SEARCH

SEARCH BY CITATION

Keywords:

  • weather research and forecasting model;
  • solar radiation;
  • Georgia Automated Environmental Monitoring

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[1] A surface dimming effect during a forest fire was observed in the incoming solar radiation measurements of the Georgia Automated Environmental Monitoring Network (AEMN). A combination of in situ AEMN and Moderate Resolution Imaging Spectroradiometer (MODIS) datasets were used to demonstrate the implications on the forecasts when aerosol radiative effects are not included in the Weather Research and Forecasting (WRF) model. The clear sky incoming radiative flux predicted by the model at the surface was overestimated when aerosol optical depths (AODs) exceeded 0.2, which in turn caused a positive temperature bias and a negative mixing ratio bias at the surface. These biases resulted from differences in the energy partitioning at the surface, where the main contribution was from enhanced sensible heat flux. The model atmosphere was also cooler and drier than the MODIS profiles, indicative of the aerosol induced warming below 6 km.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[2] The energy and water cycles of the surface and atmosphere are significantly influenced by aerosols due to their radiative forcing, which in turn influence both weather and climate [Ramanathan et al., 2001]. Remotely sensed observations of biomass burning have repeatedly evidenced the importance of aerosols on radiative forcing and the impact on air quality [Justice et al., 1996; Christopher et al., 2000; Hsu et al., 2003]. Moderate Resolution Imaging Spectroradiometer (MODIS) provides valuable information on the AOD [Remer et al., 2005]. The climatology of regional scale analyses using MODIS observations have indicated that the biomass burning over South Africa contributed to a radiative forcing of −10 Wm−2 and −30 Wm−2 at the top of the atmosphere (TOA) and at the surface, respectively [Kaufman et al., 2002] by reflecting three times more radiation than other types of pollution. In the atmospheric column above the earth's surface, radiation absorbing aerosols such as the ones in smoke [Ramanathan et al., 2001] can also cause a warming of the atmosphere. A combination of these radiative effects may result in a cooler surface layer with warmer atmosphere above, leading to stable atmospheric conditions [Ackerman et al., 2000; Taubman et al., 2004] and the suppression of rainfall [Ramanathan et al., 2001].

[3] In some instances, discrepancies in the solar irradiance predictions by the mesoscale model were traced to the inadequate representation of aerosol radiative forcing [Zamora et al., 2003; Guichard et al., 2003]. The transport of smoke from biomass burning over Mexico was reported to cause an increase in the AOD over the Great Plains of the United States [Peppler et al., 2000]. Atmospheric Radiation Measurements (ARM) taken during several clear sky conditions during this smoke plume advection event showed considerable day to day variability in the downward shortwave radiation flux, which were attributed to the presence of aerosol transport over the region [Guichard et al., 2003].

[4] The solar radiation measurements available from the Automated Environmental Monitoring Network (AEMN; www.georgiaweather.net) showed an area-wide dimming effect associated with smoke from forest fires. The main goal of this study was to evaluate the impact of forest fire events on the mesoscale forecasts from the advanced research weather research and forecasting model (WRF-ARWV2.2) [Skamarock et al., 2005], when radiative effects of aerosols are not included. In this study, remotely sensed MODIS observations of water vapor and aerosol distributions were used in conjunction with the radiation measurements from the AEMN to identify the influence of biomass burning emissions on the mesoscale weather.

