Urban environments influence precipitation formation via response to dynamic effects, while aerosols are intrinsically necessary for rainfall formation; however, the partial contributions of each on urban coastal precipitation are not yet known. Here, the authors use aerosol particle size distributions derived from the NASA aerosol robotic network (AERONET) to estimate submicron cloud condensation nuclei (CCN) and supermicron CCN (GCCN) for ingestion in the regional atmospheric modeling system (RAMS). High resolution land data from the National Land Cover Database (NLCD) were assimilated into RAMS to provide modern land cover and land use (LCLU). The first two of eight total simulations were month long runs for July 2007, one with constant PSD values and the second with AERONET PSDs updated at times consistent with observations. The third and fourth runs mirrored the first two simulations for “No City” LCLU. Four more runs addressed a one-day precipitation event under City and No City LCLU, and two different PSD conditions. Results suggest that LCLU provides the dominant forcing for urban precipitation, affecting precipitation rates, rainfall amounts, and spatial precipitation patterns. PSD then acts to modify cloud physics. Also, precipitation forecasting was significantly improved under observed PSD and current LCLU conditions.
Many studies present clear evidence that cities influence regional weather via modification of synoptic fronts, urban heat island (UHI) generation [
Shepherd et al. [
Bornstein and LeRoy [
Niyogi et al. [
To fully understand the nature of precipitation in urban environments it is important that the role of aerosols also be considered. Aerosols can change the frequency of cloud occurrence, cloud thickness, and rainfall amounts. At the large scale, aerosols alter evaporation rates of oceans by modifying the transmission and absorption of solar energy to the water [
Aerosols generated in urban and industrial sites have been shown to impact precipitation in South Asian atmospheric brown clouds [
Aerosols from urban and industrial air pollution have also been shown to suppress rain and snow, as precipitation losses over topographical barriers downwind of major coastal urban areas in California and Israel were reported to account for 15–25% of annual precipitation over these regions [
Therefore, precipitation formation involves contributions from large scale forcing and from local urban effects and modification of cloud microphysics due to aerosol ingestion from urban sources. Urban effects on precipitation may be synthesized as based on the hypotheses of Bornstein et al. (2011, personal communication) that suggests that wind, the urban heat island, convergence, and or divergence primarily affects storm formation and path, while aerosol ingestion impacts storms after dynamic effects have set the stage for precipitation to occur via modification of cloud microphysics.
Where previous studies have shown the importance of aerosol concentration and dynamic effects on rainfall over cities, the present study aims to go further in depth by investigating variation of observed PSD and land cover land use (LCLU) in precipitation events at high horizontal resolutions (e.g., 1 km) in NYC. Advancing the present state of weather prediction and analysis requires the ingestion of observed aerosol information. The present research aims to determine whether precipitation results may be improved with ingestion of aerosol PSD data from the aerosol robotic network (AERONET) and assimilation of LCLU data from the National Land Cover Database (NLCD). The present work lends support to the argument that ingestion of aerosol PSD data in numerical models is important in climate modeling and weather prediction and that its study can lead to advancements in the use of data from updated algorithms that run satellite weather instruments, LIDAR, and other remote sensing technologies.
Figure
National Weather Service (NWS) data for various sites in the NYC/NJ region for July 2007.
Data sites over topography. NWS surface stations are shown as red squares, the Upton radiosonde site is shown as a black square, and the CCNY AERONET site is shown as a light blue circle.
AERONET is a network of ground-based sunphotometers that measure sunlight intensity and can retrieve spectral aerosol optical depth, precipitable water, and PSD in diverse aerosol regimes. All products represent an average of the total aerosol column within the atmosphere. PSD may change from day to day over NYC, as chronicled by AERONET data for July 2007 displayed in Figure
PSD over CCNY AERONET station (40.83N, 73.94W) for July 2007.
For this paper, PSD is classified as high volume fine mode (HVFM) or high volume coarse mode (HVCM). PSDs characterized as high volume have volume distribution values of 0.1
LCLU data was obtained from the National Land Cover Database (NLCD 2006). Shown in Figure
RAMS land classes over the NYC/NJ Region (Grid 3).
