Pollution, Concentration and Mortality

Pollution, Concentration and Mortality

by Clifford E Carnicom
Mar 19 2016

A preliminary analytical model has been developed to estimate the impact of increased concentrations of atmospheric fine particulate pollution (PM 2.5) upon mortality rates. The model is a synthesis between an analysis of measured pollution levels (PM 2.5) and published increased mortality estimates. The model is based, in part, upon previous investigations as published in the paper “The Obscuration of Health Hazards : An Analysis of EPA Air Quality Standards“, Mar 2016.

Models for both concentration levels and visibility have now been developed; for a related model in terms of visibility, please see the paper entitled Pollution, Visibility and Mortality, Mar 2016.

 mortality-concentration-days-02
Preliminary Concentration -Exposure – Mortality Model

A substantial data base based upon direct field measurements of atmospheric fine particulate matter in the southwestern United States during the winter of 2015-2016 has been acquired. The measurements reveal clear relationships between the quality of air, the PM 2.5 concentration levels, visibility of the surrounding territory, and the existence or absence of airborne aerosol operations.

The field data shows that repeated instances of the PM 2.5 count in the range between 30-60 ug/m3 is not unusual in combination with active atmospheric aerosol operations; visibility and health impacts are obvious under these conditions. The PM 2.5 count will inevitably be less than 10 (or even 5) ug/m3 under good quality air conditions.

Additional studies based upon this acquired data may be conducted in the future. Numerous published studies make known relationships between small increases in PM 2.5 pollution and increased mortality.

 meter44Measured PM 2.5 Count, 44 ug/m3.

As an example of use of this model, if the PM 2.5 count is 44 ug/m3 as shown in the above example, and if the number of days of exposure of this level is approximately 50, then the estimated increase in annual mortality is approximately 17%. This is an extreme increase in mortality, but under observed conditions in various locales it is not beyond the range of consideration.  It is thought that reasonably conservative approaches have been adopted within the modeling process.

The field data that has been collected and this model further highlight the serious deficiencies in the current Air Quality Index (AQI) as in current use by the U.S. Environmental Protection Agency (EPA). In light of the current understanding of the health impacts of small changes in PM 2.5 counts (e.g, 10 ug/m3), a scale that gives equal prominence to values as high as 500 ug/m3 (catastrophic conditions) is an incredible disservice to the public. Please see the earlier referenced papers for a more thorough discussion of the schism between public health needs and the reporting systems that are in place.

This researcher advocates the availability of direct and real-time fine particulate matter concentration levels (PM 2.5) to the public; this information should be as readily available as current weather data is.  Cost and technology are no longer major barriers to this goal.

 

operation-01Active Aerosol Operation
City of Rocks, Southern N.M.

operation-02Demonstration of the Impact of Aerosol Banks Upon Visibility.
Concentration Levels and Subsequent Visibility Changes
Directly Impact Mortality.

As an incidental note, it may be recalled from earlier work that there is a strong conceptual basis for the development and application of surveillance systems that are dependent upon atmospheric aerosol concentrations. This application is only one of many that have been proposed over a period of many years, and readers may refer to additional details on this subject within the research library. Documentaries produced by this researcher (Aerosol Crimes, Cloud Cover) during the last decade also elaborate on those analyses. The principles of LIDAR apply here.

Current field observations continue to reinforce this hypothesis. Observation in the southwest U.S. indicates that two locale types appear to be preferred targets for application: these include the large urban areas and the border region between the U.S. and Mexico. These locations, considered in a joint sense, suggest that both people and the monitoring or tracking of those same people within an area may be a technical and strategic priority of the project. A citizen based systematic and sustained nationwide monitoring system of PM 2.5 concentrations over a sufficient time period can clarify this issue further.

The recent papers on the subject of air quality are intended to raise the awareness and involvement of the public with respect to environmental and health conditions. There are very real relationships between how far you can see, the concentration levels of particulates in the atmosphere, and ultimately our mortality. It is our responsibility as stewards, as well as in our own best interest, to not deliberately and wantonly contaminate the planet.

Clifford E Carnicom
Mar 19, 2016

Pollution, Visibility and Mortality

Pollution, Visibility and Mortality
by
Clifford E Carnicom
Mar 12 2016

A preliminary empirical model has been developed to estimate the impact of diminished visibility and fine particulate pollution upon mortality rates.  The model is a synthesis between an analysis of measured pollution levels (PM 2.5), observed visibility levels and published increased mortality estimates.  The model is based, in part, upon previous investigations as published in the paper “The Obscuration of Health Hazards : An Analysis of EPA Air Quality Standards“, Mar 2016.

 

mortality-visibility-days-04

Preliminary Visibility -Exposure – Mortality Model

Air pollution has many consequences.  One of the simplest of these consequences to understand is that of mortality and the degradation of health.  It would be prudent for each of us to be aware of the sources of pollution in the atmosphere, and their subsequent effects upon our well being.  Measurement, monitoring and auditing of airborne pollution is within range of the general public, and the role of the citizens to participate in these actions is of increased imperative.  The role of public service agencies to act on behalf of public health needs and interests has not been fulfilled and we must all understand and react to the consequences of that neglect.