2. Materials and Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[5] The Advanced Research WRF model (ARW V2.2) parameterized with the radiation scheme [Stephens, 1984] that incorporates the effects of clear-air scattering, water vapor absorption and cloud albedo and absorption was used in this study. The planetary boundary layer was parameterized with the Yonsei University (YSU) PBL scheme coupled with the Monin Obukhov (MO) similarity for the surface layer and Noah land surface scheme [Chen and Dudhia, 2001]. WRF Singe Moment 3-class (WSM3) microphysics, Kain Fritsch cumulus parameterization, and the Rapid Radiative Transfer Model (RRTM) longwave scheme were used in the simulations. Two two-way interactive nested grids at 9 km and 3 km resolutions and 40 vertical pressure levels were considered with finer resolution (15 levels in the boundary layer) close to the surface. The case study was initialized with NARR [Mesinger et al., 2006] data and the boundary of the outer domain was updated every 3 hours. The simulation was conducted for three months from April to June 2007. This period characterized several forest fire events in southern Georgia and in northern Florida. To analyze the three dimensional spread of the smoke plume, trajectories of particles released simultaneously at different vertical levels were examined with the help of VAPOR (www.vapor.ucar.edu).

[6] A combination of in situ and remotely sensed observations from the AEMN and MODIS were used in this study. The AEMN has over 75 stations across the state of Georgia (G. Hoogenboom, The Georgia Automated Environmental Monitoring Network: Experiences with the development of a state-wide automated weather station network, paper presented at 13th Symposium on Meteorological Observations and Instrumentation and 15th Conference on Applied Climatology, American Society of Meteorology, Savannah, Georgia, 2005). Routine measurements of surface wind, air temperature, relative humidity, surface pressure, solar radiation, rainfall, volumetric soil moisture content and soil temperatures at 5, 10 and 20 cm depths were available from the AEMN averaged at 15-minute intervals. The AEMN data provided an independent evaluation for model results since the AEMN are not associated with the NARR data used for model initialization. Solar radiation is measured with the LI200X pyranometer (Campbell Scientific, Inc., Logan, Utah, USA). Calibrations of the LI200X pyranometer were conducted against 4 LI-COR transfer standard pyranometers under Metal Halide lamps. The transfer standards were calibrated to the Kipp and Zonen CM21. These calibrations were carried out regularly for all sites and periodic checks were conducted every two weeks. MODIS measures the wavelength-dependant aerosol extinction in the atmospheric column referred to as the AOD and the amount of total column water vapor at a 10 km resolution. Co-located MODIS Level II cloud and corrected AOD at 0.47 μm [Remer et al., 2005] at AEMN locations were analyzed. The total column water vapor (IPW) from MODIS at several AEMN locations were compared with the IPW derived from ARW model. Since there were no direct water vapor vertical profile measurements in the study domain, MODIS atmospheric profile data were used to compare with the model. Validation of MODIS profiles with radiosonde showed a bias of ±1 K (http://modis-atmos.gsfc.nasa.gov/_docs/MOD07:MYD07_ATBD_C005.pdf). Two groups of data with AOD < 0.2 and AOD > 0.2 at all grid point locations were considered for the analysis.

3. Overview of the Study Period and Case Studies

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[7] The year 2007 was one of the worst drought years in Georgia history. Forest fires in southeastern Georgia began on April 13, 2007 at Okefenokee Swamp. This event was followed by a period with heavy rainfall. However, the fire resumed and continued until the middle of June with a major fire that burned over 243,000 ha. For May 2007, the AEMN observations showed an average monthly rainfall (for all stations) of 15.05 mm, with a standard deviation between stations of ±3.5 mm, while the ARW model predicted 9.5 (±4.8) mm. Average volumetric soil moisture decreased from 25% at the beginning of the fire to 15% at the end of May. These variations showed a continued drying effect until the end of the fire events. We selected two days from the month of May 2007 (May 1 and May 22) to illustrate the influence of the fire in South Georgia and the effect of smoke advection to north Georgia. These two cases were characterized by no rain and were cloud free; average soil moisture at the soil surface varied between 16% and 18% at the AEMN locations.

4. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[8] An examination of model radiation bias in the four sectors around the location of the fire (Figure 1, indicated with a star) showed consistently high positive biases up to 350 Wm−2 for several days in May during cloud free days. The smoke from the fire was also advected to the inland areas toward Atlanta and those situations were characterized by less incoming radiation and poor visibility. An examination of AOD distribution from MODIS showed two different scenarios: one without advection, such as on May 1 and a second case, such as on May 22, where the smoke plume got advected towards the foothills of the Appalachian mountain, indicating maximum AODs at these respective locations. The AOD distribution from MODIS and the flow trajectories from the ARW model for two selected locations are presented in Figure 1. In Case 1, the smoke from the fire was distributed across South Georgia and the aerosol distribution indicated AODs exceeding 0.6 around the source (Figure 1a). Wind trajectories on this day depicted southerly and southeasterly winds at the fire site, with easterlies in upper boundary layer (Figure 1a). In Case 2, a stagnation effect was noticed (Figure 1b) with trajectories taking a path around the terrain. The trajectories that originated at a higher vertical level in the boundary layer descended to lower layers as they approached the mountains, leading to stagnation. The trajectories that originated closer to the mountains showed flow divisions with lower layer flow turning towards the north and northeast, while the upper level flow turned towards the northwest and west. The distribution of the AODs from MODIS showed two aerosol plumes parallel to the southern Appalachian Mountains that were influenced by a complex flow division in the boundary layer.

image

Figure 1. Moderate Resolution Imaging Spectroradiometer AOD distribution over Georgia on (a) May 1 and (b) May 22. Automated Environmental Monitoring Network locations are indicated with numbers. WRF model trajectories (using VAPOR www.vapor.ucar.org) of particles released from the fire site (star) are also presented.

Download figure to PowerPoint

5. Comparison of ARW Model With Observations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[9] The incoming radiation, 2 m air temperature and 2 m mixing ratio bias (the difference between ARW model and observations of AEMN) as a function of collocated MODIS AODs are presented for the scan time of the satellite at 1000 UTC in Figures 2a and 2b for the two cases mentioned earlier. The overestimation of incoming radiation by ARW is evident on both days when AODs are exceeding 0.2, indicating the likelihood of an area-wide dimming effect for the observed incoming radiation at the surface. The sensitivity of a well-calibrated radiation sensor could also vary for clear, hazy cloudy conditions and a part of the dimming effect could be attributed to this. However, the observed biases in the current situation are much higher than such expected errors. The radiation biases are also concurrent with temperature and mixing ratio errors as described here. ARW2.2 model results showed a warmer (∼2 K), but drier surface layer when AODs exceeded 0.2 (Figure 2a). However, the model results agreed well with observations for most locations for low AODs for both the case studies (Figures 2a and 2b). On May 22, high AODs (>0.6) were observed close to the foothills of Appalachian Mountains and were coincident with the positive radiation biases (150–200 Wm−2), corresponding to an increase in positive temperature biases (1–2 K) and a mixing ratio bias of 2–4 g kg−1 (Figure 2b).

image

Figure 2. Advanced Research WRF model radiation, 2m temperature (T) and water vapor mixing ratio (r) bias at the Automated Environmental Monitoring Network stations on (a) May 1 and (b) May 22 as a function of Moderate Resolution Imaging Spectroradiometer AOD.

Download figure to PowerPoint

[10] Absence of aerosol scattering in the model leads to excess incoming radiation at the surface that was used in partitioning energy into sensible heat flux. The Bowen ratio (ratio of sensible to latent heat fluxes) increased for high AODs as sensible heat flux (50–100 Wm−2) increased and latent heat flux decreased in the model surface layer (Figure 3a). These results are consistent over most AEMN locations and for the two case studies when AOD exceeded 0.2. The ground heat flux varied between 60–120 Wm−2 for both cases and did not show a relationship with AODs. Thus, the main influence of increased irradiance in the model was on partitioning of sensible and latent heat fluxes. The effect of increased radiation and surface temperature in the model also did not contribute significantly to differences in evaporation from the already dry soil (with soil moisture content 18% on May 1 and 16% on May 22).

image

Figure 3. (a) Model Bowen ratio at different locations on May 1 as a function of aerosol optical depth, (b) difference between the Advanced Research WRF model and MODIS total column precipitable water vapor as a function of AOD, and aggregate difference between ARW and MODIS air temperature and dewpoint temperature profiles where AOD > 0.2 and AOD < 0.2 for (c) May 1 and (d) May 22.