RAMS is the main research tool for the present study as it allows modification of LCLU and microphysical parameters which include aerosol PSD, particle concentration, and particle size [
In addition to land cover modification, RAMS also facilitates ingestion of bimodal PSDs that often appear in nature [
CCN/GCCN are allowed to deplete upon nucleation/activation and to replenish upon evaporation. From this information, CCN/GCCN masses are calculated from lookup tables [
Simulations for this study incorporate three nested grids, the largest at 16 km horizontal grid spacing, the second at 4 km horizontal grid spacing, and the finest grid at 1 km horizontal grid spacing. All three grids are centered at 40.8N, 74W. Grid 1 covers the northeast US, grids 2 and 3 cover NYC, western NJ, and Long Island and allow for high resolution analysis of the NYC/NJ area. LCLU data from NLCD (2006) was assimilated onto RAMS grids in order to characterize “City” and “No City” cases. For “No City” simulations, urban grid cells were transformed into deciduous broadleaf forest grid cells (Figure
The ensemble of experiments is displayed in Table
Experimental matrix.
Run | Land | Aerosol | Duration |
---|---|---|---|
NI-C | City | Constant PSD | 744 hours (July 2007) |
I-C | City | 11 PSD updates | 744 hours (July 2007) |
NI-NC | No City | Constant PSD | 744 hours (July 2007) |
I-NC | No City | 11 PSD updates | 744 hours (July 2007) |
Run 1 | City | HVFM (July 11, 2007) | 24 hours (July 11, 2007) |
Run 2 | City | HVCM (July 18, 2007) | 24 hours (July 11, 2007) |
Run 3 | No City | HVFM (July 11, 2007) | 24 hours (July 11, 2007) |
Run 4 | No City | HVCM (July 18, 2007) | 24 hours (July 11, 2007) |
Run 1 is the July 11, 2007, precipitation event with the PSD observed above the CCNY AERONET site assimilated into the model for the same date and with City LCLU. Run 2 is the July 11, 2007, precipitation event assimilated with July 18, 2007, PSD data with City LCLU. The July 11, 2007, HVFM PSD will likely suppress precipitation, while the July 18, 2007, HVCM PSD should enhance accumulated precipitation totals. These PSD effects are attributed to hastened/reduced rates of autoconversion due to the presence of higher volumes of GCCN/CCN (reduced GCCN number concentration can result in increased GCCN volume when the modal radius is large as is the case in July 18, 2007). Run 3 is the July 11, 2007, precipitation event with PSD data for the same date and No City LCLU. Run 4 is the July 11, 2007, precipitation event ingested with July 18, 2007, PSD and No City LCLU. Each daily run endured for 24 hours. Test runs showed that spin up beyond 12 hours had no noticeable impact on model results.
In order to prove that incorporation of PSD/LCLU data from AERONET/NLCD improves precipitation estimates, monthly runs with and without observed PSD ingested and current City LCLU were compared against runs with No City LCLU. Results presented in Table
Total monthly precipitation (in mm) for July 2007 runs. Blue depicts positive biases, red depicts negative biases, and black depicts total accumulated precipitation differences of 1 mm between NWS values and simulated results.
July 2007 |
July 2007 |
July 2007 |
July 2007 |
July 2007 | |
---|---|---|---|---|---|
Mineola | 218 |
|
|
|
|
Oceanside | 110 | 110 |
|
|
|
Wantaugh | 84 |
|
85 |
|
|
Bound Brook | 120 |
|
120 |
|
|
Canistear RSVR | 112 |
|
|
|
|
Canoe Brook | 223 |
|
|
|
|
Essex Fells | 194 |
|
|
|
|
Harrison | 163 | 164 | 164 |
|
|
New Brunswick | 141 |
|
|
|
|
New Milford | 183 |
|
|
|
|
Plainfield | 139 |
|
|
|
|
JFK | 134 |
|
134 |
|
|
Central Park | 175 |
|
|
|
|
EWR | 170 |
|
171 |
|
|
LGA | 180 |
|
180 |
|
|
Secaucus | 81 |
|
83 |
|
|
Simulation NI-C over-predicts total accumulated precipitation for eight of 16 sites, while I-C overpredicts at five of 16 sites, suggesting that the presence of the city can suppress precipitation. NI-C produces more rainfall than I-C for 10 of 16 sites. These results are statistically significant to a 99.7% confidence level. Accuracy of total precipitation compared with observations is reduced for the I-NC and NI-NC scenarios. For 11 of 16 sites, the I-C scenario more accurately predicts observed precipitation totals than all other month long simulations. All runs captured the five precipitation events of July 2007 within 12 hours of their onset, consistent with findings by Comarazamy et al. [
Accumulated daily precipitation RMSE for all monthly cases. The bar graph (b) is representative of actual RMSE values for NI-C (blue), I-C (red), I-NC (green), and NI-NC (purple) over 16 sites. The spatial plots (a) are attained via interpolation of error at the closest grid point to the model errors.
