1. Introduction
A comprehensive understanding of the wind resource in the New York Bight (NYB; the region of ocean waters south of Long Island and east of New Jersey; Gunnerson 1981) is essential to the ongoing development and operation of offshore wind energy facilities. The lack of offshore meteorological measurements, especially above the surface through the top of the marine atmospheric boundary layer (MABL), makes it necessary to supplement observational data with model output.
Onshore networks such as the New York State Mesonet (NYSM; Brotzge et al. 2020) have provided profiler (e.g., winds, temperature, and moisture; Shrestha et al. 2021) measurements along coastal Long Island beginning in 2017, but publicly available data from offshore wind profiler sites within and adjacent to the NYB are limited. Temporary light detection and ranging (lidar) deployments by the New York State Energy Research and Development Authority (NYSERDA) provide measurements of wind speed and direction within Bureau of Ocean Energy Management (BOEM) wind energy areas (WEAs; OceanTech Services/DNV under contract to NYSERDA 2021; BOEM 2022; Fig. 1). However, limitations on these offshore lidar sites include length of deployment (<2 yr) and vertical scanning range (up to 200 m MSL). The NYB is a complex offshore area with various mesoscale circulations including the sea breeze and often accompanying low-level jet (LLJ) (Colle and Novak 2010; McCabe and Freedman 2023). The NYB jet maximum, or jet “nose,” varies in height between 150 and 300 m (McCabe and Freedman 2023); it thus occurs within the rotor plane of a typical offshore wind turbine [extending from 40 to 260 m or more above mean sea level (AMSL); Gaertner et al. 2020; IEA Wind Task 37 2020], making forecasting of the LLJ crucial to the accuracy of resource assessment and power production forecasts (Freedman et al. 2010). Thus, we need to rely on model data to provide a comprehensive picture of the NYB offshore wind environment.
Map of the NYB region and the data sites referenced in this study. NYSM surface, lidar, and radiometer sites (green) and NYSERDA (red) lidar buoys. Gray hatched areas indicate BOEM lease areas for offshore wind development (BOEM 2022).
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
The use of models, however, especially at the land–sea interface, can lead to significant spatiotemporal errors in forecasting or estimating wind power production. Mesoscale circulations such as the NYB sea breeze and associated LLJ, most common during the late spring and summer months, significantly increase afternoon and evening wind speeds along the coast and adjacent offshore waters (McCabe and Freedman 2023). However, models and reanalysis products often struggle to capture both the timing and strength of the sea breeze and the height, magnitude, and spatial extent of the LLJ (Nunalee and Basu 2014; Freedman et al. 2023; Fragano 2023). Reanalysis data are problematic at the scales of local circulations due to coarse spatial and vertical resolution (Draxl et al. 2015) and may be particularly biased over coastal regions due to the land–sea discontinuity (Gualtieri 2021). ERA5 reanalysis, for example, has limitations offshore, as it has been found to underestimate strong wind speeds (Gandoin and Garza 2024) in addition to a twice-daily discontinuity at the end of each data assimilation cycle observed in near-surface winds (specifically found in oceanic regions such as the NYB; Hersbach et al. 2020).
Previous studies have asserted that numerical weather prediction’s (NWP) poor ability to forecast the strength and structure of the LLJ may be related to model initialization, data assimilation, physical choices, and representation of sea surface temperature (Nunalee and Basu 2014; Colle et al. 2016; Mirocha et al. 2016; Aird et al. 2022). Aird et al. (2022) ran Weather Research and Forecasting (WRF) Model simulations covering the coastal regions of New Jersey, New York, and Massachusetts, finding that the 100 m MSL offshore wind speeds were less sensitive to boundary conditions, resolution of sea surface temperatures, number of vertical levels, or strength of spectral nudging, but strongly sensitive to the choice of planetary boundary layer (PBL) parameterization.
While there have been several studies documenting the performance of NWP models over coastal regions, very few have had the ability to perform model validation using a combination of offshore and coastal lidar networks (including the NYSERDA and NYSM profiler sites; Fig. 1). This gives us the unique opportunity to quantify model errors and understand the uncertainty in generating accurate wind energy forecasts.
This paper is organized as follows: Section 2 outlines the data used in the analysis. The model setup and methods used to run the sensitivity tests are described in section 3. Section 4 discusses the results of the WRF sensitivity analyses and discusses which model is the proper choice for the NYB region. Section 5 summarizes the findings and conclusions of our work.
2. Data
This study focuses on the region of the NYB and adjacent coastal regions, here defined as the area between 38.5°–42°N and 71°–75.5°W (Fig. 1). Model setup and observational and satellite data used in this study are detailed below. Model experiments use the WRF Model, version 4.4.1 (Skamarock et al. 2021). Analysis time is presented in UTC, which is 4 h ahead of local time [Eastern Daylight Time (EDT)] during the events discussed.
a. Observational datasets
1) The NYSM
The NYSM is a statewide meteorological network installed and operated by the University at Albany, State University of New York. The 126 standard stations were deployed during 2014–16, following World Meteorological Organization (WMO) siting guidelines and measuring surface variables such as temperature, wind speed and direction, relative humidity, global horizontal irradiance, precipitation, air pressure, and soil temperature and moisture (three levels: 5, 25, and 50 cm) at 3-s–5-min intervals (averaged to 10 min; Brotzge et al. 2020; WMO 2023). Additionally, a subnetwork of 17 profiler sites (deployed in 2017–18) includes both a Leosphere WindCube Doppler lidar (Vaisala, Inc.; measuring u, υ, and w up to 7 km) and a Radiometrics MP-3000 series microwave radiometer (estimating profiles of temperature, humidity, and vapor density at heights up to 10 km; Brotzge et al. 2020; Shrestha et al. 2021). Measurements are taken approximately every 4 s but are averaged into 10-min intervals.
The NYSM site at Wantagh is used for model verification in this study as it is positioned only 6 km inland (north) from Long Island’s southern coastline and frequently observes the NYB sea breeze and LLJ (McCabe and Freedman 2023; Fig. 1).
2) The NYSERDA
For a better understanding of the meteorological and oceanographic conditions in the NYB, NYSERDA contracted with Ocean Tech Services (OTS) to deploy two floating lidar buoys (EOLOS FLS200) in 2019 that measure wind profiles up to 200 m (20-m grid spacing; 10-min temporal resolution; OceanTech Services/DNV under contract to NYSERDA 2021). Other measurements include surface wind, air temperature, pressure, and water column temperature below the sea surface. These buoys were initially deployed at two locations: “Hudson North” (buoy E05; August 2019–September 2021) and “Hudson South” (buoy E06; September 2019–February 2022; see Fig. 1). Hudson South was decommissioned in March 2022. In January 2022, Hudson North was serviced and redeployed to the “Hudson Southwest” site and subsequently decommissioned in January 2023. Over the entire data collection period from each offshore buoy, the lidar’s data recovery is high: 91.05%, 73.32%, and 99.61% at Hudson North, Hudson South, and Hudson Southwest, respectively.
Due to the lack of data continuity across all case study events at one offshore location, we treat Hudson South (buoy E06) and Hudson Southwest (buoy E05) as the same site, given their proximity (∼15 km). We assume these sites have similar meteorological properties based on minimal differences in the daily standard deviation in the 140-m wind speeds. The 2-month (15 June–15 August) summertime 140-m wind speed average daily standard deviation for Hudson South, Hudson Southwest, and Hudson North is 2.36, 2.39, and 2.22 m s−1, respectively.
b. Satellite data
1) OSTIA
Operational Sea Surface Temperature and Ice Analysis (OSTIA) is a product of the Met Office and provides high-resolution analysis of SST and ice cover using a combination of observations and in situ platforms (Stark et al. 2007; Donlon et al. 2012; Good 2018). The SST output is available daily on a 0.05° grid and is bias corrected. OSTIA runs daily at 0600 UTC and is based on a 36-h rolling observation window, centered on 1200 UTC of the previous day (Stark et al. 2007). One of the benefits of this dataset is that it does not contain any missing data (Redfern et al. 2023). It is used operationally as a boundary condition in ECMWF models (Donlon et al. 2012; ECMWF 2019) and in NREL’s 2023 National Offshore Wind dataset (NOW-23; Bodini et al. 2024).
c. Operational models
1) HRRR
The Rapid Refresh (RAP) is a regional model produced by NOAA’s National Centers for Environmental Prediction (NCEP; Benjamin et al. 2016; Dowell et al. 2022). The High-Resolution Rapid Refresh (HRRR) is nested within the RAP and is focused over the contiguous United States at a 3-km resolution (Benjamin et al. 2016). The HRRR analysis is used in this study for initial and boundary conditions in the WRF Model.
