EMISSION INVENTORY AND SPECIATON

The Sublette County oil and gas emission inventory was provided by WDEQ and processed by both WDEQ and TRC for CALGRID model input. The emission inventory is comprised of emission sources associated with oil and gas production in Sublette County. The stationary source inventories consist of compressor stations, drill rigs, production sources (heaters, flashing, completions, etc.). Mobile sources were added in Run 5. Generally, the compressor stations and drill rigs are large sources of primary NOx, while the production sources are large sources of VOCs. The VOC emissions in the Basin tend to be of the low reactive variety, but are found at high concentrations near the oil and gas sources. According to the observed canister data, the most abundant high reactive VOCs in the Basin are xylenes (m,p and o) and toluene (ENVIRON 2008)[3], however this can vary by site or time-of-day.

Often, the emission inventory is a source of great uncertainty in photochemical grid modeling, especially the precursor VOC portion. The uncertainty arises from both an observational and modeling point of view, i.e., measurement and/or laboratory issues such as species identification and detection limits contribute to uncertainties as well as the fact that each species has specific reactivity profiles and photolysis parameters for use in modeling. In order to fulfill the chemical mechanisms requirements, the raw VOC emission inventory must be speciated (if only total VOCs provided), or grouped into like-species prior to model input. Given the amount of uncertainty and processing required for the VOC emissions, each of the five model runs represent an additional emission inventory assumption; with each assumption generally considered an improvement or a refinement over the prior analyses.

In order to increase the state of knowledge of the Basin emissions, WDEQ has compiled a detailed inventory of oil and gas sources within Sublette Co. during the months of January through March (the winter inventory). This inventory refines the temporal and spatial resolution of the ozone precursor emissions during the important late winter/early spring months. Information provided by the winter inventory provides monthly hours of operation and emissions for the drill rigs; speciated emissions for specific production activity (e.g., dehys and flashing), wellhead engines, completions and flaring. This inventory was available for winter 2008 and used in the analyses, except Run 1 which used 2007 annual production well data due to time constraints.

The use of a refined inventory, particularly for VOC sources, is expected to provide increased model performance since this is often a model input that has large uncertainty. However, even with speciated VOCs provided by the operators, several iterations and decisions were still required to understand model response and ensure consistency. For instance, some sources or source types had speciated emissions for individual hydrocarbons such as methane through octane (denoted as C1 through C8) or benzene, toluene, ethylbenzene or xylene (BTEX) while others provided only total VOCs. The effect of species treatment was assessed in each of the runs described below. Each run provided additional insights into important model responses and sensitivities to emissions and speciation.

The VOC emissions were processed to conform with the CB-IV chemical mechanism contained within the CALGRID model. CB-IV uses structural-lumping to generalize the representation of chemicals and chemical bonds. Four types of species are present: inorganic species, organic species (e.g. formaldehyde, ethane, and isoprene) that are important to represent explicitly, carbon bond surrogates (PAR = single-bonded carbon atom; OLE = carbon-carbon double bond; ALD2 = carbonyl and adjacent carbon atom), and molecular surrogates (TOL = toluene and other monoalkyl-benzenes; XYL = xylene and other dialkly-benzenes and also triakly-benzenes). Several techniques were used to remap a given emissions inventory to the CB-IV species.

The primary technique uses EPA’s Emission Modeling Clearinghouse Speciation (EMCH) methodology/data. First, a Standard Classification Code (SCC) is assigned for each source in the emissions inventory. Next, the VOC amount is scaled up to TOG (Total Organic Gases, which includes methane and ethane) using a given VOC to TOG scale-up factor lookup table (based on SCC code). Next, the SCC code is matched against a 4 digit Profile code. The Profile code provides the lookup index to a table of mass fractions of the different portion of the TOG, as represented as CB-IV species.

Given the importance of formaldehyde in the area, an optional technique (applied for some sources) replaced the formaldehyde estimated by the SCC code method with the actual formaldehyde value provided in the emissions inventory. This value “overriding” technique preserved the TOG mass by rescaling the non-formaldehyde organic CB-IV species to account for the change in formaldehyde added (or subtracted). For example, if the emission inventory formaldehyde amount for a particular source was 5 lb/hr higher than the SCC method estimation, then the non-formaldehyde species were equally scaled down to remove a total of 5 lb/hr from their emission rates.

Of particular note are Runs 1 and 2 that use the Profile 0000 (SCC 0), which is documented as an “Over All Average” profile, independent of industrial source type. This profile provides a different reactive mixture of CB-IV species than typically found in the SCC codes and attendant Profiles for oil and gas activities. Normally, its corresponding VOC to TOG scale-up factor is 1.0, which assumes that no additional methane and ethane is present (mostly VOCs). In the initial version of the emissions preprocessing program, the VOC to TOG scale-up factors from the SCC-based Profiles (rather than Profile 0000) were used. The SCC-based values are typically larger than 1.0 (e.g. Profile 1001 for Natural Gas Internal Combustion Engines has a VOC to TOG scale-up factor of 10.75). As a result, Profile 0000 runs (Runs 1 and 2) have overestimated VOC emission rates for model input. However, upon review of the total reactive species used for model input, Run 1 has similar values (~5,000 lb/hr) to Runs 3-5 and produced significantly more ozone than Runs 3-5. This is likely due to its more reactive mix of VOCs.

Rather than using the reported total VOC amount, another technique converted each reported organic pollutant (e.g. propane, pentane, benzene, etc.) in the emissions inventory into CB-IV species. Conversion scalars were obtained from Dr. William Carter’s (UC Riverside) “emitdb” database (Carter 2008)[5] which has a comprehensive list of scalars to convert a given organic pollutant into the corresponding lumped species of different chemical mechanisms, including CB-IV. This technique does not make any adjustments or estimations for unknown or missing organic species. If a particular VOC or TOG pollutant is not reported, its value is assumed to be zero.

As discussed above, VOC emissions require substantial processing. Each of the runs below used different emissions inputs, assumptions or processing and are described briefly below. Total emissions are summarized in Table 1.

Table 1. Summary of emission inventory for Runs 1-5:

Due to time constraints, CALGRID Run 1 used the 2007 annual well production inventory. The VOC emissions were speciated by default using Profile 0000, which corresponds to SCC 0. Compressor stations were run at their potential to emit values and the drill rigs were run at their reported values.

Run 2 used the 2008 winter inventory for production sources. The VOC emissions were speciated by default using Profile 0000 (SCC 0). Compressor stations were run at their potential to emit values and the drill rigs were run at their reported values.

Run 3 used the 2008 winter inventory and speciated the VOC emissions using respective SCC codes for each source type. Compressor stations were run at their potential to emit values and the drill rigs were run at their reported values.

Run 4 used the 2008 winter inventory and used the operator reported speciation when possible and respective SCC codes for all other sources. Compressor stations were run at their potential to emit values and the drill rigs were run at their reported values.

Run 5 used the 2008 winter inventory and used the operator reported speciation and field-specific speciation when possible and respective SCC codes for all other sources. Compressor stations were changed to reflect their actual emissions and the drill rigs were run at their reported values. Mobile sources were added to the inventory. It is important to note that the number of drill rig and well sources increased in Run 5 since like-drill rigs were not combined into a single source as they were in Runs 1 – 4; and similarly each well source was modeled with its own stack parameters rather than grouped together as a single source as they were in Runs 1 - 4.

Gridded emissions for Run 5 of key species such as NOx, xylene (XYL) and toluene (TOL) nearby the Jonah federal reference monitors are provided in Figures 1-2. From review of the figures it is evident that the emissions are heterogeneously spread in the Basin with localized areas of high NOx and/or VOC emissions. The heterogeneity of the emissions, when combined with local-scale transport and diffusion under stagnant conditions, increases greatly the complexity of the analysis.

Figure 1. Hourly gridded NOx emissions (lb/hr) near the Jonah Monitor (Run 5):

Figure 2. Hourly gridded XYL and TOL emissions (lb/hr) near the Jonah Monitor (Run 5):

SENSITIVITY ANALYSES

CALBOX

Following the consistent underprediction of ozone concentrations (as discussed in Section 4), analyses were performed to try and systematically test known model uncertainties, e.g. the chemical mechanism. The CALGRID CB-IV mechanism was tested by simplifying the model inputs until they represent a box model (i.e., CALBOX) with no horizontal advection (from boundary conditions or sources) or vertical advection (from reservoirs aloft).

CALBOX was applied to the February 20, 2008 observed morning canister data at Jonah such that it can be compared to the box modeling conducted by ENVIRON using the OZIPRW model updated with CB05 chemistry (ENVIRON 2010)[6]. The models were homogenized by using the same observed data for initial conditions (observed NOx, speciated VOCs and ozone on the morning of February 20, 2008 at Jonah) and preventing entrainment from layers aloft (fixed mixing height for OZIPRW and vertical advection and diffusion turned off for CALBOX), making it easier to assess the differences due to the chemical mechanisms alone. As illustrated in Figure 3, both models produce ozone at concentrations near 200 ppbv. Consistent with other studies, CB05 predicted larger ozone than CB-IV; in this case about 5% larger. These results indicate the CB-IV chemical mechanism has the ability to produce similar ozone concentrations to the CB05 chemical mechanism given the same precursor concentrations. The results of this, and several similar analyses, indicate that CB-IV by itself is likely not the sole reason for underprediction of ozone with CALGRID.

Figure 3. Comparison of CALBOX to OZIPRW using Jonah canister data on February 20, 2008:

MODEL RESULTS AND ANALYSES

The CALGRID results for Runs 1 through 5 used the above-described model inputs and emissions. The hourly timeseries of model predicted ozone versus observed ozone were plotted at several monitors throughout the Basin, including the Boulder, Daniel and Jonah monitors. The timeseries comparisons were performed for the ‘nearby’ 30 km by 30 km area and at the grid cell of the monitor location (spatially paired). Contours of ozone across the Basin were also produced for review.

Run 5

The 1 hour timeseries plots of layer 1 (10 m) modeled ozone concentrations are provided for Jonah in Figure 4. The observed ozone concentrations are provided for comparisons. Similar to Runs 3 and 4, the model continues to under-predict at all monitor locations with decreases of 20 ppbv in ozone at Boulder and Jonah and little change at Daniel when compared to Run 1.