This particular model places the emphasis upon what can be directly observed with no special means, and that is the visibility of the surrounding sky.  Visibility levels are a direct reflection of the particulate matter that is in the atmosphere, and relations between what can be seen (or not seen, for that matter) and the concentration of pollution in the atmosphere can be established.  The relationships are observable, verifiable and are well known for their impacts upon human health, including that of mortality.

All models are idealized representations of reality.  Regardless of variations in the modeling process, it can be confidently asserted that there are direct physical relationships between particulate matter in the atmosphere, the state of visibility, and your health.   There are, of course, many other relationships of supreme importance, but the objective of this article is a simple one.  It is : to look, to be aware of your surroundings, to think, to act, and to participate. The luxuries and damage from perpetual ignorance can not be dismissed or excused.

The call for awareness is a fairly simple one here.  I encourage you to become engaged;  if for nothing else than the sake of your own health.  When this has been achieved, you are in a position of strength to help others and to improve our world.  This generation has no right or privilege to deny the depths of nature to those that will follow us.

2016-03-06_11.10.49

 

Models are one thing, real life is another.  It is time to assume your place.

Sincerely,

Clifford E Carnicom
Mar 12, 2016

The Obscuration of Health Hazards :

The Obscuration of Health Hazards:
An Analysis of EPA Air Quality Standards

by
Clifford E Carnicom
Mar 12 2016

A discrepancy between measured and observed air quality in comparison to that reported by the U.S. Environmental Protection Agency under poor conditions in real time has prompted an inquiry into the air quality standards in use by that same agency. This analysis, from the perspective of this researcher, raises important questions about the methods and reliability of the data that the public has access to, and that is used to make decisions and judgements about the surrounding air quality and its impact upon human health. The logic and rationale inherent within these same standards are now also open to further examination. The issues are important as they have a direct influence upon the perception by the public of the state of health of the environment and atmosphere. The purpose of this paper is to raise honest questions about the strategies and rationales that have been adopted and codified into our environmental regulatory systems, and to seek active participation by the public in the evaluation process.  Weaknesses in the current air quality standards will be discussed, and alternatives to the current system will be proposed.

Particulate Matter (PM) has an important effect upon human health.  Currently, there are two standards for measuring the particulate matter in the atmosphere, PM 10 and PM 2.5.  PM 10 consists of material less than 10 microns in size and is often composed of dust and smoke particles, for example.  PM 2.5 consists of materials less than 2.5 microns in size and is generally invisible to the human eye until it accumulates in sufficient quantity.  PM 2.5 material is considered to be a much greater risk to human health as it penetrates deeper into the lungs and the respiratory system.  This paper is concerned solely with PM 2.5 pollution.

As an introduction to the inquiry, curiosity can certainly be called to attention with the following statement by the EPA in 2012, as taken from a document (U.S. Environmental Protection Agency 2012,1) that outlines certain changes made relatively recently to air quality standards:

“EPA has issued a number of rules that will make significant strides toward reducing fine particle pollution (PM 2.5). These rules will help the vast majority of U.S. counties meet the revised PM 2.5 standard without taking additional action to reduce emissions.”

Knowing and studying the “rule changes” in detail may serve to clarify this statement, but on the surface it certainly conveys the impression of a scenario whereby a teacher changes the mood in the classroom by letting the students know that more of them will be passing the next test.  Even better, they won’t need to study any harder and they will still get the same result.

In contrast, the World Health Organization (WHO) is a little more direct (World Health Organization 2013, 10) about the severity and impact of fine particle pollution (PM 2.5):

“There is no evidence of a safe level of exposure or a threshold below which no adverse health effects occur. The exposure is ubiquitous and involuntary, increasing the significance of this determinant of health.”

We can, therefore, see that there are already significant differences in the interpretation of the impact of fine particle pollution (especially from an international perspective), and that the U.S. EPA is not exactly setting a progressive example toward improvement.

Another topic of introductory importance is that of the AQI, or “Air Quality Index” that has been adopted by the EPA (“Air Quality Index – Wikipedia, the Free Encyclopedia” 2016).  This index is of the “idiot light” or traffic light style, where green means all is fine, yellow is to exercise caution, and red means that we have a problem.  The index, therefore, has the following appearance:

2016-02-02_11.42.35
There are other countries that use a similar type of index and color-coded scheme.  China, for example, uses the following scale (“Air Quality Index – Wikipedia, the Free Encyclopedia” 2016):

2016-02-02_11.51.45

As we continue to examine these scale variations, it will also be of interest to note that China is known to have some of the most polluted air in the world, especially over many of the urban areas.

Not all countries, jurisdictions or entities , however, use the idiot light approach that employs an arbitrary scaling method that is removed from showing the actual PM 2.5 pollution concentrations, such as those shown from the United States and China above.  For example, the United Kingdom uses a scale (“Air Quality Index – Wikipedia, the Free Encyclopedia” 2016) that is dependent upon actual PM 2.5 concentrations, as is shown below:

2016-02-02_12.04.02
Notice that the PM 2.5 concentration for the U.K. index is directly accessible and that the scaling for the index is dramatically different than that for the U.S. or China.  In the case of the AQI used by the U.S. and China (and other countries as well), a transformed scale runs from 0 to 300-500 with concentration levels that are generally more obscure and ambiguous within the index.  In the case of the U.K index, the scale directly reports with a specific PM 2.5 concentration level with a maximum (i.e., ~70 ug/m^3) that is far below that incorporated into the AQI index (i.e., 300 – 500 ug/m^3).