Using the method of factor separation [
PSD/LCLU individual contributions on precipitation.
Run 1 is used to determine how well the model captures observed precipitation totals and spatial orientation of the storms. Runs 2–4 are used to determine deviation from Run 1, thus determining ingestion/assimilation effects of different PSDs/LCLU on the precipitation event.
The weak UHI shown in the temperature plots of Figure
Temperature over the NYC region 2 m from the surface at different local standard times.
Hourly precipitation rates for Run 1 over topography (1 wind barb
Hourly total accumulated precipitation for Run 1 over topography.
Figure
Hourly total accumulated precipitation difference between Run 1 and Run 2 over topography.
Analysis of cloud base height (CBH) differences for CCNY, the northern NJ storm, and the southern NJ storm (Figure
CBH differences for three sites for Run 1 minus Run 2 (HVFM–HVCM, City Case) within the region of interest.
Model results show that Runs 3 (HVFM PSD, No City) and 4 (HVCM PSD, No City) produce more precipitation between the 40.4N and 40.9N latitudes and also extend further east than Runs 1 and 2 (Figure
PSD and LCLU variation. (a) HVFM PSD, City (Run 1). (b) HVCM PSD, City (Run 2). (c) HVFM PSD, No City (Run 3). (d) HVFM PSD, No City (Run 4).
Replacing high volume of CCN (Run 3) with high volume of GCCN (Run 4) results in precipitation enhancement due to the presence of a greater volume of GCCN. Precipitation differences for Run 1 minus Run 3 (Figure
(a) Precipitation differences for Run 1 minus Run 3. (b) Precipitation differences for Run 1 minus Run 4.
Number of grid cells per accumulated precipitation total for (a) Run 1, (b) Run 2, (c) Run 3, and (d) Run 4.
Vertical wind plots in Figure
Vertical wind speed for Run 1, Run 2, Run 3, and Run 4. All times in EDT.
CBH differences between Run 1 and Run 3 for three different sites.
Precipitation rates for all July 11, simulations for two major storms over NJ, for City and No City cases.
The aim of this paper was to investigate the role of aerosol particle size distribution (PSD) and land cover land use (LCLU) on storms over and near NYC. PSDs/LCLU obtained from AERONET/NLCD were ingested/assimilated into RAMS and precipitation results for all cases were compared. Results show that precipitation forecasting is improved when observed PSDs are and present LCLU is ingested/assimilated. For 12 of 16 different sites, runs with observed PSD/present LCLU show reduced error over runs that did not have either observed PSD updates or current LCLU.
The deviation from observations is much higher in No City cases than in City cases. Accuracy increases with PSD updates. Analysis of CBH differences suggests that added convection induced by the presence of the city elevates CBH higher. These results support the notion that cities can impact precipitation by potential aerosol concentration increases and size modification associated with the city. Convection induced by the city landscape draws natural aerosols higher into the atmosphere as well. Results show that PSD with a high volume of fine mode particles (HVFM) can suppress precipitation, while a PSD with a high volume of coarse mode particles (HVCM) can enhance accumulated precipitation totals. Spatial precipitation patterns also change.
These PSD effects are attributed to hastened/reduced rates of autoconversion due to the presence of higher volumes of GCCN/CCN (reduced GCCN number concentration can result in increased GCCN volume when the modal radius is large as is the case in July 18, 2007), which enhances/impedes droplet coalescence rates, in agreement with work by Rosenfeld [
The authors declare that there is no conflict of interests regarding the publication of this paper.
The authors would like to thank Robert D. Bornstein for his energy and interest in this paper. This paper was made possible by the National Oceanic and Atmospheric Administration, Office of Education Educational Partnership Program award NA11SEC4810004. Its contents are solely the responsibility of the award recipient and do not necessarily represent the official views of the US Department of Commerce, National Oceanic and Atmospheric Administration.