The domain-wide layer 1 ozone contours are provided for 1 pm on February 21, 2008 as shown in Figure 5(d). The areas of high ozone are similar to all previous runs, but like Run 4 exhibit weak spatial gradients and consistently lower magnitude concentrations than Runs 1 and 2 and even Run 3.

Figure 4. Comparison of timeseries for Runs 1 to 5 to observed hourly ozone concentrations at Layer 1 (10 m) at Jonah:

Figure 5. Contours of hourly ozone at Layer 1 (10 m) at Jonah on Feb. 21, 2008 13:00 LST for Run 1 (a), Run 2 (b), Run 4 (c), and Run 5 (d). Red polygon represents ozone non-attainment area. Terrain contours shown as brown lines.

REFERENCES

AUTHORS

Michael B. Newman¹, Jason F. Reed¹, Ken Rairigh², David Strimaitis¹, Gale F. Hoffnagle¹

1. TRC, 21 Griffin Road North, Windsor, CT 06095
2. Wyoming Department of Environmental Quality, Air Quality Division, Cheyenne, WY, USA

PUBLISHER NOTES

Paper originally published October 10, 2010 as an extended abstract to accompany a poster presented at the 9^th Annual CMAS Conference, Chapel Hill, North Carolina.

INTRODUCTION

The UGRB, which is located in Sublette County, is bounded by the Wind River Range to the east, the Wyoming Range to the west, the Gros Ventre Range to the north, and bounded by the Uinta Range to the south. These surrounding, significant terrain features effectively create a bowl-like basin that greatly influences the local meteorology relative to the rest of the area. The UGRB is roughly 1,000 meters to 2,000 meters lower than the terrain features to the east and west[1]. Within the UGRB, significant development of oil and gas fields has occurred recently. This development has resulted in the release of significant quantities of NOx and VOC emissions, which are both known ozone precursors. Recent monitoring data has indicated elevated ozone concentrations in the late winter that exceed the National Ambient Air Quality Standards (NAAQS) for 8-hour ozone. As a result of these high concentrations, the WYDEQ initiated the Upper Green River Winter Ozone Study (UGWOS) to understand and characterize observed phenomenon.

The 2008 UGWOS final report[2] documents the monitoring network and field operations during Intensive Operating Periods (IOPs) that were part of the study during February and March, 2008. The UGWOS field study produced a high quality database of observations for several meteorological parameters, both during IOPs as well as regular, hourly observations during February and March 2008. Regular hourly observations from the dense network of surface stations provided surface measurements of wind speed, wind direction, temperature and precipitation. A multi-level SODAR also provided hourly measurements of wind speed, wind direction and an estimate of mixing heights. The regular, hourly observations obtained during the field study are also supplemented with high resolution measurements of the wind speed, wind direction and temperature profiles during the IOPs. In the formulation of the CALMET database[3], these observations were blended with the hourly analysis fields from the 20-km Rapid Update Cycle (RUC) numerical weather prediction model and other regional surface and upper air data sites.

Other model inputs included the USGS 1:250,000-scale Digital Elevation Model (DEM) dataset with a horizontal resolution of 3 arc-seconds (approximately 90 meters); the Composite Theme Grid (CTG) dataset for land use and land cover data; and daily National Operational Hydrologic Remote Sensing Center snow depth data. The inclusion of specific snow depth data into the CALMET simulation was an important detail that was necessary to accurately reproduce mixing heights observed at the SODAR site; the snow cover data will be discussed in more detail later. Further assessment of available meteorological data allowed for a robust analysis of model input and model output over the modeling domain (spatially) and modeling period (temporally). The development and assessment of the CALMET database will be discussed more thoroughly herein.

METEOROLOGICAL CHARACTERISTICS OF WINTER OZONE IN THE UGRB

Given the atypical occurrence of high concentrations of ozone forming during a cold season in an area with little urbanization, extensive research of the recent winter ozone events was undertaken by WYDEQ. Based on this research, several common factors were identified during the late winter high ozone events. These factors include:

• continuous emissions of VOC and NOx from oil and gas field sources,
• extensive snow cover,
• sunlight,
• low wind speeds,
• wind reversals (pollutant recirculation), and
• stable atmosphere with strong temperature inversions1.

These meteorological factors were generally the result of a synoptic-scale high pressure system controlling regional weather. As the high pressure system centered over the UGRB, wind speed and mechanical turbulence were reduced. At the same time, snow cover and low wind speeds allowed the Basin atmosphere to cool rapidly and create strong temperature inversions, which persist due to the low wind speeds and quiescent weather. Finally, the increased albedo from the snow cover increased the upward radiation flux for photochemical reactions.

The importance and sequence of the above factors is confirmed by the success of the forecasting methodology utilized by WYDEQ when planning for IOPs. Prior field campaigns successfully predicted high ozone events, and as a result, additional instrumentation/monitoring equipment were in place for the IOPs to monitor the antecedent meteorological conditions leading up to and during the 2008 winter high ozone events and successful IOPs over the two month period2.

CALMET DIAGNOSTIC MODEL SETUP

Proper characterization of the meteorological conditions associated with this complex environment was necessary to assess the spatial extent of the above-described meteorological conditions. The regulatory version of the CALMET diagnostic meteorological model (version 5.8, level 070623) was used to integrate observational and RUC analysis data into a high resolution 3-D wind field. The CALMET model uses high resolution terrain and land use data to diagnostically adjust boundary layer parameters and kinematic effects across the domain, as appropriate[4].

Figure 1 shows a plot of the CALMET computational domain while Figure 2 shows the structure of the vertical layers used in the simulations. In order to capture pertinent transport and edge effects relative to the UGRB, the CALMET domain extends into northwestern Colorado, northeast Utah and southeast Idaho. The CALMET domain is defined by 464 x 400 1-km grid cells in the horizontal direction and 14 vertical layers (surface to 3500 meters at the top). The vertical layers in the CALMET simulation were selected based on observations of mixing heights in the UGRB to help better define the mixing height during high ozone events.

Figure 1. Plot Of The CALMET Modeling Domain With Terrain:

Figure 2. Plot Of The CALMET Vertical Layers Used In The Simulations:

The National Center for Environmental Prediction (NCEP) 20-km resolution Rapid Update Cycle (RUC) hourly analysis fields were used as the initial guess field and later blended with surface observations from 19 mesonet sites to produce a final CALMET database. The 19 mesonet sides provided, at a minimum, local-scale wind speed and wind direction; some monitoring sites also recorded relative humidity, temperature and air pressure. Cloud cover from the National Weather Service (NWS) site at Big Piney, Wyoming was also used. Figure 3 shows a plot of the meteorological monitoring stations within and around the UGRB used in the analysis.

Figure 3. Plot Of Surface And Upper Air Meteorological Stations:

In addition to the RUC data, the aloft meteorological data was based on the hourly SODAR wind data (valid up to 250 meters) and the sonde releases during the IOPs. As each site is within the UGRB, excellent spatial resolution is provided throughout the area; however, both of these upper air datasets also have limitations when implemented in CALMET. The SODAR does not measure temperature, thus temperature data from the RUC grid cell that is nearest to the SODAR site was used to supply the temperature data for this upper air file (UP.DAT). The sonde data provided excellent resolution of all measured parameters, but the balloon/sonde releases were often spaced farther apart in time than what is required by CALMET (more than 14 hours). In order to fulfill the time requirement, a pseudo-sounding from the closest RUC grid cell was used to fill the time gaps, when appropriate. Additional regional upper air sites were also included in the analysis, such as Riverton, Wyoming and Salt Lake City, Utah.

To the extent possible, all data collected during the IOPs were either used in the CALMET model simulations or used as validation in the CALMET model performance evaluation. The estimated mixing height data from the SODAR could not be used as explicit model input; therefore, it was used to validate the CALMET-predicted mixing heights. It was this validation/analysis that highlighted the need to replicate the boundary layer energy budget associated with the influence of snow cover, which is inherently contained within the observations. Another analysis compared the sonde vertical temperature data and the RUC vertical temperature data at the sonde grid cell for the same time periods. Since this comparison of independent datasets demonstrated that the RUC temperature data exhibited very good agreement with the sonde data, the RUC temperature data was used as the source of temperature data for the analysis (ITPROG). These analyses will be discussed in more detail below.

The CALMET model was run with default regulatory settings (MREG=1) and default settings, except for the switch settings discussed below. The project specific settings such as TERRAD, radius of influence values (R-values), etc. were based on domain topography (specifically within the UGRB) and station density, respectively. All model switch settings were qualitatively assessed via graphical analyses to determine the most representative values for TERRAD and the terrain weighting influence parameters. The final non-default switch settings used in the CALMET wind field simulations are as follows:

• TERRAD = 15 km (representative length scale for terrain feature in and around the UGRB),
• R1, R2 = 4 km (limits station influence in complex meteorology and interference with surrounding mesonet sites),
• R1MAX, R2MAX = 20 km (limits station influence in complex meteorology and interference with surrounding mesonet sites),
• ITPROG = 2 (temperature fields based on prognostic data)
• MNMDAV = 5 grid cells (the default is 1 grid cell).

The MNMDAV parameter controls the smoothing of the mixing height field. As a mixing height is calculated for each grid cell, the default value resulted in a noisy mixing height field due to the complex terrain in and around the UGRB. Through sensitivity analyses, a value of 5 was found to retain important features while reducing rapid cell-to-cell variations. The ITPROG setting will be discussed in more detail below.

GEOPHYSICAL PROCESSING AND ENHANCEMENTS

The geophysical data used in the CALMET simulations consisted of United States Geologic Survey (USGS) 90 meter resolution DEM data. The land use data from the USGS Composite Theme Grid (CTG) database, at a minimum resolution of 200 meters, were used to produce a gridded field of the dominant land use categories within each grid cell.

