highlight the ability of sequential parameter estimation to mitigate equifinality, expose model structural errors (e.g., when pumping is omitted), and enhance the use of simple groundwater models in data-poor regions.Overall, this study demonstrates that iterative/sequential parameter identification offers a practical and efficient pathway for estimating aquifer parameters, supporting the development of more complex 2-D and 3-D numerical models, and enabling realistic estimation of groundwater abstraction and aquifer properties in regions with limited hydrological information.