We can be assured that if a reading of 500 ug/m^3 is ever before us, we have a much bigger problem on our hands than discussions of air quality.  The EPA AQI is heavily biased toward extreme concentration levels that are seldom likely to occur in practical affairs; the U.K. index gives much greater weight to the lower concentration levels that are known to directly impact health, as reflected by the WHO statement above.

Major differences in the scaling of the indices, as well as their associated health effects, are therefore hidden within the various color schemes that have been adopted by various countries or jurisdictions.  Color has an immediate impact upon perception and communication; the reality is that most people will seldom, if ever, explore the basis of such a system as long as the message is “green” under most circumstances that they are presented with.  The fact that one system acknowledges serious health effects at a concentration level of  50 – 70 ug/m^3 and that another does not do so until the concentration level is on the order of 150 – 300 ug/m^3 is certainly lost to the common citizen, especially when the scalings and color schemes chosen obscure the real risks that are present at low concentrations.

The EPA AQI system appears to have its roots in history as opposed to simplicity and directness in describing the pollution levels of the atmosphere, especially as it relates to the real-time known health effects of even short-term exposure to lower concentration PM 2.5 levels.  The following statement (“Air Quality Index | World Public Library” 2016) acknowledges weaknesses in the AQI since its introduction in 1968, but the methods are nevertheless perpetuated for more than 45 years.

“While the methodology was designed to be robust, the practical application for all metropolitan areas proved to be inconsistent due to the paucity of ambient air quality monitoring data, lack of agreement on weighting factors, and non-uniformity of air quality standards across geographical and political boundaries. Despite these issues, the publication of lists ranking metropolitan areas achieved the public policy objectives and led to the future development of improved indices and their routine application.”


The system of color coding to extreme and rarified levels with the use of an averaged and biased scale versus one that directly reports the PM 2.5 concentration levels in real time is an artifact that is divorced from current observed measurements and the knowledge of the impact of fine particulates upon human health.

The reporting of PM 2.5 concentrations directly along with a more realistic assessment of impact upon human health is hardly unique to the U.K. index system. With little more than casual research, at least three other independent systems of measurement have been identified that mirror the U.K. maximum scaling levels along with the commensurate PM 2.5 counts. These include the World Health Organization, a European environmental monitoring agency, and a professional metering company index scale (World Health Organization 2013, 10) (“Air Quality Now – About US – Indices Definition” 2016) (“HHTP21 Air Quality Meter, User Manual, Omega Engineering” 2016, 10).
.

As another example to gain perspective between extremes and maximum “safe” levels of PM 2.5 concentrations, we can recall an event that occurred in Beijing, China during November 2010, and that was reported by the New York Times in January of 2013 (Wong 2013) .  During this extreme situation, the U.S. Embassy monitoring equipment registered a PM 2.5 reading of 755, and the story certainly made news as the levels blew out any scale imaginable, including those that set maximums at 500.

An after statement within the article that references the World Health Organization standards may be the lasting impression that we should carry forward from the horrendous event, where it is stated that:

“The World Health Organization has standards that judge a score above 500 to be more than 20 times the level of particulate matter in the air deemed safe.”

Not withstanding the fact that WHO also states that no there is no evidence of any truly “safe” level of particulate matter in the atmosphere, we can nevertheless back out of this statement that a maximum “safe” level for the PM 2.5 count, as assessed by WHO, is approximately 25 ug / m^3.  This statement alone should convince us that we must pay close attention to the lower levels of pollution that enter into the atmosphere, and that public perception should not be distorted by scales and color schemes that usually only affect public perception when they number into the hundreds.

Let us gain a further understanding of how low concentration levels and small changes affect human health and, shall I daresay, mortality. The case for low PM 2.5 concentrations being seriously detrimental to human health is strong and easy to make.  Casual research on the subject will uncover a host of research papers that quantify increased mortality rates with direct relationship to small changes in PM 2.5 concentrations, usually expressing a change in mortality per 10 ug / m^3.  Such papers are not operating in the arena of scores to hundreds of micrograms per cubic meter, but on the order of TEN micrograms per cubic meter.  This work underscores the need to update the air quality standards, methods and reporting to the public based upon current health knowledge, instead of continuing a system of artifacts based upon decades old postulations.

These papers will refer to both daily mortality levels as well as long term mortality based upon these “small” increases in PM 2.5 concentrations.  The numbers are significant from a public health perspective.  As a representative article, consider the following recent published paper in Environmental Health Perspectives in June of 2015, under the auspices of the National Institute of Environmental Health Sciences(Shi et al. 2015) :

 

2016-02-04_16.52.40

 

with the following conclusions:

 

2016-02-04_16.54.29

 

as based upon the following results:

 

2016-02-04_16.55.04

 

Let us therefore assume a more conservative increase of 2% mortality for a short-term exposure (i.e., 2 day) per TEN (not 12, not 100, not 500 per AQI scaling) micrograms per cubic meter.  Let us assume a mortality increase of 7% for long term exposure (i.e, 365 days).