Snow is an important component of the land/air interface that greatly affects the surface energy budget within the boundary layer. When snow is present, much of the solar insolation (short-wave UV energy) is reflected back to the atmosphere, negating potential heating and resultant mixing near the ground. Conversely, when snow is not present (as in the standard land cover data), much of the short-wave solar insolation energy is absorbed by the ground surfaces and reradiated as long-wave (infra-red) energy. As a result, much less of the incoming solar insolation flux is reflected (because of the lack of snow cover), and is instead absorbed by the ground surfaces, thereby heating the ground surfaces, and initiating convective and mechanical mixing of air near the surface and within the boundary layer, allowing mixing heights to grow throughout the day.

This effect was confirmed through the initial CALMET simulations, which did not explicitly include snow cover by default. These CALMET simulations produced mixing heights which grew as a function of daytime heating, i.e., continuously increasing heights throughout the afternoon to over 500 meters. This is in direct contrast to the SODAR estimated mixing heights, which revealed much lower mixing height values. As a sensitivity analysis, the entire domain was (through remapping GEO.DAT geophysical parameters) modeled as if covered completely by snow. Figure 4 shows a time series plot of mixing heights on February 20, 2008 from the CALMET simulation without snow cover and the CALMET simulation assuming the domain is completely covered with snow compared to the SODAR estimated mixing heights. As shown in Figure 5, the CALMET simulations with snow cover produced mixing heights that were much more in line with those observed at the SODAR, (e.g., 250 meters or less). Thus, the refinement of the geophysical data to reflect the actual amount of snow cover during the IOPs, rather than all-snow or no-snow, was considered to be an important value-added step and undertaken.

Figure 4. Time Series Plot Of February 20, 2008 CALMET Predicted Mixing Heights With And Without Snow Cover Compare To The Sonde Estimated Mixing Heights:

Figure 5. Plot Of The Albedo Within The CALMET Modeling Domain Compared To Snow Depth Over The Central Rockies Region. Snow Depth Plot From The National Hydrologic And Remote Sensing Center:

Using daily snow cover data at a 1 km resolution from the National Operational Hydrologic Remote Sensing Center, each CALMET grid cell in the surface layer (Layer 1) were assessed to determine if they were covered by snow according to the National Operational Hydrologic Remote Sensing Center database. If the grid cells were determined to have sufficient snow cover (greater than 50 % of the grid cell is covered by snow), the geophysical parameters such as albedo, Bowen ratio and surface roughness were modified to values recommended by the AERSURFACE manual for each snow covered land use type. Some exceptions to this approach were made in areas where there was barren ground or low vegetation land use. For these land use types, the surface albedo for perennial snow and ice were used. If grid cells were determined to have less than 50% snow cover, the geophysical parameters were modified to values recommended by the AERSURFACE manual for each non-snow covered land use type (late fall). The resulting geophysical database better represents those land surface conditions that actually occurred within the UGRB during IOPs. Figure 5 shows a plot of albedo within the CALMET modeling domain compared with a snow depth map from National Operational Hydrologic Remote Sensing Center for February 20, 2008. Thus when snow cover is included in the land use processing for CALMET (GEO.DAT), more accurate and realistic land surface properties are provided to CALMET which should help improve the CALMET predictions of mixing heights, temperature structure and atmospheric stability.

The determination of snow cover was evaluated on a daily basis. In order to be compatible with CALMET version 5.8, a separate snow cover-enhanced GEO.DAT file was created for each day of the modeling period rather than creating a time-varying geophysical file. The CALMET model was, in turn, run individually for each day of that period.

CALMET VERIFICATION

As part of the CALMET database development, a detailed verification of the meteorological data input/output was conducted for several monitoring sites within the UGRB, specifically the ozone monitor and SODAR locations. Through the use of time series and scatter plots, the raw data (input) and the CALMET simulated data (output) were analyzed to confirm interpretation and proper implementation by the CALMET model, (e.g., units, timing issues, diurnal variations, etc). On a domain-wide basis, descriptive performance statistics were generated from PRTMET output for each vertical layer, and for both the raw input data and derived meteorological parameters. Using these data, domain-wide statistics could be rapidly evaluated and plotted for quality-assurance (QA) purposes. This QA step also allowed for further examination of statistical outliers (minimum and maximum values) of the parameters. For example, it was verified that the maximum horizontal wind speeds (~ 40 m/s) occur in the areas of high elevation terrain features, which are at the highest vertical levels within the CALMET domain. Similarly, an analysis of the hour with the maximum and minimum vertical velocity (absolute value ~10 m/s) indicates that maximum vertical velocities occur as an isolated couplet in a small region of the domain, while the remainder of the domain had more typical vertical velocities. This is shown in Figure 6.

Figure 6. Two-Dimensional Plot Of CALMET Vertical Velocities:

Other QA procedures included two dimensional plots of specific hours or days for pertinent meteorological variables such as mixing height and wind speed. Wind vectors were also plotted during various times and at various vertical layers throughout the period that was analyzed. The plots aided in determination of representative weighting values (TERRAD, R-values) and also assisted in assessing CALMET’s terrain adjustment procedures. Figure 7 shows a set of wind vector plots over the UGRB area at the surface, 350, 1125, and 3000 meters above ground; this plot clearly shows the shallow slope flow on the surface wind field plot and reduced terrain effects with increasing height. Figure 8 shows a more detailed wind vector plot for February 20, 2008 at 0400 LST; this plot clearly shows the downslope flows as well as the terrain blocking effects occurring as the stable flow is deflected around terrain features. Overall, the kinematic terrain effects appropriately extended into the UGRB during times of downslope flows, without undue influence. Figure 9 shows 2-D plots of CALMET predicted mixing heights on February 20, 2008 at 10:00, 12:00, 14:00, and 16:00 Mountain Standard Time. The mixing height plots show the expected diurnal changes in mixing heights. The effect of snow cover is clearly shown with much higher mixing heights occurring over those parts of the modeling domain with bare ground (also see Figure 5).

Figure 7. Plot Of CALMET Predicted Wind Fields At The Surface, 350, 1125, And 3000 Meters Above Ground For February 20, 2008:

Figure 8. CALMET Predicted Wind Field Plot Over The UGRB On February 20, 2008 At 0400 Local Standard Time:

Figure 9. Plots of CALMET predicted mixing heights on February 20, 2008:

In addition to the above QA and as discussed previously, the UGWOS database contained meteorological data that allowed for a robust model validation effort. Of particular importance in the formation of winter ozone are the observed low-level mixing heights, which were estimated from the SODAR. The SODAR emitted and received acoustical impulses periodically. Based on these acoustical signals and the resulting backscattering of acoustic energy received by the SODAR, atmospheric turbulence and the associated mixing height values were inferred by a trained meteorologist. The IOP studies occurred during periods of low mixing heights that persisted throughout the period, despite the abundant sunlight and clear skies present in the area. The strong temperature inversions and low wind speeds created a stable lower atmosphere, which inhibited vertical mixing and limited transport and dispersion of emissions within the UGRB. These conditions were a result of, and were exacerbated by, the extensive snow cover across the UGRB.

The mixing heights estimated by the SODAR were compared to the mixing heights calculated by CALMET. These hourly comparisons were conducted for the entire modeling period and show good agreement in the general magnitude of daily maximums and nightly minimum mixing height values (see Figure 4 for the comparison on February 20, 2008). As previously discussed, the inclusion of snow cover into the CALMET model increased the accuracy of its mixing height predictions.

ANALYSIS OF VERTICAL TEMPERATURE PROFILES

The vertical temperature profile is an important parameter used in the computation of mixing height and kinematic effects. As mentioned above, the availability of detailed sonde data during the IOPs provided a high resolution, in-situ dataset. However, its limited availability (temporally and spatially) reduced its overall usefulness as the primary source of vertical temperature data across the domain. Given the unique topographic setting of the UGRB, the sonde data is not representative of the vertical temperature structure in other regions of the domain outside the UGRB. Since the CALMET data will be used for a variety of purposes, both local scale and regionally, the simulations must not only properly define the temperature field within the basin, but also adequately define the temperature field across the entire modeling domain.

The sonde temperature data was a critical part of the evaluation of the RUC temperature data. The evaluation of the RUC vertical temperature profile was performed through direct comparisons with the sonde vertical temperature profiles and the RUC profiles at the RUC grid cell that is closest to the location of the sonde releases.

Figure 10 shows a plot of the vertical temperature profile from the sonde measurements and for the nearest RUC grid point. This plot shows very good agreement between the sonde and RUC profiles, which was the case for other times and days as well. Generally, the RUC data performs well in replicating the temperature profile at the SODAR grid point – especially in the morning soundings. In early and late afternoon, the RUC profile mimics the general temperature profile as compared to the rawinsonde vertical temperature data. Although there are differences, the vertical temperature profile from the RUC analysis compares reasonably well with the observed upper air soundings and rawinsonde data. Other analyses, such as the consideration of the vertical temperature profiles from within and outside of the domain, confirm that the strong, persistent temperature inversions and very shallow mixing heights are generally confined to the UGRB during this period.

Figure 10. Plot Of The Vertical Temperature Profile For Both The Sonde And The Nearest RUC Model Grid Point On February 20, 2008:

Based on the above findings, the RUC temperature data was used as the primary temperature data source for the entire CALMET domain (ITPROG=2). By driving the analysis with RUC temperature data, representative, spatially and temporally refined temperature fields are used for important stability calculations rather than quasi-periodic sonde data, or relying on more distant and less representative data such as the soundings at Riverton, WY.

CONCLUSIONS

The winter ozone events in the UGRB occur during periods of complex meteorology in an area of complex terrain. The CALMET diagnostic model is well-suited for this application since it can provide a fine-scale horizontal and vertical resolution of the area while integrating all available meteorological data obtained during the 2008 UGWOS field study and associated IOPs. The high resolution in the horizontal allows refined treatment of land cover and snow cover data, which are important to boundary layer stability and growth. The high resolution in the vertical allows for generating refined CALMET-predicted mixing heights and temperature inversions, both of which are considered important meteorological components of winter ozone formation in the UGRB.