3. Methods
With a focus on the NYB sea breeze and associated LLJ, we performed several sensitivity studies to determine the optimal physical choices and initialization conditions that result in the best overall model performance. Model performance metrics are based on the timing and magnitude of wind speed maximum, placement of LLJ, and shape of wind speed profile. This study is unique, as the profiler data (lidar and microwave radiometer) available through the NYSM and NYSERDA sites will be used for the comparison of the measured and modeled wind, thermodynamic, and stability profiles, allowing for a more advanced analysis of WRF Model performance under sea-breeze and LLJ conditions. Given that the NYB sea breeze and LLJ occur on a small vertical scale in the boundary layer (typically up to ∼300 MSL; see, e.g., McCabe and Freedman 2023), it is hypothesized that the choice of PBL parameterization and land surface model (LSM) is determinative in assessing model performance.
As set forth in McCabe and Freedman (2023), there are two types of sea breezes that occur in the NYB: “classic” (a sea breeze driven primarily by both cross-shore pressure and temperature gradients, with light background winds) and “hybrid” (sea breeze occurs in combination with other larger-scale features, such as frontal systems). Using a high-resolution version of WRF (1.33-km horizontal grid and 28 vertical levels below 1 km; Skamarock et al. 2021, and see Table 1), the goal is to determine the reliability of forecasting winds associated with both a classic and hybrid sea-breeze event in the NYB and the model’s ability to reproduce an associated LLJ. In this study, four sea-breeze events (all with an associated LLJ) are considered. These events include 9 June 2020 (classic), 28 June 2021 (classic), 24 July 2022 (hybrid), and 4 August 2022 (classic).
Control setup for WRF experiments. The parameterizations to be tested include the PBL scheme, surface layer physics, land surface physics, and urban surface physics.
a. WRF Model sensitivity studies’ configuration
We use two domains with a two-way nested approach. The analysis is focused on the smallest domain (1.33-km resolution) which covers the offshore region of the NYB including the adjacent coastal regions of New Jersey, New York, and Connecticut (Fig. 2).
WRF (version 4.4.1) domain, including the outer grid (d01) and the inner grid (d02). This analysis focuses on the inner grid (d02).
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
The physical parameterizations tested include PBL, LSM (Noah and Noah-MP), and an urban surface physical scheme [building environment parameterization including the building energy model (BEP + BEM); Table 1; Zonato et al. 2021a,b; Xu et al. 2022]. The urban surface physical scheme is included in this study to consider the interaction between the New York City urban heat island and the NYB sea breeze (Gedzelman et al. 2003; Han et al. 2022). Given the importance of cross-shore temperature and pressure gradients in driving the sea breeze, it is important to evaluate how incorporating the additional urban physical scheme into the model may change performance (see later discussion in section 4b and Fig. 14).
The most common choices used for PBL schemes among studies of similar mesoscale phenomena are the Mellor–Yamada–Nakanishi–Niino Level 2.5 (MYNN2; Nakanishi and Niino 2006; Aird et al. 2022; Bodini et al. 2024), Yonsei University (YSU; Hong et al. 2006), and Mellor–Yamada–Janjić (MYJ; Mesinger 1993; Janjić 1994; Colle et al. 2016; Mirocha et al. 2016; Hughes and Veron 2018; Strobach et al. 2018; Zhang et al. 2021; Hu et al. 2022; Wermter et al. 2022; Xie et al. 2023). In a more extensive sensitivity study of PBL schemes in the NYB, Nunalee and Basu (2014) found that although all PBL schemes struggled to properly model the LLJ at coastal locations, the MYJ (Janjić 1994), Asymmetric Convection Model 2 (ACM2) scheme (Pleim 2007a,b), and quasinormal scale elimination (QNSE; Sukoriansky et al. 2005) performed best. Recently, the 2023 National Offshore Wind dataset (Bodini et al. 2024) chose the MYNN2 PBL and surface schemes for the mid-Atlantic region. Additionally, there are two PBL schemes that were designed for marine boundary layers, the Grenier–Bretherton–McCaa (GBM) and the University of Washington (UW; Park and Bretherton 2009), that have yet to be tested over this NYB region. Given the variation of PBL schemes throughout the literature, this study tests all seven for the four referenced case studies (YSU, MYJ, MYNN2, ACM2, QNSE, GBM, and UW; see Table 1).
The seven PBL parameterizations provide a variety of diagnostic (nonlocal) and prognostic (local) turbulence closure schemes. Local schemes only calculate processes using adjacent grid volumes, while nonlocal mixing schemes involve many grid volumes (better for modeling deep convection and deep flows and generally developing a better adiabatic mixed layer; Cohen et al. 2015). The YSU scheme is a nonlocal, K-theory scheme, while the MYJ, MYNN2, QNSE, GBM, and UW schemes are all turbulent kinetic energy (TKE) schemes with 1.5-order local closure. The ACM2 scheme uses nonlocal closure during unstable conditions and local eddy diffusivity under stable conditions (Pleim 2007a; Siuta et al. 2017). Therefore, the ACM2 is unique as it can represent turbulent transport at both the subgrid and supergrid scales (Pleim 2007b). To parameterize the surface layer, the ACM2, GBM, UW, and YSU use the Revised MM5 surface layer scheme, which parameterizes ocean surface properties using the COARE 3 formula (Fairall et al. 2003; Jiménez et al. 2012). The MYJ, MYNN2, and QNSE use the Monin–Obukhov similarity theory (Olson et al. 2019; Monin and Obukhov 1954; Janjić 1994, 1996, 2001). Although the MYJ and MYNN2 are both local schemes using 1.5-order closure, the MYNN2 uses a different parameterization for the mixing length scale; therefore, it can better represent mixing in stable conditions (Milovac et al. 2016; Mirocha et al. 2016).
The model runs are 30 h, initialized at 0000 UTC on the day of the desired event (allowing 6 h for model spinup) and run through 0600 UTC the following day [sea breeze onset along Long Island is generally seen around 1800 UTC (McCabe and Freedman 2023)].
b. Sea surface temperature initialization evaluation
The air–sea temperature gradient has been found to be an important contributing factor in the sea-breeze and LLJ development (Bowers 2004; Seroka et al. 2018; McCabe and Freedman 2023). To determine which SST product to use in our sensitivity analysis, we tested two different SST initializations: the HRRR [initialized using the real-time global (RTG) NCEP analysis, updated to an SST analysis using the Global Data Assimilation System (GDAS) in 2020 (Dowell et al. 2022)] compared to the OSTIA (Stark et al. 2007). The WRF initialization using the OSTIA dataset outperformed the initialization using the HRRR (Fig. 3a; Benjamin et al. 2016; Blaylock et al. 2017; Dowell et al. 2022). Across all four events, the HRRR analysis produces a cold bias in offshore SSTs but shows a warm bias near the coasts (Fig. 3). We know from previous studies that the coastlines along the New Jersey shore can feature colder water due to coastal upwelling, making the sensitivity of the satellite dataset to the localized areas of colder water crucial to accurate forecasts of the sea breeze and LLJ (Seroka et al. 2018).
(a) The 30-h model run time series comparing SST observations (black; °C) at offshore site Hudson South (9 Jun 2020 and 28 Jun 2021) and Hudson Southwest (24 Jul 2022 and 4 Aug 2022) to modeled SST (°C) when initialized with the OSTIA (blue) and the HRRR (orange). (b)–(e) The time-series location corresponds to the white circle on the contour plots. Contour plots represent the SST difference (°C) between the OSTIA and the HRRR at model initialization during the four case studies: (b) 9 Jun 2020, (c) 28 Jun 2021, (d) 24 Jul 2022, and (e) 4 Aug 2022. Gray shaded areas represent areas for offshore wind energy development (BOEM 2022).