Let us put these results into further perspective.  A sensible question to ask is, given a certain level of fine particulate pollution introduced into the air for a certain number of days within the year, how many people would die as a consequence of this change in our environment?  We must understand that the physical nature of the particulates is being ignored here (e.g., toxicity, solubility, etc.) other than that of the size being less than 2.5 microns.

The data results suggest a logarithmic form of influence, i.e. a relatively large effect for short term exposures, and a subsequently more gradual impact for long term exposure.  A linear model is the simplest approach, but it also is likely to be too modest in modeling the mortality impact. For the purpose of this inquiry, a combined linear-log approach will be taken as a reasonably conservative approach.

The model developed, therefore, is of the form:

Mortality % Increase (per 10ug/m^3) = 1.65 +. 007(days) + 0.48 * ln(days)

The next step is to choose the activity level and time period for which we wish to model the mortality increase.  Although any scenario within the data range could be chosen, a reasonably conservative approach will also be adopted here.  The scenario chosen will be to introduce 30 ug/m^3 of fine particulate matter into the air for 10% of the days within a year.

The model will therefore estimate a 3.6% increase in mortality for 10 ug/ m^3 of introduced PM 2.5 materials (36.5 days).  For 30 ug/m^3, we will therefore have a a 10.9% increase in mortality.  As we can see, the numbers can quickly become significant, even with relatively low or modest PM 2.5 increases in pollution.

Next we transform this percentage into real numbers. During the year of 2013, the Centers for Disease Control (CDC) reports that 2,596,993 people died during that year from all causes combined (“FastStats” 2016).  The percentage of 10.9% increase applied to this number results in 283, 072 additional projected deaths per year.

Continuing to place this number into perspective, this number exceeds the number of deaths that result from stroke, Alzheimer’s, and influenza and pneumonia combined (i.e, 5th, 6th, and 8th leading causes of death) during that same year.  The number is also much higher than the death toll for Chronic Pulmonary Obstructive Disease (COPD), which is now curiously the third leading cause of death.

We should now understand that PM 2.5 pollution levels are a very real concern with respect to public health, even at relatively modest levels.  Some individuals might argue that such a scenario could never occur, as the EPA has diminished the PM 2.5 standard on an annual basis down to 12 ug/m^3.  The enforcement and sensitivity of that measurement standard is another discussion that will be reserved for a later date.  Suffice it to say that the scenario chosen here is not unduly unrealistic here for consideration, and that it is in the public’s interest to engage themselves in this discussion and examination.

 


 

The next issue of interest to discuss is that of a comparison between different air quality scales in some detail.  In particular, the “weighting”, or influence, of lower concentration levels vs. higher concentration levels will be examined.  This topic is important because it affects the interpretation by the public of the state of air quality, and it is essential that the impacts upon human health are represented equitably and with forthrightness.

The explanation of this topic will be considerably more detailed and complex than the former issues of “color coding” and mortality potentials, but it is no less important.  The results are at the heart of the perception of the quality of the air by the public and its subsequent impact upon human health.

To compare different scales of air quality that have been developed; we must first equate them.  For example, if one scale ranges from 1 to 6, and another from 0 to 10, we must “map”, or transform them such that the scales are of equivalent range.  Another need in the evaluation of any scale is to look at the distribution of concentration levels within that same scale, and to compare this on an equal footing as well.  Let us get started with an important comparison between the EPA AQI and alternative scales that deserve equal consideration in the representation of air quality.

Here is the structure of the EPA AQI in more detail (U.S. Environmental Protection Agency 2012, 4) .

 

 AQI Index AQI Abitrary Numeric  AQI Rank PM 2.5 (ug/m^3) 24 hr avg.
Good  0-50  1  0-12
Moderate  51-100  2  12.1-35.4
Unhealthy for Sensitive Groups  101-150  3  35.5-55.4
Unhealthy  151-200  4  55.5-150.4
Very Unhealthy  201-300  5  150.5-250.4
Hazardous  301-500  6  250.5-500

 

Now let us become familiar with three alternative scaling and health assessment scales that are readily available and that acknowledge the impact of lower PM 2.5 concentrations to human health:

 

United Kingdom Index U.K. Nomenclature PM 2.5 ug/m3 24 hr avg.
1 Low 0-11
2 Low 12-23
3 Low 24-35
4 Moderate 36-41
5 Moderate 41-47
6 Moderate 48-53
7 High 54-58
8 High 59-64
9 High 65-70
10 Very High >=71

 

Now for a second alternative air quality scale, this being from Air Quality Now, a European monitoring entity:

 

Air Quality Now EU Rank Nomenclature PM 2.5  Hr PM 2.5 24 Hrs.
1 Very Low 0-15 0-10
2 Low 15-30 10-20
3 Medium 30-55 20-30
4 High 55-110 30-60
5 Very High >110 >60

 

And lastly, the scale from a professional air quality meter manufacturer:

 

Professional Meter Index Nomenclature PM 2.5 ug/m^3 Real Time Concentration
0 Very Good 0-7
1 Good 8-12
2 Moderate 13-20
3 Moderate 21-31
4 Moderate 32-46
5 Poor 47-50
6 Poor 52-71
7 Poor 72-79
8 Poor 73-89
9 Very Poor >90

 