Data which could not be, or was not used directly by CALMET was employed to verify CALMET and RUC model performance. The use of SODAR-derived mixing heights highlighted the importance of snow cover on the energy budget within the UGRB. If default values were used for land cover, the mixing heights would have been consistently biased high during the day. The inclusion of snow cover provides the CALMET model with a more realistic energy budget consistent with the actual conditions. Similarly, the good agreement between the sonde and RUC vertical temperature data demonstrates the value of using high quality high resolution prognostic data as the primary data set to initialize the 3-D wind and 2-D temperature fields, which allows for a better characterization of the wind and temperature fields on a spatial and temporal basis. These two factors, directly attributable to the model validation effort, increased the confidence of the analyses, results, and decisions made thereupon.

The results of these CALMET simulations show that with suitable observational data and mesoscale numerical weather prediction model data, the CALMET model provides a very viable and efficient approach that can yield a realistic representation of three-dimensional meteorological fields important for air quality modeling. These CALMET simulations provided a more refined and improved representation of meteorological conditions essential for air quality modeling that could not be achieved using observations or mesocale modeling alone. This is largely due to the capability of CALMET to directly incorporate meteorological and geophysical observations (i.e. snow data) and couple these data with gridded prognostic data. The model also has the ability to efficiently diagnose kinematic effects and terrain-adjusted flows that are important in valley areas such as the UGRB.

This study shows that the CALMET model configuration should be based on the following considerations:

• detailed quality assessment of the meteorological observations,
• detailed quality assessment of the mesoscale model gridded data,
• comparisons of mesoscale model gridded data with available observations, and
• the consideration of terrain and land use characteristics of the modeling domain.

Accurate diagnostic modeling requires a careful analysis of both observational data and gridded prognostic data to understand fully both the data quality and relationship between these data sources. This information, coupled with an understanding of the terrain and land use characteristics of the region and their affect on air quality, is important when selecting the various CALMET model switch settings that control the simulation (e.g. radius of influence settings or use of observations versus prognostic data).

Additional QA procedures to evaluate the full CALMET simulations, such as the automation of PRTMET to review various meteorological data, and the statistical procedures employed in the CALMET wind field validation add to the confidence and the overall veracity of the results.

REFERENCES

KEYWORDS

CALMET, Wyoming, validation, ozone, mixing heights

AUTHORS

Ken Rairigh¹, P.E.; Jason F. Reed²; Alfred M. Klausmann³, CCM; Michael B. Newman², P.E.; Joseph S. Scire³, CCM; Gale F. Hoffnagle²,CCM, QEP

1. The Wyoming Department of Environmental Quality – Air Quality Division, Herschler Building, 122 West 25th Street, Cheyenne, WY 82002
2. TRC Environmental Corp., 21 Griffin Road North, Windsor, CT, 06095
3. TRC Environmental Corp., 650 Suffolk St. #200, Lowell, MA, 01854

PUBLISHER NOTES

Paper originally published October 26, 2009 as Paper 47 for the meeting of the Air and Waste Management Association: Guideline on Air Quality Models: Next Generation of Models, Raleigh, North Carolina.

What is spam?

Spam is typically defined as “unsolicited e-mail”. (Spam can also be used as a verb in the sense of sending unsolicited e-mail to someone.) This is along the lines of “junk mail”, which is unsolicited postal mail. Most people argue “I didn’t ask for it, I shouldn’t get it.” But I believe that statement is wrong, and on top of that the issues (or true problems) are more complex.

Spam should not be confused with SPAM (TM) the luncheon meat. Please see Hormel’s explanation.

I recommend spelling the e-mail-type of “spam” in all lowercase letters when stylistically possible.

Additional Terms:

• “Spammer” - someone who voluntarily spams
• “Spam bucket” - e-mail account that you use publicly in order to draw attention away from a different e-mail address that you want to use with a more select private audience
• “harvesting” - seeking out and gathering e-mail addresses with the intent to spam, done automatically (by spam bot) or manually (by spammer)
• “Open relay” - a mail server that accepts e-mail from anyone and sends it anywhere
• “UCE” - Unsolicited Commercial E-mail; Spam

What is the problem?

An engineer always asks “what is the problem?”. Most people will just rush and say “kill spammers” or “we need Spam filters”. But an engineer takes a more global look by directly addressing the underlying problems. Often, more effective (and humane :-) solutions can be found this way. Unfortunately, there are multiple problems:

• User’s attention/time being wasted/abused.
• Bandwidth being wasted: clogging up preventing/slowing down other data transfer (once again user’s time wasted), money wasted
• Messages offend user (emotional damage/peace of mind)
• Messages corrupt minds of children (emotional damage)
• Waste of storage space (hard drive on either system or user’s computer - territorial encroachment? probably not a valid problem since even nonspammers’ messages take up your hard drive space, but perhaps its a “money wasted” issue, especially for server maintainer if he/she has to buy more storage space to handle spam influx)
• Worst of all, a message can contain many forms of viruses (macro/script, executable) which can damage your computer (time wasted, money wasted, emotional damage) or spam other people

So in summary, Spam is troublesome because it:

• Wastes time
• Wastes money
• May cause emotional damage

But wait, there are counter-arguments to that!

• Saves time (you don’t have to waste your time look for bargains)
• Saves money (they offer/give you bargains)
• Conveys good stuff, make people feel good (I suppose this is possible, but I never got a spam message with this tone of message)
• Educate/enlighten people (once again, never have seen this, but its possible!)

I dispute such counter-arguments because spam does not save net-time nor net-money (ironic play on the word “net”) nor does it on average make people feel better or properly (factually, honestly) educate people. Unfortunately, I love to overanalyze things. Its not illegal to waste someone’s time from what I understand. But perhaps you are holding someone’s attention hostage. :-). Hostage = “A prisoner who is held by one party to insure that another party will meet specified terms” (WordWeb dictionary). They hold your attention hostage so that you buy their product. Now wasting someone’s money is a property damage from what I understand and can be litigated in court. However, their is a minimum amount of damages that must be present before such a case can be brought before a court (I think). Plus, regular messages consume your money too. That leads up to “social contract”:

Social Contract

John Locke, where are you? Perhaps what we are failing to address is the existence of an unstated “social contract”. That is “an implicit agreement among people that results in the organization of society”, or how society acts (WordWeb dictionary). Some people believe that in terms of a “social contract”, when you have an Internet e-mail account, you are agreeing to receive any message from any person. This viewpoint unfortunately lets spammers off the hook. I think its wrong to allow people to send messages that cause you emotional damage. So, the “social contract” instead should be you are agreeing to receiving non-threatening messages from any person. So your time and money can still be wasted. :-( Hopefully someone has a better idea.

Right to Receive Spam (Argh!)

Its hard to ban spam outright though because it seems people should have the right to receive spam if they want to. Otherwise, we would be dealing with censorship and extreme curtailing of the freedom of speech.

Reasons for wanting to receive spam:

• People do buy products from spam (scary as it is, telemarketing works too because of this)
• People might want to collect threatening messages in order to seek legal action or know nature of threat to protect themselves
• Joy of receiving e-mail each day (I admit partially to this :-P )

So What’s So Evil Here?

Sounds like I’m being nice to spammers, but I’m just being fair to the issues, as well as fair to the rights of individuals. There is still lots of wrong doings being done via spam:

• Misidentification / false identification (e-mail address, name, business name, IP address)
• Identity Theft
• Violation of Privacy
• Scams (resulting in the theft of money)
• Usage of stolen e-mail accounts
• Unaccountability (of spammer, ISP, e-mail administrator)

Why do Spammers Spam?

While we know a woodchuck woodchucks wood to sharpen its teeth, why does a spammer spam? Let’s look at the motivating factors involved:

• Greed (lust for money - when you want more money than you need)
• Financial Obligations (instead of lust, money needed instead to pay off debts, support family, etc.)
• Propaganda / Support for some idea
• Joke / Prank
• No reason whatsoever (damn irrational spammers!)
• Tricked into spamming (you ever reply to those messages about a kid with cancer who wants an e-mail before he dies? Well, most likely you fell for a spam prank to mail bomb someone)
• Mail bomb (attack someone)
• Hate
• Love (spam for love? anything’s possible when you’re in love! :-) - As for spamming to be loved, I’m afraid that won’t happen! :-P
• Sex (spam for sex? I guess its possible. A fair trade in some horny minds.)
• Loneliness (at least, that’s the punch for the “kid in the hospital” prank. But its still a valid motive if genuine)
• Desire to be famous (I suppose its possible to spam oneself into fame, but you’ll like also land yourself in jail)

Motivating factors are nice (especially for explaining dumb criminal actions), but the reason dumb and smart “criminals” spam is that SPAMMING WORKS! That is, people make money off it, or can raise awareness for an issue, cause emotional damage, or worse yet can effectively clog up someone’s e-mail account.

What is the system being analyzed?

Engineers like to deal with systems, that is: put something in a theoretical box! But what is the “box” when dealing the previously mentioned spam problems? Possiblities are:

• E-mail account on POP server
• Mail client software (Inbox) on user’s computer
• User’s attention (not sure if this can count as a system, but then again a “user” can be a system :-)

Analogous Situations

Sometimes it helps to consider similar situations and the issues that surround them. Situations that are similar to spam include:

• Junk Mail (via postal service)
• Telemarketers
• Prank Phone Calls (Is your refrigerator running? Yeah? Well go catch it!)
• Door-to-door activities (salesmen, solicitors, doorbell ringing pranksters!)
• Fan Mail (getting way too much mail - only if you are famous)
• Fame (people fighting over your attention - people possibly wasting your time)
• Advertisements (both print and online) - people fighting over your attention
• Office mailbox (stupid memos)

Balance Analysis

In chemical engineering, balances help us analyze how some quantity (such as mass or energy) moves in and out of a system, as well as where it accumulates in places. Formula-wise, a balance is:

INPUT - OUTPUT + GENERATION - CONSUMPTION = ACCUMULATION

Let’s now look at what each term means in term of Spam, and also what the sources are:

Sources of INPUT (Enters system by crossing system boundary):

• Spammer (definition: one who sends spam) sends message to your e-mail account

Sources of OUTPUT (Leaves system by crossing system boundary):

• User voluntarily sends spam (their own spam, or spam from company they agreed to allow send spam through them)
• User involuntarily sends spam (spyware/spamware, virus)

Sources of GENERATION (Appears in system without crossing system boundary):

• Users don’t generally generate spam, but they can, see #1 for output
• User decides that a regular message is now Spam (it was already in the system!)