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Although the HRRR (3 km) has higher horizontal grid spacing than OSTIA (6 km), the OSTIA produced a lower bias in SST predictions (Figs. 3d,e). The OSTIA is produced once daily; therefore, the sea surface temperature in the model is initialized at 0000 UTC on the day of the sea-breeze event and held constant for the duration of the 30-h model run. This is a reasonable assumption given the length of the model runs and the small amplitude (0.3°–0.7°C) of the SST diurnal cycle as shown in the observations (Fig. 3a).
4. Results and discussion
A total of 18 sensitivity experiments are run for each case study (Table 2). The naming convention of each test is based on the PBL scheme, urban parameterization, and land surface model (PBL_urban_LSM). To evaluate model performance in forecasting the sea breeze and LLJ, we consider time series and vertical profiles of wind speed, rotor equivalent wind speed [REWS; see Eq. (3)], wind direction, temperature, virtual potential temperature, specific humidity, and dewpoint. For model validation, WRF output is paired with observations using the WRF grid point collocated with the latitude and longitude of the observations. The grid cell used for the coastal site, Wantagh, has been verified to be a land point in the WRF Model and is classified as land-use category 12 (“croplands”).
Summary of 18 total WRF sensitivity experiments.
a. Planetary boundary layer schemes
Based on previous findings, we expect the PBL scheme choice to play the largest role in improving model performance (Nunalee and Basu 2014; Colle et al. 2016; Aird et al. 2022). Model validation is done using an NYSM coastal site (Wantagh) and the NYSERDA offshore buoy-based lidar locations: Hudson North, Hudson South, and Hudson Southwest (Fig. 1). Validation at the Hudson North location is limited in this study due to data availability constraints. Data from the Hudson North location is only available through September 2021, and therefore, the bulk of the analysis focuses on the Hudson South and Hudson Southwest locations. The coastal site, Wantagh, provides insight into understanding model performance at the land–sea interface. Wantagh’s location on the south shore of western Long Island makes it an ideal site for LLJ identification, as it is proximate to where the core of the LLJ moves onshore (Fig. 1; McCabe and Freedman 2023). As mentioned previously, all sea-breeze events discussed feature an accompanying LLJ.
When modeling the wind profiles at Wantagh, all model configurations, regardless of the PBL scheme, have difficulty capturing the magnitude and height of the LLJ maximum (Figs. 4a–d). Generally, the models overpredict the height of the LLJ maximum while underpredicting the maximum wind speed, as seen during the early season events (9 June 2020 and 28 June 2021). This can be directly related to WRF’s difficulty in modeling the stable layer observed near the surface during these sea-breeze events. The virtual potential temperature profiles at Wantagh, measured from the microwave radiometer (note that the microwave radiometer has a cold bias near the surface and tends to smooth gradients in the boundary layer thermodynamic profile, so the 2-m virtual potential temperature is also shown; Bianco et al. 2017; Djalalova et al. 2022), show stable conditions on 9 June 2020, 28 June 2021, and 24 July 2022 (see further discussion in section 4b and Fig. 12). Meanwhile, each PBL scheme configuration depicts a mixed layer up to 200–300 m AGL. This emphasizes the difficulties that the WRF Model has in adjusting at the land–sea discontinuity, as all model configurations tend to overpredict mixing onshore. This leads to a modification of the jet at the coastline and a redistribution of the wind profile (weakening the jet in the mixed layer while accelerating the jet at the top of the mixed layer; Fig. 12). This is substantiated by what is seen on 4 August 2022, as the observations indicate a weaker LLJ and mixed layer at Wantagh, meaning weaker forcing of the sea breeze and better agreement between the model configurations and observations (Figs. 4d and 12j,k,l). As a result, the height of the LLJ wind speed maximum is generally seen to occur above the observations (Fig. 4).
Scatterplots showing model error in the time (x axis; min), height (y axis; m), and velocity (color bar; m s−1) of the LLJ maximum. Different PBL schemes (ACM2, QNSE, YSU, MYNN2, MYJ, UW, and GBM) are represented through varying symbols. The cross at zero represents the observations at Wantagh, and the time represents the lidar-measured time of the LLJ maximum. Shown for (a) 9 Jun 2020, (b) 28 Jun 2021, (c) 24 Jul 2022, and (d) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Although all WRF Model physical configurations tested perform better in capturing the offshore LLJ, forecasted variations in the strength, height, and timing of maximum wind speeds can have a large effect on shear across the rotor plane and alter wind energy forecasts (Fig. 5). No model configuration is able to correctly depict the time, strength, and height of the LLJ maximum associated with the sea breeze, typically predicting the jet to reach its maximum too early (Figs. 5a,c,d). These timing errors related to the sea breeze and LLJ at the offshore and coastal locations are observed throughout the duration of the sea-breeze event (Figs. 4, 5, and 7). On 9 June 2020, the modeled sea-breeze onset lags the observed by nearly 2 h at Hudson North and Hudson South yet accelerates more quickly in the model predictions and reaches a wind speed maximum earlier when compared to the observations (Figs. 5a and 7a,e). At Wantagh, models lag the observations of the peak wind speed by nearly 2 h (Fig. 4a).
As in Fig. 4, but for Hudson South on (a) 9 Jun 2020 and (b) 28 Jun 2021 and Hudson Southwest on (c) 24 Jul 2022 and (d) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Changing the PBL scheme alters the strength and extent of the strongest winds, especially apparent offshore (Fig. 6). For example, the MYNN2 leads to higher 150-m wind speeds across much of NYB when compared to the ACM2 (Figs. 6c,e). Although wind speed magnitude varies, all seven PBL schemes show the southern extent of the sea breeze on 9 June 2020 extends to just south of the Hudson South buoy location (Fig. 6). During this event, the eastward extent of the sea breeze diminishes over and just west of the lease areas south of New England (Fig. 6). Knowing the offshore horizontal extent of the sea breeze and LLJ in the NYB is crucial for wind energy purposes, as wind speed under sea-breeze and LLJ conditions can vary significantly from one lease area to another. The northward extent of the sea breeze, however, cannot be determined given the difficulties of all PBL schemes in penetrating the coastline of Long Island. The 150-m winds quickly decelerate as they reach the south shore of Long Island (Fig. 6). Offshore, a stable boundary layer allows for the development and propagation of the LLJ, but the modeled instability at the coastline mixes out higher wind speeds at low levels, forcing the jet height to increase (and winds diminish) and contributing to the poor model forecasts of the LLJ at Wantagh (Figs. 4a–d and 12, and continued discussion below; Doyle and Warner 1991; Colle and Novak 2010).
2100 UTC plots of modeled 150-m wind speed (contours; m s−1) and wind direction (wind barbs; 1 full barb = 5 m s−1) shown for seven different model runs: (a) YSU_nUrb_Noah, (b) MYJ_nUrb_Noah, (c) MYNN2_nUrb_Noah, (d) QNSE_nUrb_Noah, (e) ACM2_nUrb_Noah, (f) UW_nUrb_Noah, and (g) GBM_nUrb_Noah. Small white markers represent the location of the offshore NYSERDA sites, Hudson South and Hudson North, and the coastal NYSM site, Wantagh.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
In addition to evaluating model performance based on wind speed, we also consider capacity factor. Capacity factor is defined as the ratio of gross energy generation produced over a period of time relative to the energy that could have been produced at continuous full-power operation over the same time period [based on a generic 15 megawatt (MW) wind turbine; Gaertner et al. 2020; International Energy Agency 2024]. Due to the errors in the timing of sea-breeze onset, rate of wind speed increase, and wind speed magnitude, we see the largest errors in capacity factor forecasts during the hours of 1200–1800 UTC, just prior to and during sea-breeze onset (Figs. 7d,h). This is problematic for wind energy forecasts, as turbines are generally still operating below their rated power and small variations (or speed ramps) can cause large deviations in power predictions. From a wind energy perspective, knowing the timing and magnitude of these intermediate wind speeds [those between cut-in (3 m s−1) and turbine-rated power speeds (10.59 m s−1) for a 15 MW turbine] is an important driver in seeking NWP improvement (Gaertner et al. 2020).