We can see that the only true common denominator between all scaling systems is the PM 2.5 concentration.  Even with the acceptance of that reference, there remains the issue of “averaging” a value, or acquiring maximum or real time values.  Setting aside the issue of time weighting as a separate discussion, the most practical means to equate the scaling system is to do what is mentioned earlier:  First, equate the scales to a common index range (in this case, the EPA AQI range of 1 to 6 will be adopted).  Second, inspect the PM 2.5 concentrations from the standpoint of distribution, i.e., evaluate these indices as a function of PM 2.5 concentrations.  The results of this comparison follow below, accepting the midpoint of each PM 2.5 concentration band as the reference point:

PM 2.5 (ug/m^3) EPA AQI UK EU (1hr) Meter
1-10 1 1 1 1
10-20 2 1.6 1 2.1
20-30 2 2.1 2.2 2.7
30-40 2 2.1 3.5 3.2
40-50 3 3.2 3.5 3.2
50-60 3 4.3 3.5 4.3
60-80 4 5.4 4.8 4.9
80-100 4 6 4.8 6
100-150 4 6 6 6
150-200 4 6 6 6
200-250 5 6 6 6
250-300 5 6 6 6
300-400 6 6 6 6
400-500 6 6 6 6

 

This table reveals the essence of the problem; the skew of the EPA AQI index toward high concentrations that diminishes awareness of the health impacts from lower concentrations can be seen within the tabulation. 

This same conclusion will be demonstrated graphically at a later point.

Now that all air quality scales are referenced to a common standard, i.e., the PM 2.5 concentration), the general nature of each series can be examined via a regression analysis.  It will be found that a logistical function is a favored functional form in this case and the results of that analysis are as follows:

EPA Index (1-6) = 5.57 / (1 + 2.30 * exp(-.016 * PM 2.5))
Mean Square Error = 0.27

Mean (UK – EU – Meter) Index (1-6) = 6.03 / (1 + 5.65 * exp(-.046 * PM 2.5))
Mean Square Error = 0.01

The information that will now be of value to evaluate the weighting distribution applied to various concentration levels is that of integration of the logistical regression curves as a function of bandwidth.  The result of the integration process (Int.) applied to the above regressions is as follows:

PM 2.5 Band EPA AQI (Int.)
[Index * PM 2.5]
Mean Index (Int.)
(UK-EU-Meter)
[Index * PM 2.5]
% Relative Overweight or Underweight of PM 2.5 Band Contribution Between EPA AQI and Mean Alternative Air Quality Index Scale (Endpoint Bias Removed)
1-10 16.1 10.1 +42%
10-20 19.8 15.8 +27%
20-30 21.9 21.6 +8%
30-40 24.1 28.3 -10%
40-50 26.3 35.2 -27%
50-60 28.5 41.5 -39%
60-80 63.6 98.0 -47%
80-100 72.1 110.4 -46%
100-150 211.7 295.0 -32%
150-200 243.7 300.8 -16%
200-250 261.7 301.4 -8%
250-300 270.7 301.5 -4%
300-400 551.8 603.0 -2%
400-500 555.9 603.0 0%

 

A graph of a regression curve to the % Relative Overweight/Underweight data in the final column of the table above is as follows (band interval midpoints selected; standard error = 4.1%).

 

EPA Underweight Function Feb 09 2016 - 01

 

And, thus, we are led to another interpretation regarding the demerits of the EPA AQI.  The EPA AQI scaling system unjustifiably under-weights the harmful effects of PM 2.5 concentrations that are most likely to occur in real world, real time, daily circumstances.  The scale over-weights the impacts of extremely low concentrations that have little to no impact upon human health.  And lastly, when the PM 2.5 concentrations are at catastrophic levels and the viability of life itself is threatened, all monitoring sources, including the EPA, are in agreement that we have a serious situation.  One must seriously question the public service value under such distorted and disproportionate representation of this important monitor of human health, the PM 2.5 concentration.

 


 

Let us proceed to an additional serious flaw in the EPA air quality standards, and this is the issue of averaging the data. It will be noticed that the current standard for EPA PM 2.5 air quality is 12 ug/m^3 , as averaged over a 24 hour period. On the surface, this value appears to be reasonably sound, cautious and protective of human health. A significant problem, however, occurs when we understand that the value is averaged over a period of time, and is not reflective of real-time dynamic conditions that involve “short-term” exposures.

To begin to understand the nature of the problem, let us present two different scenarios:

Scenario One:

In the first scenario, the PM 2.5 count in the environment is perfectly even and smooth, let us say at 10 ug/m^3. This is comfortably within the EPA air quality standard “maximum” per a 24 hour period, and all appears well and good.

Scenario Two:

In this scenario, the PM 2.5 count is 6 ug/m^3 for 23 hours out of 24 hours a day. For one hour per day, however, the PM 2.5 count rises to 100 ug/m^3, and then settles down back to 6 ug/m^3 in the following hour.

Instinctively, most of us will realize that the second scenario poses a significant health risk, as we understand that maximum values may be as important (or even more important) than an average value. One could equate this to a dosage of radiation, for example, where a short term exposure could produce a lethal result, but an average value over a sufficiently long time period might persuade us that everything is fine.

And this, therefore, poses the problem that is before us.