Sources of CONSUMPTION (Leaves system without crossing system boundary):

• User deletes message.
• User’s software (spam filter) deletes message.
• User decides message is not Spam.

Locations of ACCUMULATION (Rate of amount staying in system):

• User’s POP e-mail account (server)
• User’s hard drive

Where do spammers get your e-mail address from?

High Risk (According to this CNET article [inactive link 1] [no longer available online as of 2012-09-30]) - results in many (5+ per cause) spam messages per day:

• USENET or online message boards
• online lotteries and sweepstakes
• AOL chatrooms (all they need is your screen name to spam you)
• companies that buy/sell e-mail address lists (but they need to get it from some other reason listed here first)

Medium Risk - results in about 5 spam messages per day:

• Your e-mail address on record if you register a domain name (WHOIS database); Note: I can attest that your regular postal mail address will get lots of postal junk mail too from this.
• spam bots (scour internet for mailto: addresses, especially front pages of dot coms)
• online phonebook or address book (thank you UConn)

Low Spam / No Spam:

• your ISP (not sure if this is true)
• your friends :-(
• your enemies :-( :-(
• virus
• product registration (not bad if you check/uncheck appropriate opt-out boxes; overall though, this can be argued to be solicited e-mail)
• website access registration / creating member profile on website (like to NYTimes)
• online shopping (generally can unsubscribe if present)
• by just having a free web-based e-mail account (Getting spam this way is a myth which is not true. Opening an account and never using it publicly will not generate spam on its own. See CNET article [inactive link 2] again [unfortunately, that article is no longer available online as of 2012-09-30])
• subscribing to an e-mail newsletter

Please tell me they don’t make my address up out of thin air!

Unfortuantely, as it has been pointed out to me, if your e-mail address is not very unique, chances are you’ve been hit by spam due to a method called a “dictionary attack”. For example, spammers can send spam messages to simple first name addresses at each mailserver they know of: bob@microsoft.com, bob@vgmusic.com, bob@bob.com... Other simple combinations of words and numbers can be generated in a quick batch method by spammers.

Solutions to Spam

I’m not jumping-the-gun here, but just listing possible solutions that I’ve heard through the grapevine:

Solutions Regarding Sources of INPUT (stop Spam from being sent to your e-mail address):

• Direct Action Against Spammers

  • Kill spammers (sorry, I don’t believe in the death penalty :-)
  • Take computers or Internet access away from spammers (steal [evil] or legal force [good])

• Legislation

  • Make spam illegal (I don’t think its fair to make all unsolicited mail illegal. I think someone, especially a cute girl, is welcome to e-mail me even if I never knew her before. Heaven forbid we make flirting illegal :-P )
  • Make unsolicited messages from businesses illegal (that sounds more reasonable, but will this include non-profit organizations?) Also, wouldn’t this make business-to-business e-mail offers illegal? How can we differentiate between business and personal addresses? Different first level domains in e-mail addresses! How about “dot person” (TM)? :-P My new e-mail address would be michael.newman@vgmusic.person)
  • Make spoofing e-mail origin and return addresses illegal, also spoofing IP address
  • Make illegal to deliver messages to someone whom the message isn’t addressed to (how is this possible to do in the first place?) - mailing lists would be illegal; e-mail address w/o name of addressee is illegal; your name has to be listed in message or the message is illegal; Is it too much to ask for full name OR first name OR last name OR approved nickname in body of message?

• White Lists (Approval) / Black Lists (Banned)

  • Make it so e-mail servers must all be registered through a central authority. Unregistered servers would have their DNS or IP rights terminated. Better yet, there would be an approved server list to compare Spam against. Sounds like ORBS, which got sued I think.
  • Ban e-mail from open relays (MAPS Realtime Blackhole List)

• Tests to verify sender is human, not automated spammer

  • Have people pass human-only-passable tests the first time they e-mail you (I think this what Spamcop.net uses; doesn’t work since most people get confused by it)

• Avoiding spam bots and/or e-mail address collection

  • When posting/putting an e-mail address online (like a message board), post your e-mail address using one or a combination of the follow techniques:

    • Unlinked Image[1]
    • Text Insertion: “newman@no-spam-please.vgmusic.com”
    • Punctuation Replacement: “newman at vgmusic dot com”
    • Mirror Image: “moc.cisumgv@namwen”
    • URL Context Reference: “newman@ this page’s domain name”
    • Name Context Reference: “My Last Name @vgmusic.com” - This may not be the best method since it results in the generation of an “error message” if a spammer tries using the invalid truncated address “Name @vgmusic.com”. It is best to avoid spam ever arriving to the real server. Additionally, error messages can sometimes be forwarded to the postmaster of a mail server, which is a further unforunate consequence. A solution that wastes bandwidth and other people’s time (besides the spammer’s) should be frowned upon.
    • Spaces: “n e w m a n @ v g m u s i c . c o m”
    • NATO phonetic alphabet: november echo whiskey mike alpha november at victor golf mike uniform sierra india charlie dot charlie oscar mike
    • Unicode Obfuscation: The following e-mail address is written in browser-readable Unicode: “newman@vgmusic.com”. To generate such an address easily, use the free-price program E_Cloaker. View the source code for this page to see how this works. You can also generate a linking mailto version of with Unicode: newman@vgmusic.com. While the link looks normal when rendered by a web browser, the underlying HTML code is not easily readable by e-mail address harvesting bots.

  • Don’t post messages online (that’s no fun!)

  • Only post anonymous messages online (still not fun!)

  • Post messages online that don’t require your e-mail address (ok, that’s ok)

• Using different e-mail addresses

  • E-mail addresses that expire / Disposable e-mail addresses:

    • Sneakemail.com
    • SpamMotel.com
    • Mailexpire.com [Service no longer available as of 2002-06-30]

  • Change your e-mail address every few months (not an easy option for me, plus “I’m not going to run anymore!” :-)

  • Use spam-bucket (use webmail or another e-mail account in public, but only give out private e-mail to only a select group of people)

• Quit using e-mail (sorry, I’m not the Unabomber); this may include just quitting to receive e-mail, but someone who only sends but does not receive sounds like a spammer to me!

Solutions Regarding Sources of OUTPUT (stop Spam from being sent to others FROM YOU):

• Use virus scanner / virus shield program
• Don’t use spamware software

Solutions That Utilize OUTPUT (sending, replying to, or forwarding messages to stop Spam):

• Reporting service (Spamcop.net)
• Writing your governmental representative (how does this stop Spam??) / Or sending your Spam to them.
• Consumer protection agency
• Request to be removed from a spammer’s e-mailing list. (There are tons of problems with this, including: bogus companies that don’t give a damn if you make such a request; reply addresses can be fake; you often get put on a sucker’s list if you reply to such messages; my having to exert more effort than the original time that the message wastes defeats the whole purpose. Thankfully this technique works with real companies - in theory)
• Send out fake bounced message (possible with called Bounce Spam Mail. This program is only good for messages with valid return addresses. Also, spammers can put down someone else’s return address and you might end up spamming them unknowingly!)
• Mail bomb spammers (Flood their e-mail inbox with messages). (Evil, and doesn’t work if the return address is fake or stolen - its also philosophically hypocritical)

Solutions Regarding ACCUMULATION (dealing with Spam when it has already arrived):

• Filtering

  • Filter Spam (after it arrives to your computer)

    • Custom Filter (one you design yourself). See my example of recommended filter keywords and phrases for Outlook Express [Note: This list is outdated, and not updated lately since I’ve disregarded this method.]
    • Automatic Filter (provided by other entity)
    • Distributed Filter (mutiple people working constantly together to have a dynamically updated filter: CloudMark)

  • Filter Spam (at mail server; censorship??)

  • Filter Spam at each mail router (censorship??)

  • Filtering service (Spamcop.net)

  • Have a filter that only accepts people you have pre-approved. This is a perfect solution for non-solicited e-mail, but you won’t get to hear from anyone new. :-(

  • Get a secretary! (human to filter e-mail for you; this option makes you lose privacy and subjects you to possible censoring)

  • Use artificial intelligence to filter e-mail. (not advanced enough yet)

• Manual Deletion

  • The DELETE key :-)
  • Read your subject headers only, and based on those you can personally decide to view or delete messages (pretty good method)

• Unsubscribe (on one hand it sometimes works, while on the other hand it may tell spammers your address really exists - sorta like the equivalent of a “sucker list” for postal junk mail. I think its a good method only with businesses and organizations you recognize and trust.)

• Contact your ISP about spam message (AOL keyword “tosspam”) so they can block it on their end

• Contact your State’s Attorney General (for U.S. residents) - some states have laws currently against spammers, see state list at SpamCon Foundation, example of law in Connecticut

Solutions from Outside the SYSTEM:

• Denial of service attack to offending mail servers or spammer’s IP addresses (evil, ineffective, especially since spammers can use stolen ISP accounts)
• Act of God (keep praying!)
• Class-Action Lawsuit against spammers
• Sue spammer on your own
• Bot Bait - put up tons of fake e-mail addresses on websites in order to overload spam bots

Solutions that Change the SYSTEM::

• Create new e-mail system that is not compatible with current e-mail and have it institute Spam protection measures
• Hashcash/camram - change e-mail from a “receiver-pays” to a “sender-pays” system. This method upsets some people because it introduces inefficiency into the e-mail system in order to solve yet another inefficiency problem (spam). However, I like this idea very much since there is a clear need for “a mechanism to throttle systematic abuse of un-metered internet resources”, as explained in Hashcash’s overview. Also check out camram for an explanation of how hashcash can potentially benefit e-mail.