Time series comparing observations (black) to the seven model runs for the 30-h period from 0000 UTC 9 Jun 2020 to 0600 UTC 10 Jun 2020 at (a)–(d) Hudson South and (e)–(h) Hudson North. Figure panels show (a),(e) wind speed (m s−1) and (b),(f) direction (1 barb = 5 m s−1) at 140 m MSL, (c),(g) 2-m air temperature (°C), and (d),(h) 140-m capacity factor (%) based on a 15-MW wind turbine.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Given this uncertainty in the modeled wind speeds during the hours surrounding sea-breeze onset, Taylor diagrams are used to evaluate model accuracy in forecasting the wind speed closest to hub height (based on a 140 m MSL measurement height) at the offshore Hudson South/Hudson Southwest location for a 12-h period from 1200 to 2359 UTC on each sea-breeze event (Fig. 8; Taylor 2001). Given the 18 different model configurations tested across four different events, the MYJ_nUrb_NoahMP configuration consistently performs well during the 12-h period at the offshore location. The GBM_nUrb_Noah, ACM2_nUrb_Noah, YSU_nUrb_Noah, and YSU_Urb_Noah also do well forecasting most of the events (Fig. 8).
Taylor diagrams of 140-m wind speed forecasts at (a),(b) Hudson South and (c),(d) Hudson Southwest for all 18 WRF sensitivity runs. The radial axis shows the standard deviation, the angular axis is the coefficient of determination r2, and the radial arcs represent the centered RMSE difference between the model and the observation (Taylor 2001). Statistics are measured over a 12-h period, from 1200 to 2359 UTC on four sea-breeze events: (a) 9 Jun 2020, (b) 28 Jun 2021, (c) 24 Jul 2022, and (d) 4 Aug 2022. The legend on the upper right indicates which symbols and colors are associated with each test run. The observation is indicated on each plot by a black star. The best forecasts are located closest to the black dashed line (standard deviation of observation) and closest to the star.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
For each sensitivity test, the normalized bias of the REWS is calculated at 30-min intervals over the entire WRF simulation. Since the bias is normalized as a percent of the wind speed magnitude, it is common to see higher biases during periods of higher wind speeds. However, using the normalized bias allows us to more easily compare results site to site.
At Hudson South (9 June 2020 and 28 June 2020; Fig. A3) and Hudson Southwest (24 July 2022 and 4 August 2022; Fig. A4), the MYJ PBL scheme performs well across all four case studies, especially during the midday to evening period (1200–2359 UTC; sea breeze onset time) when errors (and wind speeds) are generally highest (Fig. 8). The MYNN2_nUrb_Noah scheme also does well during both June cases but shows higher biases during the late summer events (Figs. A3 and A4). Note that the MYNN2 PBL scheme was chosen for the mid-Atlantic region in NREL’s recent NOW-23 (Bodini et al. 2024).
The same metrics are shown for the model performance of REWS prediction at the NYSM Wantagh coastal site (Figs. A5 and A6). At Wantagh, the YSU, UW, and MYNN2 PBL schemes perform worse when compared to the MYJ, ACM2, QNSE, and GBM (Fig. 9). To determine which schemes do best under sea-breeze and LLJ conditions, we average the metrics across the four events at the two sites, Wantagh and Hudson South/Hudson Southwest (Figs. 9 and 10). As discussed in section 2, for analysis purposes, we average Hudson South and Hudson Southwest, treating them as the same site (Figs. 9 and 10b). To determine which model is statistically best, we use the bootstrapping technique to create 95% confidence intervals of MAE, RMSE, CRMSE, r2, and normalized bias. The overlap of the confidence intervals helps to determine the statistical significance of these results. If the 95% confidence intervals do not overlap (or have minimal overlap), we know that one test is statistically better. At Wantagh, the improvement in model performance across the varying schemes is clear (Fig. 9). For example, there is little to no overlap when comparing the MYNN2-based schemes to the MYJ-based schemes, indicating the clear improvement in model performance near the coast when using the MYJ compared to the MYNN2. As we have discussed throughout this analysis, overall, model performance is better at offshore locations and statistical differences remain small. However, as discussed, even small differences in model predictions of wind speed can have significant impacts from a wind energy perspective, and the results clearly show that the MYJ- and ACM2-based sensitivity studies continue to demonstrate better model performance (Fig. 9).
MAE, RMSE, CRMSE, R2, and normalized bias with 95% confidence intervals (vertical bars) shown for Wantagh (orange) and Hudson South/Southwest (blue). Data are averaged from the 24-h time period from 0600 to 0600 UTC. The dot represents the mean value, and the interval boundaries represent the range from 10 000 bootstrapping resamples.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Twenty-four-hour statistics of REWS, including MAE, RMSE, CRMSE, and r2 averaged across the four sea-breeze cases, 9 Jun 2020, 28 Jun 2021, 24 Jul 2022, and 4 Aug 2022, for (a) Wantagh and (b) Hudson South/Southwest.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
As model performance varies between the offshore and coastal sites, the top three model configurations from each location are chosen, providing five runs we consider in more detail (Fig. 10). These include the MYJ_Urb_Noah, MYJ_Urb_NoahMP, ACM2_nUrb_Noah, MYJ_nUrb_NoahMP, and MYJ_nUrb_Noah. The MYJ scheme performs best at both the coastal and offshore locations. Therefore, we next consider how the use of the urban parameterization (on/off) and choice of LSM (Noah or Noah-MP) change model performance.
Note that while model performance ranking based on MAE, RMSE, and CRMSE is consistent, the r2 values are not (Fig. 11). To better understand this dissimilarity, we consider the r2 for three cases at Wantagh, the MYJ_Urb_Noah [ranked first, high RMSE, MAE, CRMSE, and r2 value (0.823)], the YSU_Urb_Noah [ranked 16th, a low MAE value, and low r2 (0.596)], and the UW_nUrb_NoahMP [ranked 18th, featuring a low MAE but high r2 (0.802); Fig. 10a]. As shown, there are limitations in using r2 alone to determine model accuracy. The correlation of determination r2 looks to fit the best linear relationship; however, it may not always be the best statistic to determine the overall agreement between the model and observations, as demonstrated when comparing the REWS of UW_nUrb_NoahMP and MYJ_Urb_Noah (Fig. 11; Janse et al. 2021).
Time series of REWS at Wantagh measured from the observations (black) and three model runs; YSU_Urb_Noah (blue), UW_nUrb_NoahMP (orange), and MYJ_Urb_Noah (green), for (a) 0600 UTC 9 Jun 2020–0600 UTC 10 Jun 2020, (b) 0600 UTC 28 Jun 2021–0600 UTC 29 Jun 2021, (c) 0600 UTC 24 Jul 2022–0600 UTC 25 Jul 2022, and (d) 0600 UTC 4 Aug 2022–0600 UTC 5 Aug 2022. The r2 value is calculated and shown (upper left) between observations and each model over the 24-h period.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
b. Land surface models and urban surface physics
Although the PBL scheme generally has the largest influence on model performance, other physical parameterizations such as the land surface scheme (LSM) and urban parameterization are considered (Nunalee and Basu 2014; Colle et al. 2016; Aird et al. 2022). By using the five model setups found to statistically best represent the sea breeze and LLJ in the NYB (Figs. 9 and 10), we evaluate model performance using two different LSMs, Noah and Noah-MP, and urban surface physical parameterization using the building energy parameterization which includes BEP + BEM and the MYJ compared to the ACM2 PBL scheme. Model performance is evaluated based on the ability to forecast important driving mechanisms in the sea breeze and LLJ, including timing, magnitude, and height of the wind speed maximum, land–sea temperature gradient, baroclinicity of the offshore environment, and overall boundary layer stability.
We again emphasize that our ability to validate model performance using observations is limited by data availability in the offshore environment. While the NYSM network allows us to compare vertical profiles of observed virtual potential temperature and humidity to model output, those measurements are not available offshore. At the land–sea interface, the model has difficulties thermodynamically and the LLJ is not as well defined. The observations at Wantagh show that with a strong sea breeze the surface layer remains stable at Wantagh (Figs. 12a–i). Meanwhile, the model predictions are more consistent with weaker forcing as they allow for mixing at the coastline, confirmed by the observations of 4 August 2022 (Fig. 12).