In the first scenario, the weighted average PM 2.5 count over a 24 hour period is 10 ug/ m^3.

In the second scenario, the weighted average PM 2.5 count over a 24 hour period is 10 ug/m^3.

Both scenario averages are within the current EPA air quality maximum pollution standards.

Clearly, this method has the potential for disguising significant threats to human health if “short-term” exposures occur on any regular basis. Observation and measurement will show that they do.

Now that we have seen some of the weaknesses of the averaging methods, let us look at an additional scenario based upon more realistic data, but that continues to show a measurable influence upon human health. The scenario selected has a basis in recent and independently monitored PM 2.5 data.

The situation in this case is as follows:

This model scenario will postulate that the following conditions are occurring for approximately 10% of the days in a year. For that period, let us assume that for 13.5 hours of the day that the PM 2.5 count is essentially nil at 2 ug/m^3. For the remaining 10.5 hours of the day during that same 10% of the year, let us assume the average PM 2.5 count is 20 ug/m^3. The range of the PM 2.5 count during the 10.5 hour period is from 2 to 60 ug/m^3, but the average of 20 ug/m^3 (representing a significant increase) will be the value required for the analysis. For the remainder of the year very clean air will be assumed at a level of 2 ug/m^3 for all hours of the day.

A more extended discussion of the nature of this data is anticipated at a later date, but suffice it to say that the energy of sunlight is the primary driver for the difference in the PM 2.5 levels throughout the day.

The next step in the problem is to determine the number of full days that correspond to the concentration level of 20 ug/m^3, and also to provide for the fact that the elevated levels will be presumed to exist for only 10% of the year.  The value that results is:

0.10 * (365 days) * (10.5 hrs / 24 hrs) = 16 full days of 20 ug/m^3 concentration level.

As a reference point, we can now estimate the increase in mortality that will result for an arbitrary 10 ug/m^3 (based upon the relationship derived earlier):

Mortality % Increase (per 10ug/m^3) = 1.65 +. 007(16 days) + 0.48 * ln(16 days)

and

Mortality % Increase (per 10ug/m^3) = 3.1%

The increase in this case is 18 ug/m^3 (20 ug/m^2 – 2 ug/m^3), however, and the mortality increase to be expected is therefore:

Mortality % Increase (per 18ug/m^3 increase) = 1.8 * 3.1% = 5.6%.

Once again, to place this number into perspective, we translate this percentage into projected deaths (as based upon CDC data, 2013):

.056 * (2, 596, 993) = 145, 431 projected additional deaths.

This value is essentially equivalent (again, curiously) to the third leading cause of death, namely Chronic Pulmonary Obstructive Disease (COPD), with a reported value of deaths for 2013 of 149, 205.

It is understood that a variety of factors will ultimately lead to mortality rates, however, this value may help to put the significance of  “lower” or “short-term” exposures to PM 2.5 pollution into perspective.

It should also be recalled that the averaging of PM 2.5 data over a 24 hour period can significantly mask the influences of such “short-term” exposures.

A remaining issue of concern with respect to AQI deficiencies is its accuracy in reflecting real world conditions in a real-time sense. The weakness in averaging data has already been discussed to some extent, but the issue in this case is of a more practical nature. Independent monitoring of PM 2.5 data over a reasonably broad geographic area has produced direct visible and measurable conflicts in the reported state of air quality by the EPA.

After close to twenty years of public research and investigation, there is no rational denial that the citizenry is subject to intensive aerosol operations on a regular and frequent basis. These operations are conducted without the consent of that same public. The resulting contamination and pollution of the atmosphere is harmful to human health.  The objective here is to simply document the changes in air quality that result from such a typical operation, and the corresponding public reporting of air quality by the EPA for that same time and location.

Multiple occasions of this activity are certainly open to further examination, but a representative case will be presented here in order to disclose the concern.

 

clear_01

Typical Conditions for Non- Operational Day.
Sonoran National Monument – Stanfield AZ

op_01

Aerosol Operation – Early Hours
Jan 19 2016 – Sonoran National Monument – Stanfield AZ

op_02

Aerosol Operation – Mid-Day Hours
Jan 19 2016 – Sonoran National Monument – Stanfield AZ

 

op_day-crop

EPA Website Report at Location and Time of Aerosol Operation.
Jan 19 2016 – Sonoran National Monument – Stanfield AZ
Air Quality Index : Good
Forecast Air Quality Index : Good
Health Message : None

Current Conditions : Not Available
(“AirNow” 2016)

 

The PM 2.5 measurements that correlate with the above photographs are as follows:

With respect to the non-operational day photograph, clean air can and does exist at times in this country, especially in the more remote portions of the southwestern U.S. under investigation.  It is quite typical to have PM 2.5 counts from 2 to 5 ug/m^3, which fall under the category of very good air quality by any index used.  Low PM 2.5 counts are especially prone to occur after periods of heavier rain, as the materials are purged from the atmosphere.  The El Nino influence has been especially influential in this regard during the earlier portion of this winter season.  Visibility conditions of the air are a direct reflection of the PM 2.5 count.