Related Links

• Yahoo! News Full Coverage-Spam and Junk Email [link no longer works as of 2012-09-30]
• Yahoo! Computers and Internet > Communications and Networking > Email > Spam
• CNET Article: “The great CNET spam-off!!!”, 2001-07-26 [link no longer works as of 2012-09-30]

Concluding Remarks

The Best Solution So Far: I feel hashcash/camram has the most potential to alleviate the spam problem. This revelutionary idea gives e-mail value, or more importantly, associates a cost in sending out e-mail (CPU time). Used in combination with filters, it would provide a very important measure of whether a message should be considered spam. In the meantime (until hashcash/camram gets past the concept stage), I am very hopeful about the success that distributed filters like CloudMark may be able to provide. It takes my old favorite solution to spam, “Get a secretary!”, and expands its to an even higher level.

Good luck fighting spam! And happy analyzing to all! :-)

Keywords

“manifesto on spam”, “manifesto against spam”, spam, “spam problem”

About the Author

Publisher Notes

This manifesto was originally written July 15, 2001. For a decade, the article resided at http://www.vgmusic.com/~mike/an_analytical_look_at_spam.html I have moved it to this blog location to encourage discussion using the Disqus system below. Your feedback would be appreciated!

Note due to the age of this article, many external links no longer work. I have replaced as many as possible with Internet Archive copies or Wikipedia articles. Any other broken links have been marked as broken, but left unchanged.

As of September 2012, I have been using gmail as my e-mail system of choice. Their spam filtering methods get rid of most of my spam. The issues discussed above are still of great concern for people who run their own e-mail servers, and it continues to be an on-going battle even a decade after this article was written.

INTRODUCTION

Air quality dispersion modeling is used to predict ambient air concentrations of pollutants emitted by a variety of sources. To calculate the total ambient air concentration, modeled concentrations attributable to modeled local sources must be summed with the representative background pollutant concentration. The background concentration represents the contribution from all other sources not modeled, which may include distant and/or small anthropogenic sources as well as naturally occurring atmospheric concentrations of the pollutant. This paper explores methods of determining 24-hour average background concentrations for PM_2.5 that are protective of ambient air quality standards while still permitting environmentally responsible projects.

Typically, background air quality concentrations are derived from historical air quality monitoring data, which tend to be log normally distributed. In this case, the proposed source is located in Connecticut and the Connecticut Department of Environmental Protection (CTDEP) specifies that the initial PM_2.5 background concentration (i.e., the design value) be determined as the average of the yearly 98^th percentile 24-hour concentrations measured over the last three years of available data. Generally, the data are averaged for the nearest three monitoring sites as shown in Table 1. The background concentration is added to the highest eighth high (H8H) modeled concentration to determine NAAQS compliance. This approach is statistically incorrect, since log normal distributions are not additive, although some authors have suggested approximations for the summations.[1] Note that the average background concentration yields only 2 µg/m³ available for the source impact to remain below the NAAQS. Unfortunately, multisource modeling predicted H8H concentrations exceeding 9 µg/m³, and thus further analyses were required to demonstrate compliance.

Table 1. Design value monitored background concentrations:

APPROACH

The U.S. Environmental Protection Agency (EPA) “Guideline on Air Quality Models,”[2] Section 8.2, discusses the use of monitored air quality data to incorporate the background concentration into the total predicted ambient concentration.

The Guideline states:

“Use air quality data collected in the vicinity of the source to determine the background concentration for the averaging times of concern. Determine the mean background concentration at each monitor by excluding values when the source in question is impacting the monitor. The mean annual background is the average of the annual concentrations so determined at each monitor. For shorter averaging periods, the meteorological conditions accompanying the concentrations of concern should be identified. Concentrations for meteorological conditions of concern, at monitors not impacted by the source in question, should be averaged for each separate averaging time to determine the average background value. Monitoring sites inside a 90° sector downwind of the source may be used to determine the area of impact. One hour concentrations may be added and averaged to determine longer averaging periods.”

Thus, the Guideline provides an approach based on average monitored concentrations to determine the appropriate background concentrations based on concurrent meteorological conditions.

Further, the CTDEP Interim PM2.5 New Source Review Modeling Policy and Procedures[3] (August 21, 2007) states:

“CTDEP may allow an applicant to define background values that are less than the observed design values, provided the applicant provides sound technical reasoning for such an approach (e.g., a directional-specific analysis of monitored levels).”

Thus, both the Guideline and CTDEP’s Interim Procedures provide guidance to determine the appropriate background concentrations based on concurrent meteorological conditions or other “sound technical reasoning.”

The approach used to derive a representative background PM_2.5 concentration included the following steps:

• The PM_2.5 data from the three closest air quality monitoring stations were reviewed together with the concurrent meteorology.
• Air quality dispersion modeling was conducted to identify the meteorological conditions of interest that are associated with the maximum predicted PM_2.5 impacts.
• The monitored concentrations corresponding to the conditions of interest were identified and the appropriate background determined.

OBSERVED CONCENTRATIONS

Air quality dispersion modeling using off-site, National Weather Service (NWS) data is generally performed using five years of meteorological input data to obtain climatologically representative concentration predictions. In order to provide a complete view of the background air quality concentrations during the modeling period, meteorological data and monitored PM_2.5 concentrations from the three sites were reviewed for the entire modeling period, 2002-2006.

CTDEP operated the PM_2.5 monitoring stations in the region during the 2002-2006 time period. These included Federal Reference Method (FRM) monitoring as well as two sites equipped with Beta Attenuation Monitors (BAM). Table 2 summarizes the data periods and monitor types available at the CTDEP sites in the area. All of the listed monitors continued operating through the end of 2006, but some came on-line at various times during the period. The East Hartford FRM site collected daily filter samples. The other primary FRM sites, the New Haven Agricultural Station and the Waterbury site collected data on an every third day basis. A collocated FRM filter monitor at Waterbury collected data every sixth day.

Table 2. Monitoring period of record:

*Total possible daily observations 2002-2006 = 1826

Figure 1 shows concentrations from all other monitors versus the East Hartford FRM. The data track reasonably well with only two obvious significant outliers, both in the BAM data. This figure suggests that daily PM_2.5 concentrations at all the sites are dominated by regional influences, since they all track up and down together, with some scatter due to local sources and instrument noise.

Figure 1. East Hartford FRM vs. All Other Monitored Daily PM_2.5 Concentrations (µg/m³):

In order to have the most complete record of daily concentrations at each of the three monitoring locations, the data for each day at each site were combined. To generate the combined datasets, the FRM data were selected preferentially and the BAM data were “filled-in” on days when the FRM data were unavailable. In accordance with the “Interpretation of the National Ambient Air Quality Standards for PM_2.5,”[4] the two collocated FRM observations from Waterbury were averaged when both were present. This procedure resulted in at least one background concentration observation on 1,774 days out of the 1,826 days (> 97 percent) in the five year modeling period.

Figure 2 shows that the observed data for all years from all three stations combined are log-normally distributed and that the 98^th percentile of the log-normal fitted distribution is representative of the 98^th percentile of the observed data at all sites.

Figure 2. Log Probability Plot for All Stations Combined:

The closest NWS station that collects the meteorological data used for dispersion modeling is at Bradley Airport, approximately 26 miles north of the project site. A wind rose for the five years of modeled hourly meteorological data (2002-2006) from the Bradley Airport NWS station (BDL) is presented in Figure 3.

Figure 3. Bradley Airport Windrose 2002-2006:

Monitored concentrations were compared to the daily resultant wind direction, the daily average wind speed and the wind direction persistence. The most visible correspondence was between monitored concentrations and resultant wind directions. Figure 4 presents the PM_2.5 monitored concentrations for the combined data sets at the three selected sites versus the daily resultant wind direction from Bradley. The southwest wind quadrant has the most monitored concentrations that are greater than 20 µg/m³. This is indicative of transport of polluted air from urban areas to the southwest of Connecticut. The highest frequency of monitored concentrations less than 10 µg/m³ is with winds from the northern directions, reflecting cleaner air advection from the north.

Figure 4. Monitored Concentration vs. Daily Resultant Wind Direction:

PREDICTED CONCENTRATIONS

To determine predicted source contributions, EPA’s AERMOD[5] dispersion model was run. AERMOD uses hourly observed meteorological data, in this case hourly data from Bradley Airport to predict 24-hour average concentrations at an array of receptors surrounding the modeled sources.

The AERMOD model was run for the period 2002-2006 to identify the meteorological conditions of interest leading to high predicted concentrations from all NAAQS sources combined. The pattern of the predicted eighth high 24-hour average concentrations over the three consecutive worst-case years at each receptor, as well as the location of the three year average H8H predicted concentration are presented in Figure 5. The concentrations shown on this figure are the averages of the eighth highest concentrations at each receptor for the three highest consecutive years out of the five years modeled or, in other words, the isopleths for the controlling predicted concentrations for all sources in the multi-source inventory. The high AERMOD predicted 24-hour average concentrations occur on the nearest elevated terrain southeast of the project site. This direction also corresponds to high wind direction frequency as indicated by the Bradley Airport wind rose (Figure 3). Note that the highest predicted PM_2.5 concentrations are a small fraction of the highest observed concentrations.

Figure 5. Model Predicted 8^th Highest 24-hour PM_2.5 Impacts:

Figure 6 shows the pattern of the resultant 24-hour wind directions versus the highest daily predicted concentrations from all sources in the NAAQS inventory. Figure 6 shows that wind directions from approximately 300 to 360 degrees cause the highest predicted concentrations, with resulting maximum concentration impacts to the southeast or south of the proposed facility. In contrast to Figure 6, Figure 4 shows that the high observed concentrations tend to be clustered within the 180 to 225 degree wind directions, which are just the opposite of the wind directions associated with the highest predicted concentrations.

Figure 6. Maximum Daily Predicted Concentration vs. Resultant Wind Direction:

Given both the federal and state recommendations for refined background concentrations, and the demonstrated tendency for high predicted concentrations in general and the controlling concentration in particular to occur with northwest winds, the average observed 24-hour average PM_2.5 concentrations from the closest three monitoring stations have been calculated as presented in Table 4. As expected for log-normal distributions, the mean is higher than the geometric mean or the median, which are better statistical indicators of central tendency for a log–normal distribution. Thus, the mean background concentration is a conservative indicator of the background concentration.