Wantagh vertical profiles between observations (black) and top five model runs [MYJ_Urb_Noah (blue), MYJ_Urb_NoahMP (light orange), MYJ_nUrb_Noah (green), MYJ_nUrb_NoahMP (dark orange)], and ACM2_nUrb_Noah (pink)]; surface measurements at Wantagh are shown using a black star. (left) 2100 UTC vertical profiles of wind speed (m s−1), (center) virtual potential temperature (K), and (right) specific humidity (g kg−1) are shown for (a)–(c) 9 Jun 2020, (d),(e) 28 Jun 2021, (g)–(i) 24 Jul 2022, and (j)–(l) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Although all configurations have errors, when modeling the virtual potential temperature and the wind speed at Wantagh, the MYJ_nUrb_Noah overall performs best (Fig. 12). It places the inversion a bit lower, indicating a shallower mixed layer, allowing the LLJ to form at lower heights and therefore closer to observations. On 4 August 2022, it models the jet almost perfectly; however, forecasted wind speeds are too strong on 28 June 2021 and 24 July 2022, and the jet nose is too high on 9 June 2020 (Figs. 12a,d,g,j).
Offshore, only observations of the vertical profiles of wind speed are available, but given the more homogeneous offshore environment and distance from the land–sea interface, model performance is better (Fig. 13). The virtual potential temperature profiles from the models show a stable layer in which the LLJ maximum forms directly above (Fig. 13). We assume that the model simulations show a better representation of the stability of the offshore environment, due to the improvement in modeling the wind speed profile when compared to Wantagh.
As in Fig. 12, but for offshore site Hudson South (9 Jun 2020 and 28 Jun 2021) and Hudson Southwest (24 Jul 2022 and 4 Aug 2022). Observations of virtual potential temperature and specific humidity are not available at this site.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Regardless of the sea-breeze case, the MYJ_Urb_NoahMP forecasts the strongest LLJ (Fig. 13). This is consistent with the bias in the air temperature gradient measured between the onshore and offshore locations (Fig. 14). Prior to sea-breeze onset, the MYJ_Urb_NoahMP consistently models the strongest 2-m air temperature gradient and sea level pressure gradient (measured between the onshore Bronx and offshore sites Hudson South and Hudson Southwest; ∼152 and ∼156 km apart, respectively). These stronger cross-shore gradients directly correlate with a higher 140-m wind speed (and stronger LLJ) at the offshore location (Figs. 15 and 16). We see the opposite relationship when considering the MYJ_nUrb_NoahMP, as the model run indicates a smaller 2-m air temperature and pressure gradient (the urban scheme “off” lowers NYC regional 2-m air temperatures onshore) and correlates with the weakest sea breeze (Fig. 14). As shown, the Noah-MP LSM paired with the urban physical scheme is too extreme, as urban physics “on” overheats the NYC regional area and inflates the regional temperature gradient, while urban physics off weakens the gradients.
(top) Time series of 2-m air–sea temperature difference (solid lines) and sea level pressure gradient (right y axis; dashed lines) measured between (a),(c) Bronx and Hudson South and (e),(g) Hudson Southwest compared to the time series of (bottom) 140-m wind speed offshore at (b),(d) Hudson South and (f),(h) Hudson Southwest. Time series is shown from 1200 to 0000 UTC on (a) and (b) 9 Jun 2020, (c) and (d) 28 Jun 2020, (e) and (f) 24 Jul 2022, and (g) and (h) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
(top) Time series of 150-m wind speed (m s−1), (middle) 2-m air temperature (°C), and (bottom) dewpoint temperature (°C) at Wantagh, comparing observations (black) to the top five model runs [MYJ_Urb_Noah (blue), MYJ_Urb_NoahMP (light orange), MYJ_nUrb_Noah (green), MYJ_nUrb_NoahMP (dark orange), and ACM2_nUrb_Noah (pink)]: (a) 1200 UTC 9 Jun 2020–0200 UTC 10 Jun 2020, (b) 1200 UTC 28 Jun 2021–0200 UTC 29 Jun 2021, (c) 1200 UTC 24 Jul 2022–0200 UTC 25 Jul 2022, and (d) 1200 UTC 4 Aug 2022–0200 UTC 5 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Twenty-four-hour time–height cross sections of wind speed (m s−1; contours) from (a)–(c) 0600 UTC 9 Jun 2020 to 0600 UTC 10 Jun 2020 at Hudson South, (d)–(f) 0600 UTC 28 Jun 2021 to 0600 UTC 29 Jun 2021 at Hudson South, (g)–(i) 0600 UTC 24 Jul 2022 to 0600 UTC 25 Jul 2022 at Hudson Southwest, and (j)–(l) 0600 UTC 4 Aug 2022 to 0600 UTC 5 Aug 2022. Time–height cross sections are shown for (a),(d),(g),(j) lidar observations and (b),(c),(h),(k) WRF Model output from the MYJ_nUrb_Noah sensitivity test. The wind speed difference (observation − WRF Model forecast) is shown in (c), (f), (i), and (l). (bottom) The black dashed line at 140 m indicates the height of the wind speed difference time series in (c), (f), (i), and (l).
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Interestingly, when pairing the BEP + BEM urban physical scheme with the Noah LSM, we see little impact on the temperature gradient or wind speed at the offshore locations (Fig. 14). At Wantagh, we see that the sensitivity run without urban physics produces higher wind speeds than with the urban physics on (Fig. 15). When pairing the BEP + BEM urban physical scheme with Noah, we see a larger change in the profile of the virtual potential temperature (Figs. 12b,e,h,k). The urban scheme decreases the stability at the coastline, resulting in a deeper mixed layer and therefore mixing out higher wind speeds. This makes it more difficult for the model to capture the inland propagation of the LLJ, and as a result, 150-m wind speeds are reduced at Wantagh (Figs. 12a,d,g,j, and 15).
Although the Noah-MP LSM (Niu et al. 2011) improves upon the Noah LSM (Chen and Dudhia 2001; Ek et al. 2003), it may not be the best choice for the NYB region. The Noah-MP LSM was designed to improve the model representation of vegetation and canopy, radiative transfer processes, snowpack, and both soil and hydrological processes (He et al. 2023). This study, however, is considering a subset of sea-breeze events over a region characterized by offshore, coastal, and urban land surfaces, and factors such as vegetation, canopy, snowpack, and soil hydrology will have less influence on model performance.
We see an increased negative bias in the 150-m wind speed predictions and an increased positive bias in 2-m air temperature at Wantagh when using the Noah-MP LSM (Fig. 15). The time series at Wantagh shows us that the best model performance excludes the urban scheme and employs the Noah LSM (the MYJ_nUrb_Noah and ACM2_nUrb_Noah; Fig. 15). The MYJ_nUrb_Noah overall provides the best forecast for the coastal LLJ, 150-m winds, and variations in 2-m air temperature and dewpoint (Figs. 12a,d,g,j, and 15). It also performs well offshore, as it accurately represents the maximum afternoon wind speeds and captures the LLJ magnitude and height on most events (Figs. 13 and 14). However, timing errors are still present, as the model still has difficulties in forecasting the exact time of sea-breeze onset (Figs. 14–16).
The MYJ_nUrb_Noah offshore errors in wind speed prediction are consistent from case to case. The time–height cross sections of the wind speed at Hudson South/Southwest show an underprediction of the wind speed prior to sea-breeze onset, but an overprediction shortly thereafter (Fig. 16), related to event transition timing errors. The models do perform well in matching the pattern of the winds during the entire event, featuring light, low-level winds in the morning, transitioning to stronger, and higher (presence of an LLJ) toward the afternoon and evening (Fig. 16). The largest errors, as we have shown throughout this analysis, occur during the transition from light to strong winds accompanying sea-breeze onset. The peak in the 140-m wind speed error occurs between 1500 and 1800 UTC (Figs. 16c,f,i,l).
