Comparison of on-site versus NOAA’s extreme precipitation intensity-duration-frequency estimates for six forest headwater catchments across the continental United States
Given the continually available digitized up-to-date, long-term, and fine resolution precipitation dataset from the United States Department of Agriculture Forest Service’s (USDAFS) Experimental Forests (EF) based rain gauge stations, it is both important and relevant to develop onsite Precipitation-Intensity- Duration-Frequency (Onsite-PIDF) estimates that incorporate the recent effect of climate change, aiding in the design, and planning of forest road-stream crossing structures (RSCS) to maintain resilient forest ecosystems. Despite such available resources, the RSCS on US National Forests and similar other landscapes are commonly designed using the sub-hourly and sub-daily precipitation intensities (PIs) provided by the National Oceanic and Atmospheric Administration Atlas 14 (NOAA-Atlas14), which have multiple limitations, including, lack of recency, irregularity and remoteness of NOAA station locations, unavailability of PI estimates for multiple forest locations, and oversimplified derivation of partial duration series (PDS) and assumption of stationarity in PIDF analysis. To address some of these key issues, we estimated Onsite-PIDFs using annual maxima series (AMS), as well as peak-over-threshold (POT) series of PIs fit to stationary and non-stationary Generalized Extreme Value (GEV) and Generalized Pareto (GPA) distributions, respectively, with L-moment and Bayesian parameter estimation methodologies at six USDAFS EFs. Onsite-PIDFs are then evaluated against NOAA-Atlas14 based on the relative differences in PIs specific to given design return intervals (RIs) and durations of storm events, and through the comparison of onsite RIs specific to PIs based on NOAA’s 25-yr, 50-yr, and 100-yr RIs. Results show considerable relative differences between the Onsite- and NOAA-Atlas14 PIs at all EF station locations, which are strongly dependent on the storm durations and elevation of stations, particularly in Horace Justin Andrews (HJA) EF, and are found to be significantly higher for the POT based non-stationary PIDFs. Compared to the onsite stationary AMS- and POT-PIDFs, NOAA-Atlas14 PIDFs substantially underestimated PIs, especially, for longer (> 1hr) duration storms in the Santee (SAN), and overpredicted PIs for the stations located in Coweeta Hydrological Laboratory (CHL), Alumn Creek (ALC), Fraser (FRS), and Hubbard Brook (HBR) EF. More importantly, at the majority of EF station locations, NOAA PIDFs lie within the uncertainty range of onsite non-stationary PIDFs, however, they considerably underestimate the PIs, and frequency of storm events, especially, at the HJA, SAN, CHL, ALC, and HBR EF station locations. Our results recommend use of Onsite-PIDF estimates for the estimation of design storm peak discharge rates at the EF locations, where recent changes in extreme precipitation storm events exhibit non-stationarity effects of anthropogenically forced climate change.