Table 4. Background concentrations by meteorological condition:

EPA conservatively recommends using the mean concentration for the background, which is biased high in the sense that for most days the actual background will be lower than the mean. It is obvious from the data presented and discussed above that the background concentrations are wind direction dependent, thus it would be overly conservative to calculate a total potential ambient concentration based on the sum of the 98^th percentile modeled and monitored concentrations if more representative total concentrations can be calculated by summarizing the background concentrations expected during periods of maximum modeled impacts. From Table 4, the appropriate mean background concentration for the days with the maximum modeled AERMOD concentrations (which are associated with wind directions from 300 to 360 degrees) is 8 µg/m³. However, this background concentration may not be conservative for other days when the modeled concentrations are smaller since the modeled impact is only a small part of the total concentration, which is dominated by the background. Although this approach complies with EPA’s Modeling Guideline technique for determining background concentrations, CTDEP would not accept the directionally dependant mean background concentration as appropriate for the NAAQS compliance demonstration.

DAILY COMBINED OBSERVED AND PREDICTED CONCENTRATIONS

Figure 7 shows the coincidence of the highest daily AERMOD-predicted concentrations anywhere on the receptor array and the monitor-observed concentrations over all sites and years. The 98^th percentile observed and predicted concentrations are noted as vertical and horizontal lines on the graph. Predicted exceedances of the 24-hour standard (observed + predicted > 35 µg/m³) are indicated to the right of the diagonal line in the figure. Note that high predicted concentrations (those above the 98^th percentile prediction line) tend to occur with low observed concentrations and vice versa. There is a statistically significant negative correlation (r ≈ -0.31) between the observed and predicted concentrations, showing that high predicted concentrations tend to occur with low observed concentrations. This is consistent with the analyses provided above which show the meteorological conditions that lead to the greater than 98^th percentile predicted concentrations are not the same as the conditions that lead to the greater than 98^th percentile observed concentrations (e.g., north winds lead to high predicted concentrations but generally low observed concentrations, while southwest winds lead to lower predicted concentrations but frequently higher observed concentrations). Viewing Figure 7, it is clear that there are no cases when observed and predicted concentrations are both greater than the 98^th percentiles (i.e., there are no points in the upper right-hand quadrant of the figure). This null set corresponds to the generally followed compliance demonstration approach of adding together the 98^th percentile observed and the 98^th percentile predicted to determine the total concentration for comparison to the NAAQS, even though these two events are unlikely to occur together.

Figure 7. Highest Daily Predicted Concentrations vs. Observed Concentrations at All Sites (µg/m³):

If the occurrence of the predicted and observed peak concentrations is considered to be independent of one another, then the probability of their joint occurrence can be calculated as the product of the frequency of their occurrences. There is a 4 in 10,000 (0.02 X 0.02) chance that the 98^th percentile or higher predicted value will occur on the same day as the 98^th percentile or higher observed value. Actually, the probability is even less since the predicted and observed concentrations are negatively correlated. This situation corresponds to the upper right-hand region of Figure 7 (both distributions ≥ 98^th percentile) and, not surprisingly, there are no occurrences in this very low probability region. Stated differently, there is a 9,996 in 10,000 chance that the observed value corresponding to the 98^th percentile predicted value will be less than the 98^th percentile observed value.

The preceding situation is similar to the current regulatory scheme using the 98^th percentile predicted and the 98^th percentile observed concentrations to predict the total concentration. The expected number of days with concentrations greater than this value over a three-year compliance period is 0.4 days. Note that, assuming all years have the same distribution of concentrations, there could be 23 exceedance days before a violation occurs, which is a large safety margin of predicted exceedance days.

While this analysis shows that the current regulatory scheme is very conservative and that the 98^th percentile or higher observed and predicted concentrations are very unlikely to occur simultaneously, unfortunately, to the best of our knowledge, there is no way of adding two log–normal distributions together (i.e., the observed background and the model predicted concentrations) and mathematically determining the resulting distribution. Therefore, rather than combining the distributions, AERMOD modeling was conducted and the daily observed background concentrations were added to the daily predicted concentrations at each receptor to predict the total concentration on each day.

AERMOD was run to predict the impacts of all sources in the NAAQS inventory for the project and the daily background was added to the daily predicted concentrations at all receptor locations. The resulting concentration files were subsequently analyzed to find the highest three-consecutive-year average of eighth high predicted concentrations over all receptors for all modeling years. Figure 8 shows the resulting eighth highest combined daily predicted plus daily background concentrations. The isopleths are for the eighth high concentration at each receptor for consecutive three-year block averages and the maximum concentration (33.8 µg/m³) shown is the overall H8H daily predicted plus background value, which occurred in the years 2003 through 2005. Comparing this with the model predicted concentrations shown in Figure 5 (with high concentrations to the south), it becomes clear that the model predicted concentrations do not drive the total predicted concentration analysis, but that the daily background concentrations (which are highest during southwest winds) do.

Figure 8. Total Model-Predicted Plus Background 8^th Highest 24-Hour PM_2.5 Concentrations:

The relationship between the daily average background concentrations and the model predicted concentrations is illustrated by showing the time series of both. Figure 9 presents an example of both the daily average monitored concentration and the predicted concentration for 2003 at the controlling receptor location shown in Figure 8 for the combined predicted concentrations and monitored background concentrations. The monitored concentrations (shown in brown) are the daily average of the data available from all three monitoring sites. All missing monitored concentrations (52 days out of 1,826 in the modeling period) were filled with the mean monitored daily PM_2.5 concentration of 11.8 µg/m³. The contribution from all NAAQS modeled sources are shown as the AERMOD-predicted green portion of the bars. The concentrations attributable to the modeled local sources are small fractions of the total concentrations. The green dashed line on the figure is the 35 µg/m³ 24-hour PM_2.5 NAAQS concentration.

Figure 9. Time Series of 2003 Observed and Predicted Concentrations:

The overall 24-hour average H8H PM_2.5 concentration found by combining the daily predicted concentrations with the daily average background concentration is 33.8 µg/m³ based on the three-years of 2003 through 2005, as shown in Figure 8. As discussed above, the background concentration is generally low but occasionally spikes up, sometimes above the 35 µg/m³ concentration level. Examining the figure, it is clear that the EPA Modeling Guideline background approach of using the mean monitored concentrations of 8 to 12 µg/m³ as shown in Table 4 will not yield conservative total predictions.

SUMMARY

Appropriate 24-hour average PM_2.5 background concentrations have been derived for modeling analyses based on modeling guidance and concurrent meteorological, air quality modeling and air quality monitoring data. An approach was demonstrated based on an analysis of the monitored and modeled concentrations. It was shown that modeled peak concentrations do not occur simultaneously with high observed background concentrations. The unlikely chance of high background concentrations corresponding to high predicted concentrations is borne out by comparing the actual concurrence of the predicted and observed concentrations. This approach calls for combining the daily predicted PM_2.5 concentrations at all model receptors with the daily average observed concentrations from the nearest three monitoring locations. Following the addition of the predicted and observed concentrations, the resulting total concentration files are analyzed to find the average of the H8H total concentration over three consecutive years. This average H8H concentration is compared to the 24-hour PM_2.5 NAAQS to demonstrate compliance.

REFERENCES

KEYWORDS

MODELING, PM2.5, BACKGROUND, DESIGN CONCENTRATIONS

AUTHORS

Douglas R. Murray, CCM; Michael K. Anderson; QEP, Michael B. Newman, PE; Dana L. Lowes

TRC, 21 Griffin Road North, Windsor, CT 06095

PUBLISHER NOTES

Paper originally published June 24, 2008 as Paper 344 at the 101^st Annual Meeting of the Air and Waste Management Association, Portland, Oregon.

The Idea

A good friend of mine suggested to me that if two chat bots were engaged in a conversation with one another, they would end up talking forever. I was curious to find out if his theory was correct. At the time, I only knew one online chat bot (LeknorChat4) that could be instructed to contact and talk to another person through AOL Instant Messenger, and could not test the theory because I didn’t have another bot to contact. But recently, I found out about another chat bot, SmarterChild, that works through AIM, and the following results were acquired.

The Bots

• LeknorChat4 - a chat bot created by Leknor, based on Alicebot/AIML bot source/language. This bot has the unique ability to be instructed to contact someone else, as well as let you watch the resulting conversation.
• SmarterChild - Created by Activebuddy, Inc.; an information and chat bot.

Conversation #1

2001-08-27-Mon

Conversation #2

2001-08-27-Mon

Conclusions

Well, my friend was wrong. The programs stop talking to each other after a finite period of time. This happened because LeknorChat4 apparently halts conversation when it is asked the same thing twice. What a shame!

Another interesting finding was that ASCII art can be used as a way to figure out if someone/something is artificial intelligence. I would imagine this is cheating in terms of the Turing Test, but I still plan to use it in the future when I’m talking online to someone who is acting strange. This occurred in Conversation #1 at 12:46:57 PM. SmarterChild was trying to show LeknorChat4 what it looks like:

            XXXX
            X  X
       XXXXXXXXXXXXXX
       XX          XX
       XX  XX  XX  XX
       XX          XX
       XX XX    XX XX
       XX   XXXX   XX
  XX   XX          XX   XX
XXXXXXXXXXXXXXXXXXXXXXXXXXXX
X   XXXXXXXXXXX XXXXXXXX   X
XXXXXXXXX    XX    XXXXXXXXX
    XXXXXXXX XX XX XXXXX
      XXX    XX XX XXX
      XXX XX XX XX XXX
      XXX    XX    XXX
      XXXXXXXXXXXXXXXX
      XXXXXXXXXXXXXXXX
      XXX          XXX
    XXXXX          XXXXX

LeknorChat4’s conversation monitoring script however garbled the ASCII art up.

Page Created 2001-08-28-Tue

Last Updated 2012-09-18-Tue

What we should seek is knowledge and wisdom, not pure information.

We are currently living in a period that many people call the “Information Age”. This new age has been spawned by the microchip revolution and the Internet. It is important to understand what “information” actually is, and how it is lower in importance than knowledge and wisdom:

1. Data is a collection of observations, measurements, and facts. Data itself has no context, and thus has little meaning.
2. Information is data that has been given a context, and thus meaning.
3. Knowledge is the act of understanding information.
4. Wisdom is the accumulation and the resulting useful, moral application of knowledge.