5. Findings and conclusions
We analyzed model performance across four sea-breeze events, all featuring an associated LLJ. The events featured examples of both classic and hybrid sea breezes; however, model performance was found to be consistent regardless of the sea-breeze type. For each case study, we conducted 18 WRF sensitivity experiments, which tested different combinations of PBL schemes, including YSU, MYJ, MYNN2, ACM2, QNSE, UW, and GBM, LSMs (Noah and Noah-MP), and urban surface physical parameterization (BEP + BEM). The MYJ PBL scheme best simulates wind fields associated with the sea breeze and LLJ both offshore and at a coastal site. The MYJ PBL scheme, initially designed for its skill in forecasting severe summertime convective storms, is a TKE closure scheme that, after extensive research to improve the forecast at the ocean–air interface, includes a marine viscous sublayer at the surface (Janjić 1994). The MYJ is a good choice in parameterizing the NYB because it shows better performance in forecasting the characteristic warm season sea breeze and LLJ.
We found that the urban scheme did not add value to the model forecasts. Despite the proximity of the NYB and adjacent coastal regions to New York City and surrounding urban areas, under summertime sea-breeze conditions, the urban scheme, when paired with Noah-MP, leads to an artificial enhancement of the air–sea temperature gradient, therefore overestimating the near-hub height wind speeds and the LLJ. When BEP + BEM was paired with Noah, we saw more unstable conditions, which resulted in increased mixing, therefore decreasing wind speeds at the coast and ultimately degrading model performance. However, regardless of the urban physical setting, Noah outperforms Noah-MP.
These findings emphasize the importance of fine-tuning NWP models to a specific region, as minor changes in model physics can have a significant effect on overall model performance. Accurate sea-breeze forecasts are complicated in coastal locations and are often region specific, so sea-breeze forecast techniques are not always transferable (Miller et al. 2003). While this study is focused on wind energy applications, model forecasts of the sea breeze and LLJ can be of value in other applications. For example, John F. Kennedy (JFK) International Airport is located just 20 km west of Wantagh and situated on Long Island’s south shore. Proper forecasts of the sea breeze and LLJ play a critical role in planning JFK operations, as a shift in wind direction or increase in wind speed, especially those associated with the sea-breeze front, can introduce additional shear and lead to rerouting of arriving or departing aircrafts [National Research Council (NRC) 1983; Chan and Hon 2023]. Better forecasts are also important from a power demand perspective, as the sea breeze cools air temperatures onshore, easing energy loads in coastal locations.
With the development of offshore wind facilities in the NYB, improving model performance and understanding factors such as magnitude, location, height, and timing are crucial to making accurate wind energy forecasts. Especially important from a wind energy perspective is that the hours just prior to and during sea-breeze onset are where errors are largest, but also when the greatest changes in wind speed are observed. While all models have inherent biases, it is important to understand their limitations and how they can best be used to improve forecasts of offshore wind power production.
Acknowledgments.
This research is made possible by the New York State (NYS) Mesonet and National Oceanic and Atmospheric Administration’s New York State Mesonet Profiler Network: Empire State Vertical Sensing Evaluation Regional Testbed Experiments (VERTEX 1 and VERTEX 2), Awards NA22NWS4690023 and NA23NWS4690025. Original funding for the NYS Mesonet (NYSM) buildup was provided by the Federal Emergency Management Agency Grant FEMA-4085-DR-NY. The continued operation and maintenance of the NYSM are supported by the National Mesonet Program, the University at Albany, federal and private grants, and others.
Data availability statement.
The buoy data are available upon request from NYSERDA. Neither NYSERDA nor OceanTech Services/DNV have reviewed the information contained herein and the opinions in this report do not necessarily reflect those of any of these parties. The HRRR data archive is accessed through the University of Utah page, developed by Brian Blaylock (https://home.chpc.utah.edu/∼u0553130/Brian_Blaylock/cgi-bin/hrrr_download.cgi?model=hrrr&field=prs&date=2020-06-09&link2=grib2; Blaylock et al. 2017). The metadata for the High-Resolution Rapid Refresh Model Archives are located at https://doi.org/10.7278/S5JQ0Z5B. ERA5 reanalysis data are downloaded at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-pressure-levels?tab=overview, and OSTIA satellite data are downloaded from https://data.marine.copernicus.eu/product/SST_GLO_SST_L4_NRT_OBSERVATIONS_010_001/description. The New York State Mesonet (NYSM; copyrighted by RFSUNY) data can be requested at http://www.nysmesonet.org/about/data and are cited in Brotzge et al. (2020).
APPENDIX
Rotor Equivalent Wind Speed Statistics
Figures A1–A6 show summarized observations and model output of REWS and corresponding statistics including normalized bias, MAE, RMSE, CRMSE, and r2.
Wantagh REWS (m s−1) from observations and all model sensitivity runs at six time intervals for (a) 9 Jun 2020, (b) 28 Jun 2021, (c) 24 Jul 2022, and (d) 4 Aug.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
As in Fig. A1, but for Hudson South on (a) 9 Jun 2020 and (b) 28 Jun 2021 and Hudson Southwest on (c) 24 Jul 2022 and (d) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
Grid shows the REWS normalized bias at 0000 UTC (initialization), 1200 UTC, 1500 UTC, 1800 UTC, 2100 UTC, and 0000 UTC, as well as the model run statistics (calculated over a 24 h period from 0600 to 0600 UTC) MAE, RMSE, CRMSE, and r2 for sea-breeze cases at Hudson South for (a) 9 Jun 2020 and (b) 28 Jun 2021.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
As in Fig. A3, but for Hudson Southwest on (a) 24 Jul 2022 and (b) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
As in Fig. A3, but for Wantagh on (a) 9 Jun 2020 and (b) 28 Jun 2021.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
As in Fig. A3, but for Wantagh on (a) 24 Jul 2022 and (b) 4 Aug 2022.
Citation: Weather and Forecasting 40, 3; 10.1175/WAF-D-24-0086.1
REFERENCES
Aird, J. A., R. J. Barthelmie, T. J. Shepherd, and S. C. Pryor, 2022: Occurrence of low-level jets over the eastern U.S. Coastal Zone at heights relevant to wind energy. Energies, 15, 445, https://doi.org/10.3390/en15020445.
Benjamin, S. G., and Coauthors, 2016: A North American hourly assimilation and model forecast cycle: The Rapid Refresh. Mon. Wea. Rev., 144, 1669–1694, https://doi.org/10.1175/MWR-D-15-0242.1.
Bianco, L., K. Friedrich, J. M. Wilczak, D. Hazen, D. Wolfe, R. Delgado, S. P. Oncley, and J. K. Lundquist, 2017: Assessing the accuracy of microwave radiometers and radio acoustic sounding systems for wind energy applications. Atmos. Meas. Tech., 10, 1707–1721, https://doi.org/10.5194/amt-10-1707-2017.
Blaylock, B. K., J. D. Horel, and S. T. Liston, 2017: Cloud archiving and data mining of High-Resolution Rapid Refresh forecast model output. Comput. Geosci., 109, 43–50, https://doi.org/10.1016/j.cageo.2017.08.005.
Bodini, N., and Coauthors, 2024: The 2023 National Offshore Wind data set (NOW-23). Earth Syst. Sci. Data, 16, 1965–2006, https://doi.org/10.5194/essd-16-1965-2024.
BOEM, 2022: New York Bight. Accessed 4 April 2022, https://www.boem.gov/renewable-energy/state-activities/new-york-bight.
Bowers, L. A., 2004: The effect of sea surface temperature on sea breeze dynamics along the coast of New Jersey. M.S. thesis, Graduate program in Oceanography, Rutgers, The State University of New Jersey, 13 pp., https://rucool.marine.rutgers.edu/media/downloads/papers/thesis/bowers_thesis_front.pdf.
Brotzge, J. A., and Coauthors, 2020: A technical overview of the New York State mesonet standard network. J. Atmos. Oceanic Technol., 37, 1827–1845, https://doi.org/10.1175/JTECH-D-19-0220.1.