On the day of the aerosol operation, the PM 2.5 counts were not low and the visibility down to ground level was highly diminished.  The range of values throughout the day were from 2 to 57, with the low value occurring prior to sunrise and post sundown.  The highest value of 57 occurred during mid-afternoon.  A PM 2.5 value of 57 ug/m^3 is considered poor air quality by many alternative and contemporary air quality standards, and the prior discussions on mortality rates for “lower” concentrations should be consulted above.  This high value has no corollary, thus far, during non-aerosol-operational days.  From a common sense point of view, the conditions recorded by both photograph and measurement were indeed unhealthy.  Visibility was diminished from a typical 70 miles + in the region to a level of approximately 30 miles during the operational period.  Please refer to the earlier papers (Visibility Standards Changed, March 2001 and Mortality vs. Visibility, June 2004; also additional papers) for additional discussions related to these topics.

The U.S. Environmental Protection Agency reports no concerns, no immediate impact, nor any potential impact to health or the environment during the aerosol operation at the nearest reporting location.

 


Summary:

This paper has reviewed several factors that affect the interpretation of the Air Quality Index (AQI) as it has been developed and is used by the U.S. Environmental Protection Agency (EPA). In the process, several shortcomings have been identified:

1. The use of a color scheme strongly affects the perception of the index by the public. The colors used in the AQI are not consistent with what is now known about the impact of fine particulate matter (PM 2.5) to human health. The World Health Organization (WHO) acknowledges that there are NO known safe levels of fine particulate matter, and the literature also acknowledges the serious impact of low concentration levels of PM 2.5, including increased mortality.

2. The scaling range adopted by the AQI is much too large to adequately reveal the impact of the lower concentration levels of PM 2.5 to human health. A range of 500 ug/m^3 attached to the scale when mortality studies acknowledge significant impact at a level of 10 ug/m^3 is out of step with current needs by the public.

3. The underweighting of the lower PM 2.5 concentration levels relative to more contemporary scales that adequately emphasize lower level health impacts obscures health impacts which deserve more prominent exposure.

4. The AQI numeric scale is divorced from actual PM 2.5 concentration levels. The arbitrary scaling has no direct relationship to existing and actual concentrations of mass to volume ratios. The actual conditions of pollution are therefore hidden by an arbitrary construct that obscures the impact of pollution to human health.

5. The AQI is a historic development that has been maintained in various incarnations and modifications since its origin more than 45 years ago. The method of presentation and computation is obtuse and appears to exist as a legacy to the past rather than directly portraying pollution health risks.

6. The averaging of pollution data over a time period that filters out short term exposures of high magnitude is unnecessary and it hinders the awareness of the actual conditions of exposure to the public.

7. Presentation of air quality information through the authorized portal appears to present potential conflicts between reported information and actual field condition observation, data and measurement.

Recommendations:

In the opinion of this researcher the AQI, as it exists, should be revamped or discarded. Allowing for catastrophic pollution in the development of the scale is commendable, but not if it interferes with the presentation of useful and valuable information to the public on a practical and daily basis.

There is a partial analogy here with the scales used to report earthquakes and other natural events, as they are of an exponential nature and they provide for extreme events when they occur. It is now known, however, that very low levels of fine particulate matter are very harmful to human health. Any scaling chosen to represent the state of pollution in the atmosphere must correspondingly emphasize and reveal this fact. This is what matters on a daily basis in the practical affairs of living; the extreme events are known to occur but they should not receive equal (or even greater) emphasis in a daily pollution reporting standard. It is primarily a question of communicating to the public directly in real-time with actual data, versus the adherence to decades old legacies and methods that do not accurately portray modern pollution and its sources.

It seems to me that a solution to the problem is fairly straightforward; this issue is whether or not such a transformation can be made on a national level and whether or not it has strong public support. Many other scaling systems have already made the switch to emphasize the impact of lower level concentrations to human health; this would seem to be admirable based upon the actual needs of society.

It is a fairly simple matter to reconstruct the scale for an air quality index. THE SIMPLEST SOLUTION IS TO REPORT THE CONCENTRATION LEVELS DIRECTLY, IN REAL TIME MODE. For example, if the PM 2.5 pollution level at a particular location is, for example, 20 ug/m^3, then report it as such. This is not hard to do and technology is fully supportive of this direct change and access to data. We do not average our rain when it rains, we do not average our sunlight when we report how clear the sky is, we do not average the cloud cover, and we do not average how far we can see. The environmental conditions exist as they are, and they should be reported as such. There is no need to manipulate or “transform” the data, as is being done now. A linear scale can also be matched fairly well to the majority of daily life needs, and the extreme ranges can also be accommodated without any severe distortion of the system. The relationship between visibility and PM 2.5 counts will be very quickly and readily assimilated by the public when the actual data is simply available in real-time mode as it needs to be and should be. Of course, greater awareness of the public of the actual conditions of pollution may also lead to a stronger investigation of their source and nature; this may or may not be as welcome in our modern society. I hope that it will be, as the health of our generation, succeeding generations, and of the planet itself is dependent upon our willingness to confront the truths of our own existence.

Clifford E Carnicom
Mar 12, 2016

Born Clifford Bruce Stewart
Jan 19, 1953

 

References

“AirNow.” 2016. Accessed March 13. https://www.airnow.gov/.

“Air Quality Index | World Public Library.” 2016. Accessed March 13. http://www.worldlibrary.org/articles/air_quality_index.