It is important to understand that the Internet is chock full of data, and to a lesser extent has a quantity of information. The quality of the information (whether it is true of false) is questionable. Always be skeptical of the source! Until artificial intelligence is realized, people are the only storage place of knowledge. Ultimately, we hope to attain wisdom. In essence, engineering itself is technical wisdom!

I argue American schools are doing a disservice by filling young minds with too much information, rather than encouraging their ability to understand information (gain knowledge) and apply it wisely in totality (wisdom). Masters of trivia and memorization will serve our world poorly.

Technology is created with knowledge, but is generally used in ignorance.

Ask someone today what “technology” is, and most people will probably not be able to define it. They will however be able to give you examples of it, with the prime example being computers. The vagueness of the idea of technology underlies its current state of use in society: in ignorance. To understand technology, I argue there are 3 levels of handling knowledge:

1. Scientist - acquires knowledge [pure research]
2. Engineer - applies knowledge to solve a problem [designs technology]
3. Technologist - utilizes technology made via engineering - [can be ignorant on how thing works]

Of course, there are also Outsiders to this system, including:

• Obsoletist - doesn’t know how to use a particular technology - [can refuse to use it, or is simple unable to use in ability]
• Luddite - refuses to use technology all together
• Refuse-nik - refuses to use a particular technology
• Receptor - affected by technology used by others

Thus, technology is a process (way of doing something) or a product (physical object) that is created with scientific knowledge to solve a problem (which includes meeting a need - figuring out how to meet a need is a problem in itself).

For example, computers were created in order to calculate numbers rapidly and efficiently (meeting a need, such as that of the U.S. Census Bureau - an avid first user of computers). But technology also includes numerous other things such as light switches, buildings, and cars. As mentioned above, you are a technologist when you use technology, but you don’t need to know how the technology specifically works.

The ignorance involved is a double-edged sword: a wide range of technology can be utilized without understanding how it works, thus saving you time and freeing yourself to do other things. However, since you don’t know how the technology works, and you especially don’t know how to make it, you risk losing the technology if engineers disappear from society. Many science fiction stories have played on such a dreadful scenario. In Jack McDevitt’s Eternity Road, a post-apocalyptic world stands in awe of the remnants of old city ruins and especially of highway ruins that crisscross the world. Left-over combustion engines are just magic, unable to be reproduced in new products. Such a future is possible if the knowledge of production is left only in the hands of technical elite. For the few are easy to disappear in times of trouble.

Engineers, Managers, and Leaders are of the same breed.

With such morbid thoughts aside, let us retain the idea of the three levels. (Levels will be abbreviated here on as #1, #2, and so on.) Scientific knowledge is acquired (#1), applied (#2), and utilized (3). The beauty of this system is that reaches far beyond just science, but into any other field that solves problems or meets needs: business, health care, politics, ethics, and more. The general 3 levels in knowledge transfer theory are:

1. Ponderers / Analysts / Fundamental Researchers - job is to get knowledge
2. Designers / Problem Solvers / Controllers - WHERE THE POWER LIES :-)
3. Followers / Executors / The Ignorant - where the majority of people lie

The previous discussion has dealt with the field Science, or “Problems dealing with the physical world”. Let us see how these 3 levels apply to other fields:

Business: Problems dealing with acquiring money.

1. Economist / Analyst - acquisition of knowledge of how to make money
2. Manager / Consultant - application of knowledge to make money [designs strategies, solves problems] - “applied economics”
3. Worker - follows orders of manager or consultant [can be ignorant] [caries out strategies/orders]

Health Care: Problems dealing with the maintaining/regaining good health.

1. Research Doctors / Biologists - figure out health and recovery works.
2. Clinical Trial Doctors / Experimental Doctors / Biomedical and Biochemical Engineers - design health care practices, surgery, and medicine
3. Doctors, Nurses, Pharmacists [only know what works, don’t have to know how it works] - simply carry out what has already been figured out

Politics: Problems dealing with the gaining/retaining power (control over others).

1. Political Scientists / Sociologists / Psychologists / Think Tanks / Historians / Interest Groups / (Lobbyists?) / Political Advisers / Activists - figure out how people and nations/governments/organizations do act or should act.
2. Politicians / Leaders / Administrators / (Lobbyists?) / Monarchs / Dictator - decide policy
3. Bureaucracy / Military / Police [can be ignorant] - carry out policy

Ethics: Problems dealing with moral and ethical issues / what to believe.

1. Philosophers / Theologians - figure out basis of what we believe. - investigates foundation and expansion of doctrine
2. Pope / Religious Councils - execute, make decisions based on above. - turns doctrine into dogma
3. Clergy / Followers - carry out, simply believe [can be ignorant of basis] - follows dogma

An interesting occurrence in each field is a “bi-product” of the problem solving:

• Science: technology
• Business: strategies
• Health Care: practices
• Politics: policy
• Ethics: religious doctrine

Understanding the World through the 3 Levels

If Level #2 is where the power is, then there is likely to be fighting among the levels:

• Anarchy - group #3 does not follow group #2. Can use all benefits up to one point and time and use the benefits ignorantly for rest of time. Note this is a dissociation, not take over.
• Stagnation in Progress - group #2 does not listen to #1 anymore, dissociation. Uses all knowledge gained up to that point only.
• Coup - group #1 tries to take over group #2’s power. Especially in government situation. [I think I have this wrong, might be #3 vs #2]
• Riot - group #3 tries to take over group #2’s power. Group #3 has no clue on how to do what group #2 did, so its a mess. Similar to anarchy. But more effort involved.
• Censorship - one group tries to silence the other.
  • Example: #2 tries to silence #1. Although this generally doesn’t happen. Usually “Stagnation in Progress” happens instead. Group #2 usually likes more info from Group #1.
  • Group #3 may also not like #1 or #2’s ideas. Example, #1 may be study/believe in evolution, #3 may not like it. Example, #2 may use genetic engineering, #3 may not like it.
  • The groups get confusing when you cross science and government situations in the different fields/areas below.
  • Group #2 generally doesn’t silence group #3 because if group #3 speaks up, it generally turns into another section of group #1. Then again group #2 may silence group #3 if #3 speaks out of ignorance.
• Standard Politics - group #2 listens to group #1 (hopefully active citizens or accurate experts / policy advisers)
• Bureaucratic Politics - group #2 listens to group #3 (ignoring or in opposition of group #1)
  • Example: President enacts policy encouraged by FBI, but opposed by active citizens or interest groups (#1) - like encryption standards.
  • Example: President enacts policy encouraged by military (#3) - like missile defense shield. Bureaucratic politics need not be in conflict with group #1, since some think tanks support missile defense shield.
  • Not always bad, since group #3 deals with practicalities of real world while group #1 is theory about real world.
• Populist Politics - group #2 follows whims of Outsiders, ignoring or opposing info from #1, or experience knowledge of #3. A cause for “Stagnation in Progress” (see above).
• “George W. Bush” Politics - group #2 decides without any input of any other group in governmental field. Decides based on what he feels is right, so he borrows from other fields in a non-governmental way. Effectively is “Stagnation in Progress” while also ignoring #3 and Outsiders. May be same as Dictator. :-)
• Stability - roles #1 thru #3 stay in place.
• Class Society - same people remain in roles #1 thru #3.
• Aristocracy - same people remain in role #2.
• Status Quo - “Stagnation in Progress” to the extreme. Even accidental discoveries are not utilized.

Don’t just sit there, be something!

It is important to consider: Are roles (the 3 levels) assigned arbitrarily ??? No effort to be Outsider. Lesser effort required to be in position #3. Seems you can also be in position #1 sometimes, but may have no influence on #2 or #3. Overall, #2 may be hardest position to obtain. Smart/good people in group #3 needed, but people willing to do this are rare.

There are consequences to being in Level #2: more responsibility, especially over other people.

The Beauty of Dilbert

The comic world of Dilbert exemplifies not only Knowledge Transfer Theory in one field, but the cross interaction of different fields.

One Field Interactions:

Business #3 (Dilbert, Secretary) vs Business #2 (Pointed Hair Boss, Catbert) Note that Dogbert is Business #1 (knows how businesses work)

Cross Field Interactions:

But Dilbert also plays on different fields fighting against one another: Engineering #2 (Dilbert) vs Business #2 (Pointed Hair Boss)

What the future holds.

Conflict Between the Bane of Intellectual Property and the Joy of Privacy

Intellectual property is the ownership of ideas (hence, information), effectively a cog into the acquisition, application, and utilization of knowledge.

Privacy, like IP, is also control over information, ultimately to protect oneself or others. IP used by businesses to protect their monetary interests. Privacy is used to protect one’s safety and reputation/respect.

Privacy does not equal IP. Not all info of privacy is IP, nor is it legally owned IP.

But why is privacy held to a “purer” standard than IP? Well, because privacy is used for safety and respect and IP for money.

Today, corporations have the philosophy that more money is to be gained by violating privacy (turning private info their IP), rather than money to be made by protecting privacy (see Database Nation by Simson Garfinkel).

Control over information leads to more money...

Corporations are artificial people who live on money, while regular people need food, shelter, etc. which is acquired also by money in today’s society. Thus, people need corporations, and corporations need people... until corporations can be completely run by machines. (Yikes!) Imagine someday entire corporations run by AI and robots. Beware! :-P

The right to privacy of people is a fundamental reason why a completely open information transfer system is not right.

However, when information does not threaten the safety or reputation of a person, shouldn’t the information be open? Well, what about the reward of artists for making music for example? Is it right to share their “information”, i.e. music, without their permission? We as a society are struggling to re-address this issue today with a more open information transfer system: the Internet. Privacy is also becoming a more pertinent issue.

But all this transferring of information is trivial in important to the transfer of knowledge. What knowledge we posses is more valuable than any information we hold. Knowledge transfer systems will always deal with people (until A.I. is made). The future should be interesting...

Page Created: 2001-08-11

Page Last Updated: 2012-09-17

Keywords / Key phrases: “science”, “engineering”, “technology”

Copyright 2001 Michael B. Newman