Chan, P. W., and K. K. Hon, 2023: Observations and numerical simulations of sea breezes at Hong Kong International Airport. Weather, 78, 55–63, https://doi.org/10.1002/wea.4282.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569–585, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Cohen, A. E., S. M. Cavallo, M. C. Coniglio, and H. E. Brooks, 2015: A review of planetary boundary layer parameterization schemes and their sensitivity in simulating southeastern U.S. cold season severe weather environments. Wea. Forecasting, 30, 591–612, https://doi.org/10.1175/WAF-D-14-00105.1.
Colle, B. A., and D. R. Novak, 2010: The New York Bight Jet: Climatology and dynamical evolution. Mon. Wea. Rev., 138, 2385–2404, https://doi.org/10.1175/2009MWR3231.1.
Colle, B. A., M. J. Sienkiewicz, C. Archer, D. Veron, F. Veron, W. Kempton, and J. E. Mak, 2016: Improving the Mapping and Prediction of Offshore Wind Resources (IMPOWR): Experimental overview and first results. Bull. Amer. Meteor. Soc., 97, 1377–1390, https://doi.org/10.1175/BAMS-D-14-00253.1.
Djalalova, I. V., D. D. Turner, L. Bianco, J. M. Wilczak, J. Duncan, B. Adler, and D. Gottas, 2022: Improving thermodynamic profile retrievals from microwave radiometers by including radio acoustic sounding system (RASS) observations. Atmos. Meas. Tech., 15, 521–537, https://doi.org/10.5194/amt-15-521-2022.
Donlon, C. J., M. Martin, J. Stark, J. Roberts-Jones, E. Fiedler, and W. Wimmer, 2012: The Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) system. Remote Sens. Environ., 116, 140–158, https://doi.org/10.1016/j.rse.2010.10.017.
Dowell, D. C., and Coauthors, 2022: The High-Resolution Rapid Refresh (HRRR): An hourly updating convection-allowing forecast model. Part I: Motivation and system description. Wea. Forecasting, 37, 1371–1395, https://doi.org/10.1175/WAF-D-21-0151.1.
Doyle, J. D., and T. T. Warner, 1991: A Carolina coastal low-level jet during GALE IOP 2. Mon. Wea. Rev., 119, 2414–2428, https://doi.org/10.1175/1520-0493(1991)119<2414:ACCLLJ>2.0.CO;2.
Draxl, C., A. Clifton, B. M. Hodge, and J. McCaa, 2015: The Wind Integration National Dataset (WIND) toolkit. Appl. Energy, 151, 355–366, https://doi.org/10.1016/j.apenergy.2015.03.121.
ECMWF, 2019: Global reanalysis: Goodbye ERA-Interim, Hello ERA5. Accessed 21 March 2023, https://www.ecmwf.int/en/newsletter/159/meteorology/global-reanalysis-goodbye-era-interim-hello-era5.
Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, https://doi.org/10.1029/2002JD003296.
Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and J. B. Edson, 2003: Bulk parameterization of air–sea fluxes: Updates and verification for the COARE algorithm. J. Climate, 16, 571–591, https://doi.org/10.1175/1520-0442(2003)016<0571:BPOASF>2.0.CO;2.
Fragano, C., 2023: Validation of offshore winds in the ERA5 reanalysis, HRRR, and NOW-23 analyses using two floating LiDARs South of Long Island, NY. M.S. thesis, Dept. of Marine and Atmospheric Science, Stony Brook University, 115 pp.
Freedman, J. M., B. H. Bailey, J. Manobianco, K. T. Waight, J. W. Zack, and M. Brower, 2010: Peak load matching and the offshore wind resource. First Conf. on Weather, Climate, and the New Energy Economy, Atlanta, GA, Amer. Meteor. Soc., J10.4, https://ams.confex.com/ams/90annual/techprogram/paper_161343.htm.
Freedman, J. M., B. Colle, E. McCabe, and C. Christopher, 2023: Integrating unique observations and models to improve offshore wind resource assessment and energy production forecasts. Sunrise Wind Final Rep., 19 pp.
Gaertner, E., and Coauthors, 2020: Definition of the IEA 15-megawatt offshore reference wind turbine. National Renewable Energy Laboratory Tech. Rep. NREL/TP-5000-75698, 54 pp., https://www.nrel.gov/docs/fy20osti/75698.pdf.
Gandoin, R., and J. Garza, 2024: Underestimation of strong wind speeds offshore in ERA5: Evidence, discussion and correction. Wind Energy Sci., 9, 1727–1745, https://doi.org/10.5194/wes-9-1727-2024.
Gedzelman, S. D., S. Austin, R. Cermak, N. Stefano, S. Partridge, S. Quesenberry, and D. A. Robinson, 2003: Mesoscale aspects of the urban heat island around New York City. Theor. Appl. Climatol., 75, 29–42, https://doi.org/10.1007/s00704-002-0724-2.
Good, S., 2018: The Operational Sea Surface Temperature and Ice Analysis (OSTIA) system. ECMWF SST and Sea Ice Workshop, Exeter, UK, Met Office, 30 pp., https://pdfs.semanticscholar.org/a064/8799f8546f8adf7d07697b8b246c0159f1df.pdf.
Grell, G. A., and S. R. Freitas, 2014: A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling. Atmos. Chem. Phys., 14, 5233–5250, https://doi.org/10.5194/acp-14-5233-2014.
Gualtieri, G., 2021: Reliability of ERA5 reanalysis data for wind resource assessment: A comparison against tall towers. Energies, 14, 4169, https://doi.org/10.3390/en14144169.
Gunnerson, C. G., 1981: Chapter 11 The New York Bight ecosystem. Elsevier Oceanogr. Ser., 27, 313–378, https://doi.org/10.1016/S0422-9894(08)71416-8.
Han, Z. S., J. E. González-Cruz, H. N. Liu, D. Melecio-Vázquez, H. Gamarro, Y. H. Wu, F. Moshary, and R. Bornstein, 2022: Observed sea breeze life cycle in and around NYC: Impacts on UHI and ozone patterns. Urban Climate, 42, 101109, https://doi.org/10.1016/j.uclim.2022.101109.
He, C., and Coauthors, 2023: Modernizing the open-source community Noah with Multi-Parameterization options (Noah-MP) land surface model (version 5.0) with enhanced modularity, interoperability, and applicability. Geosci. Model Dev., 16, 5131–5151, https://doi.org/10.5194/gmd-16-5131-2023.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Hu, Y., J. Tan, S. Grimmond, X. Ao, Y. Yan, and D. Liu, 2022: Observed and modeled urban heat island and sea-breeze circulation interactions: A Shanghai case study. J. Appl. Meteor. Climatol., 61, 239–259, https://doi.org/10.1175/JAMC-D-20-0246.1.
Hughes, C. P., and D. E. Veron, 2018: A characterization of the Delaware sea breeze using observations and modeling. J. Appl. Meteor. Climatol., 57, 1405–1421, https://doi.org/10.1175/JAMC-D-17-0186.1.
Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.
International Energy Agency, 2024: Capacity factor. Accessed 18 October 2024, https://www.nrel.gov/docs/fy20osti/75698.pdf.
Janjić, Z. I., 1994: The Step-Mountain Eta Coordinate Model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, https://doi.org/10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Janjić, Z. I., 1996: The surface layer in the NCEP ETA model. 11th Conf. on Numerical Weather Prediction, Norfolk, VA, Amer. Meteor. Soc., 354–355.
Janjić, Z. I., 2001: Nonsingular implementation of the Mellor-Yamada level 2.5 scheme in the NCEP Meso model. NCEP Office Note 437, 61 pp., https://repository.library.noaa.gov/view/noaa/11409.
Janse, R. J., T. Hoekstra, K. J. Jager, C. Zoccali, G. Tripepi, F. W. Dekker, and M. van Diepen, 2021: Conducting correlation analysis: Important limitations and pitfalls. Clin. Kidney J., 14, 2332–2337, https://doi.org/10.1093/ckj/sfab085.
Jiménez, P. A., J. Dudhia, J. F. González-Rouco, J. Navarro, J. P. Montávez, and E. García-Bustamante, 2012: A revised scheme for the WRF surface layer formulation. Mon. Wea. Rev., 140, 898–918, https://doi.org/10.1175/MWR-D-11-00056.1.
