Date of Award

1-1-2021

Language

English

Document Type

Master's Thesis

Degree Name

Master of Science (MS)

College/School/Department

Department of Atmospheric and Environmental Sciences

Content Description

1 online resource (xii, 85 pages) : color illustrations, color maps.

Dissertation/Thesis Chair

Justin R Minder

Committee Members

Junhong Wang

Keywords

National Water Model, New York State Mesonet, Noah-MP, Northeast United States, Snow, SWE, Mesometeorology

Subject Categories

Atmospheric Sciences | Hydrology | Meteorology

Abstract

Snow is a critical component in the hydrologic cycle and critical to runoff in many regions. While not as deep or persistent as snow in the Western United States (WUS) , snow in the Northeastern US (NEUS) is critical to water resource management and flood forecasting. For hydrological applications, snow is simulated using coupled hydrology models . These models couple numerical weather models, land surface models, and channel routing models that simulate water transport. One such coupled hydrology model is the NOAA National Water Model (NWM), implemented in 2016. The NWM runs a specific configuration of the WRF-Hydro community model. The underlying LSM used to simulate snow processes is Noah-MP. The goal of this research is to evaluate the performance of NWM snow simulations in the NEUS through: (1) determining the effects of forcing error on NWM snow simulations in the NEUS, (2) evaluating snow accumulation and ablation errors to target physics parameterizations for improvement, (3) to evaluating if difference in land-use classification between model and observations contribute to model error. To explore these goals, model output and meteorological forcing from retrospective runs of NWMv2.0 and NWMv2.1 with data available from 2010-2021 are analyzed. Observations from: the NYS Mesonet (NYSM), a network of 126 weather stations with 20 snow water equivalent (SWE) monitoring sites, NY Snow Survey (NYSS), a collection of snow course measurements conducted around New York roughly every two weeks from January to April every winter dating back to the late 1930s , and snow course measurements taken at a subset of 5 NYSM SWE sites as part of this research are used to evaluate model performance. A comparison of NWMv2.0 and NWMv2.1 snow simulations are conducted for the winter 2017/18 season using NYSM snow depth measurements. Additionally, SWE from both model versions is compared with NYSS snow survey data from 2010-2018. Forcing precipitation and temperature is analyzed for both versions to determine the contribution of forcing error to snow simulation error Finally, NWMv2.1 and forcing is analyzed against, NYSM SWE, snow depth, and meteorological data as well as research snow survey data for the 2019-2021 winters. Results of the model comparison show that NWMv2.1 performs better than NWMv2.0 but both versions under simulate snow in the NEUS. In the winter 2017/18 analysis, NWMv2.1 improved on snow depth bias and RMSE. Regional median time series in the Adirondack Mountains, Tug Hill Plateau and Catskill Mountains showed that NWMv2.1 better simulated date and magnitude of max snow depth for all regions, with the largest improvements seen in the Tug Hill Plateau. Date of melt-out was the same for both versions in all three regions, even though snow depth was substantially lower in NWMv2.0 leading up to melt-out. The results of the NYSS SWE analysis showed similar results. The largest improvements where NWMv2.0 had an average bias of -89mm while NWMv2.1 had an average bias of -51mm. While -51mm is a substantial bias, its an improvement of 20% from NWMv2.0. Analysis of the meteorological forcing biases for the 2017/18 winter found negative precipitation biases in both versions with the forcing for NWMv2.0 having mean daily precipitation biases of -0.63 in the Adirondack Mountains, -1.7mm in the Tug Hill Plateau, and -0.51mm in the Catskill Mountains. NWMv2.1 showed improvement with daily mean precipitation biases of -2.6mm in the Adirondack Mountains, -0.43mm in the Tug Hill Plateau, and -0.43mm in the Catskill Mountains. Temperature was largely biased positive in the forcing for NWMv2.0 while relatively unbiased in NWMv2.1. In the Adirondack Mountains, the average daily mean temperature bias was 0.96oC in NLDAS-2 and 0.13oC in AORC. In the Tug Hill Plateau, the average daily mean temperature bias was 1.13oC in NLDAS-2 and 0.12oC in AORC. In the Catskill Mountains, the average daily mean temperature bias is 0.41oC in NLDAS-2 and 0.19oC in AORC. Analysis of the correlation between forcing and snow depth bias during the 2017/18 winter found significant positive correlations between precipitation bias and snow depth bias in both model versions. A significant negative correlation was seen between NWMv2.0 temperature and snow depth bias while there was no correlation between temperature and snow depth bias in NWMv2.1, likely due to temperature being relatively unbiased. Analysis of NWMv2.1 in the 2019-2021 winter showed mixed results. The mean snow depth bias in the Adirondacks was -0.03 m with a maximum positive bias of 0.56 m and a maximum negative bias of -0.36m. The mean snow depth bias in the Tug Hill Plateau was -0.06 and, in the Catskills, the mean snow depth bias was -0.01m. The mean SWE bias at NYSM snow sites for the 2019 winter was -4mm. The maximum positive SWE bias was 55mm and the maximum negative bias was -137mm. The regional average mean daily precipitation biases were -0.52 mm/day in the Adirondack Mountains, -0.32mm in the Tug Hill Plateau, and -0.9mm in the Catskill Mountains. The regional average daily mean temperature biases were 0.19oC in the Adirondack Mountains, 0.29oC in the Tug Hill Plateau, and 0.23oC in the Catskill Mountains. For winter 2019/20, in the Adirondack Mountains, the mean daily snow depth bias was 0.14m. In the Tug Hill Plateau, the daily mean snow depth bias 0.08m. In the Catskill Mountains, the daily snow depth bias was 0.12m. Mean SWE bias at NYSM snow sites for this season were 39mm and the maximum SWE bias was 76mm. The regional average mean daily precipitation biases were -0.45mm in the Adirondack Mountains, -0.31mm in the Tug Hill Plateau, and +0.03mm in the Catskill Mountains. The regional daily mean temperature biases were 0.16oC in the Adirondack Mountains, 0.13oC in the Tug Hill Plateau, and 0.25oC in the Catskill Mountains. For winter 2020/21, the daily mean snow depth bias in the Adirondack mountains were -0.05m. In the Tug Hill Plateau, daily mean snow depth bias was 0.06m. In the Catskill Mountains, daily mean snow depth bias was 0.10m. SWE network mean bias was -2.6mm and maximum bias was 50mm. The regional average mean daily precipitation biases were -0.32mm in the Adirondack Mountains, +0.21mm in the Tug Hill Plateau, and +1.1mm in the Catskill Mountains. Regional average daily mean temperature were -0.19oC in the Adirondack Mountains, -0.28oC in the Tug Hill Plateau, and -2.3oC in the Catskill Mountains. Analysis of research snow courses showed that the observed error in snow simulations could not be contributed to the difference in land-use classification between the model and observations. Even though all the model grid cells analyzed were considered forested, simulated snow depth and SWE was more consistent with observations in clearings than in the forests. The results show that simulations got timing of max SWE, magnitude of max SWE, and date of melt-out wrong.

Download

Share

COinS