Quantifying uncertainty in annual runoff due to missing data.

Area of Science:

• Hydrology
• Environmental Science
• Data Science

Background:

• Long-term streamflow datasets often contain data gaps.
• Accurate estimation of runoff and catchment water budgets requires filling these gaps.
• The uncertainty associated with filling streamflow data gaps is seldom reported.

Purpose of the Study:

• To characterize and quantify the uncertainty introduced by filling streamflow data gaps.
• To assess the contribution of gap-filling to overall uncertainty in annual streamflow estimates.
• To evaluate the effectiveness of gap-filling methods in a reference watershed.

Main Methods:

• Simulated artificial gaps of varying durations and flow rates in streamflow data.
• Employed an ensemble of regression models to fill discharge data gaps using nearby streamflow records.
• Compared predicted streamflow with actual streamflow to quantify differences.

Main Results:

• Differences between predicted and actual runoff increased with gap length and flow rate, averaging 2.8% of runoff during the gap.
• Annual uncertainty due to gap filling at the Hubbard Brook Experimental Forest (HBEF) ranged from 0.2% to 2.1% of annual runoff.
• Uncertainty was found to be small at HBEF due to the use of multiple, similar catchments for infilling models.

Conclusions:

• Uncertainty in annual runoff due to streamflow data gap filling is generally small, particularly when using robust infilling models.
• The demonstrated method provides a framework for quantifying gap-filling uncertainty in any long-term streamflow dataset.
• Accurate gap filling is essential for reliable hydrological and water budget assessments.