McCabe, E. J., and J. M. Freedman, 2023: Development of an objective methodology for identifying the sea-breeze circulation and associated low-level jet in the New York Bight. Wea. Forecasting, 38, 571–589, https://doi.org/10.1175/WAF-D-22-0119.1.
Mesinger, F., 1993: Forecasting upper tropospheric turbulence within the framework of the Mellor-Yamada 2.5 closure. UCAR Visiting Research Program at the U.S. National Meteorological Center Doc., 2 pp., https://www2.mmm.ucar.edu/wrf/users/physics/phys_refs/PBL/MYJ_part2.pdf.
Miller, S. T. K., B. D. Keim, R. W. Talbot, and H. Mao, 2003: Sea breeze: Structure, forecasting, and impacts. Rev. Geophys., 41, 1011, https://doi.org/10.1029/2003RG000124.
Milovac, J., K. Warrach-Sagi, A. Behrendt, F. Späth, J. Ingwersen, and V. Wulfmeyer, 2016: Investigation of PBL schemes combining the WRF model simulations with scanning water vapor differential absorption lidar measurements. J. Geophys. Res. Atmos., 121, 624–649, https://doi.org/10.1002/2015JD023927.
Mirocha, J. D., M. D. Simpson, J. D. Fast, L. K. Berg, and R. L. Baskett, 2016: Investigation of boundary-layer wind predictions during nocturnal low-level jet events using the Weather Research and Forecasting model. Wind Energy, 19, 739–762, https://doi.org/10.1002/we.1862.
Monin, A. S., and A. M. Obukhov, 1954: Basic laws of turbulent mixing in the surface layer of the atmosphere. Tr. Akad. Nauk SSSR Geophiz. Inst., 24, 163–187.
Nakanishi, M., and H. Niino, 2006: An improved Mellor–Yamada Level-3 Model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397–407, https://doi.org/10.1007/s10546-005-9030-8.
Niu, G., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res., 116, D12109, https://doi.org/10.1029/2010JD015139.
NRC, 1983: Low-Altitude Wind Shear and Its Hazard to Aviation. The National Academies Press, 112 pp.
Nunalee, C. G., and S. Basu, 2014: Mesoscale modeling of coastal low-level jets: Implications for offshore wind resource estimation. Wind Energy, 17, 1199–1216, https://doi.org/10.1002/we.1628.
OceanTech Services/DNV under contract to NYSERDA, 2021: Floating lidar offshore wind measurement data. Accessed 5 June 2024.
Olson, J. B., J. S. Kenyon, W. A. Angevine, J. M. Brown, M. Pagowski, and K. Sušelj, 2019: A description of the MYNN-EDMF scheme and the coupling to other components in WRF–ARW. NOAA Tech. Memo. OAR GSD 61, 42 pp., https://doi.org/10.25923/n9wm-be49.
Park, S., and C. S. Bretherton, 2009: The University of Washington shallow convection and moist turbulence schemes and their impact on climate simulations with the Community Atmosphere Model. J. Climate, 22, 3449–3469, https://doi.org/10.1175/2008JCLI2557.1.
Paul, A., M. Afroosa, B. Baduru, and B. Paul, 2023: Showcasing model performance across space and time using single diagrams. Ocean Modell., 181, 102150, https://doi.org/10.1016/j.ocemod.2022.102150.
Pleim, J. E., 2007a: A combined local and nonlocal closure model for the atmospheric boundary layer. Part I: Model description and testing. J. Appl. Meteor. Climatol., 46, 1383–1395, https://doi.org/10.1175/JAM2539.1.
Pleim, J., 2007b: A combined local and nonlocal closure model for the atmospheric boundary layer. Part II: Application and evaluation in a mesoscale meteorological model. J. Appl. Meteor. Climatol., 46, 1396–1409, https://doi.org/10.1175/JAM2534.1.
Redfern, S., M. Optis, G. Xia, and C. Draxl, 2023: Offshore wind energy forecasting sensitivity to sea surface temperature input in the Mid-Atlantic. Wind Energ. Sci., 8, 1–23, https://doi.org/10.5194/wes-8-1-2023.
Seroka, G., E. Fredj, J. Kohut, R. Dunk, T. Miles, and S. Glenn, 2018: Sea breeze sensitivity to coastal upwelling and synoptic flow using Lagrangian methods. J. Geophys. Res. Atmos., 123, 9443–9461, https://doi.org/10.1029/2018JD028940.
Shrestha, B., J. A. Brotzge, J. Wang, N. Bain, C. D. Thorncroft, E. Joseph, J. Freedman, and S. Perez, 2021: Overview and applications of the New York State mesonet profiler network. J. Appl. Meteor. Climatol., 60, 1591–1611, https://doi.org/10.1175/JAMC-D-21-0104.1.
Siuta, D., G. West, and R. Stull, 2017: WRF hub-height wind forecast sensitivity to PBL scheme, grid length, and initial condition choice in complex terrain. Wea. Forecasting, 32, 493–509, https://doi.org/10.1175/WAF-D-16-0120.1.
Skamarock, W. C., and Coauthors, 2021: A description of the Advanced Research WRF Model Version 4.3. NCAR Tech. Note NCAR/TN-556+STR, 162 pp., https://doi.org/10.5065/1dfh-6p97.
Stark, J. D., C. Donlon, M. Martin, and M. Mcculloch, 2007: OSTIA: An operational, high resolution, real time, global sea surface temperature analysis system. OCEANS 2007 - Europe, Aberdeen, UK, Institute of Electrical and Electronics Engineers, 1–4, https://doi.org/10.1109/OCEANSE.2007.4302251.
Strobach, E., L. C. Sparling, S. D. Rabenhorst, and B. Demoz, 2018: Impact of inland terrain on mid-Atlantic offshore wind and implications for wind resource assessment: A case study. J. Appl. Meteor. Climatol., 57, 777–796, https://doi.org/10.1175/JAMC-D-17-0143.1.
Sukoriansky, S., B. Galperin, and V. Perov, 2005: Application of a new spectral theory of stably stratified turbulence to the atmospheric boundary layer over sea ice. Bound.-Layer Meteor., 117, 231–257, https://doi.org/10.1007/s10546-004-6848-4.
Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106, 7183–7192, https://doi.org/10.1029/2000JD900719.
Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 5095–5115, https://doi.org/10.1175/2008MWR2387.1.
Wermter, J., S. Noble, and B. Viner, 2022: Impacts of the thermal gradient on inland advecting sea breezes in the southeastern United States. Atmosphere, 13, 1004, https://doi.org/10.3390/atmos13071004.
WMO, 2023: Guide to instruments and methods of observations. WMO, 574 pp., https://library.wmo.int/records/item/68695-guide-to-instruments-and-methods-of-observation.
Xie, Y., Y. Ishida, H. Watanabe, and A. Mochida, 2023: Impacts of urban development between 2002 and 2022 on the effects of sea breezes in Sendai, Japan—Analyzing heat balance mechanism in urban space. Atmosphere, 14, 677, https://doi.org/10.3390/atmos14040677.
Xu, Y., and Coauthors, 2022: Comparison of urban canopy schemes and surface layer schemes in the simulation of a heatwave in the Xiongan New Area. Atmosphere, 13, 1472, https://doi.org/10.3390/atmos13091472.
Zhang, L., J. Xin, Y. Yin, W. Chang, M. Xue, D. Jia, and Y. Ma, 2021: Understanding the major impact of planetary boundary layer schemes on simulation of vertical wind structure. Atmosphere, 12, 777, https://doi.org/10.3390/atmos12060777.
Zonato, A., A. Martilli, and E. Gutierrez, 2021a: README The new version of the BEP+BEM scheme in WRF 4.3: Local climate zones, mitigation Strategies, building materials permeability and new buildings drag coefficient. 3 pp., https://ral.ucar.edu/sites/default/files/public/product-tool/urban-canopy-model/README_BEP_BEM_MODIFICATIONS.pdf.
Zonato, A., A. Martilli, E. Gutierrez, F. Chen, C. He, M. Barlage, D. Zardi, and L. Giovannini, 2021b: Exploring the effects of rooftop mitigation strategies on urban temperatures and energy consumption. J. Geophys. Res. Atmos., 126, e2021JD035002, https://doi.org/10.1029/2021JD035002.