Download figure to PowerPoint

[11] Further analysis above the boundary layer suggested that the absorption of radiation by aerosols in the atmospheric layers was also very important in these cases. A comparison of total column precipitable water predicted by the ARW model and from MODIS observations is presented in Figure 3b. A negative model bias when AOD exceeded 0.2 indicated a dry atmosphere in the model; the predicted column water vapor levels are lower than observations from MODIS. These results were consistent for both cases considered. The difference between ARW predicted and MODIS (ARW-MODIS) observed temperature and dew point temperature profiles is presented for two AOD classes (Figures 3c and 3d). A negative model bias in the air temperature and dew point temperature is noted for all vertical levels below 6 km, indicating a cooler and drier atmosphere, and a positive dew point bias is noted above that layer, indicating a wetter atmosphere. The atmosphere warming rate due to the presence of aerosols is evident in MODIS profiles, which is absent in the model profile. Due to the warming, the boundary layer was 3–5 K warmer than the model predicted on May 1(Figure 3c) and 2–3 K warmer on May 22 (Figure 3d), when AODs exceeded 0.2. This difference may be attributed to advection of smoke and its layering. Drying in the model was found throughout the lower atmosphere, including the surface (as evident from the mixing ratio biases in Figure 2) at all locations. However, drying was more prevalent at locations where AOD exceeded 0.2.

[12] The solar irradiance error of the ARW2.2 model contributed to errors in the surface and boundary layer predictions over the region when AODs exceeded 0.2. Zamora et al. [2003] found that the MM5 model predictions were accurate when the AODs were below 0.1. We noticed similar problems with ARW2.2 model for AODs exceeding 0.2. Recent observational studies [Koren et al., 2008; Rosenfeld et al., 2008] on the radiative/microphysical role of aerosols showed that for AOD > 0.25 the radiative effects get predominant. More recent versions of the ARW model incorporate aerosol radiative effects and might rectify radiation errors in the presence of aerosols. However, ARWV2.2 could be used as a benchmark to verify aerosol radiative impact. Aerosol loading events, such as biomass burning, have a significant impact on the radiative forcing. Our results showed that the aerosol loading effect from biomass burning could influence model predictions through the surface boundary layer and radiative feedbacks. Therefore, a proper representation of fire events from satellite data such as MODIS/MISR in radiative transfer schemes is a necessary step to improve model performance.

6. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[13] Incoming solar radiation observations from the Georgia AEMN showed an area-wide dimming effect due to a forest fire event in the southeastern United States that occurred in May 2007. One of the key findings of this study is that errors in the surface temperature and moisture are closely related to the incoming radiation errors, attributed to aerosol loading during cloud free conditions, which gets modified through surface energy partitioning. The impact of these aerosol loading events on the precipitation is crucial as it appears that atmospheric stability increases (which may inhibit cloud formation) due to aerosol-induced warming at the higher levels. The additional scattering and absorption by aerosols as incorporated in the newer versions of WRF model could be evaluated for such observed aerosol loading events. The ARW V2.2 could potentially be used as a tool to isolate the aerosol radiative effects.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References

[14] This study was funded by a partnership with the USDA Risk Management Agency and by a Seed Grant from the College of Agricultural and Environmental Sciences, the University of Georgia. The computational support was through WRF-DTC from Computational and Information Systems Laboratory (CISL), National Center for Atmospheric Research (NCAR) and Research and Computing Center (RCC) University of Georgia, Athens, Georgia are acknowledged. The authors are grateful to two anonymous reviewers for their valuable comments.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Overview of the Study Period and Case Studies
  6. 4. Results
  7. 5. Comparison of ARW Model With Observations
  8. 6. Conclusions
  9. Acknowledgments
  10. References