“Air Quality Index – Wikipedia, the Free Encyclopedia.” 2016. Accessed March 13. https://en.wikipedia.org/wiki/Air_quality_index.

“Air Quality Now – About US – Indices Definition.” 2016a. Accessed March 13. http://www.airqualitynow.eu/about_indices_definition.php.
———. 2016b. Accessed March 13. http://www.airqualitynow.eu/about_indices_definition.php.

“FastStats.” 2016. Accessed March 13. http://www.cdc.gov/nchs/fastats/deaths.htm.

“HHTP21 Air Quality Meter, User Manual, Omega Engineering.” 2016.

Shi, Liuhua, Antonella Zanobetti, Itai Kloog, Brent A. Coull, Petros Koutrakis, Steven J. Melly, and Joel D. Schwartz. 2015. “Low-Concentration PM2.5 and Mortality: Estimating Acute and Chronic Effects in a Population-Based Study.” Environmental Health Perspectives 124 (1). doi:10.1289/ehp.1409111.

U.S. Environmental Protection Agency. 2012. “Revised Air Quality Standards for Particle Pollution and Updates to the Air Quality Index (AQI).”

Wong, Edward. 2013. “Beijing Air Pollution Off the Charts.” The New York Times, January 12. http://www.nytimes.com/2013/01/13/science/earth/beijing-air-pollution-off-the-charts.html.

World Health Organization. 2013. “Health Effects of Particulate Matter, Policy Implications for Countries in Eastern Europe, Caucasus and Central Asia.”

MORTALITY REQUIRES EXAMINATION

MORTALITY REQUIRES EXAMINATION
Clifford E Carnicom
Mar 22 2004
Edited Jan 16 2006

The following information has been provided by a citizen as a basis for further inquiry and examination:

“Dear Clifford,

I have been very curious as to whether or not the chemtrails have increased the death rate. I found it very difficult to get any official figures to compare from different years so I decided to do an informal check using the classified death listing archives from my local newspaper.

I was shocked at what I found. These figures are truly alarming. I compared the number of death notices for the same two months (Jan and Feb) going back for approx ten years (to compare pre chemtrails years with chemtrails years). This is what I found:

Year

Death Notice Totals for Jan-Feb:

1995

191

1996

134

1997

105

1998

98

1999

144

2000

196

2001

1680

2002

1734

2003

1728

2004

2000

As you can see the totals for 2003 are almost 10 times higher than the totals for 1995. This seems way out of line and very alarming even taking into consideration an aging population. Perhaps there is another reason for these figures but I suspect it is a result of the chemtrails and what we are seeing may be the result of a deliberate depopulation campaign. These figures are from the San Francisco Chronicle death notice archives. For accuracy the actual newspaper archives should be doublechecked aginst the online archives because death notices that were originally listed in the newspaper are not necessarily included in the online archives. This is only done with the families permission.

Perhaps you could ask people in other communities to run this check to see what figures they come up with.

I appreciate your efforts on behalf of the greater community.

Best Regards,”

Anonymous by Request

DRASTIC pH CHANGES

DRASTIC pH CHANGES
Clifford E Carnicom
September 24 2000


1. The most significant chemical species in the clouds and
precipitation is the hydrogen ion (or hydroxide ion, correspondingly)
concentration, as measured by the pH, according to the 1995 Nobel
Prize winner for chemistry, Paul J. Crutzen, Director of Air
Chemistry Division of the Max Planck Institut.

2. The magnitude of recently measured pH values of rainfall across
the country shows a twenty fold increase in the number of hydroxide
ions in the year 2000 vs. both 1990 and 1999 baseline data. This
translates directly to a major change in pH and atmospheric chemistry
during the recent year.

3. A statistical Student’s t test applied to the year 2000 measured
differences in rainfall pH is statistically significant at the 99.9%+
level.

4. A Wilcoxon’s Signed Rank non-parametric statistical test, which
makes no assumptions about the underlying distribution of the data
(normal or otherwise), shows a statistically significant difference
in the atmospheric chemistry of the year 2000 pH data at the
99.9999%+ level.

5. A 95% confidence interval for the average 2000 pH change relative
to 1999 data indicates the average 2000 pH difference is expected to
fall between +1.0 and +1.7. This corresponds to a 10 to 50 times
increase in the hydroxide ion concentration in the atmosphere,
occurring primarily within a twelve month period.

6. The atmospheric changes are correlated directly with the presence
of sustained and extensive aircraft aerosol operations since the
beginning of 1999.

7. These drastic changes and the results of these studies
demonstrate the urgent need for a formal investigation into recent
and radical changes in the atmospheric chemistry of the nation and
globe. Citizens across the country are urged to organize and to
demand this investigation without delay.

Clifford E Carnicom
September 24 2000
Santa Fe, NM
Authored at Rio Chama, NM

RAINFALL pH TEST REPORTS

RAINFALL
pH TEST REPORTS
Measurements taken by involved citizens across the country.
Posted by Clifford E Carnicom
September 2000

1990 Difference Statistics:
Number of Observations: 87
Average of Differences : 1.41
Sample Standard Deviation of Differences: 0.72
t Statistic: 18.3
Significance Level: 99.999%+

1999 Difference Statistics:
Number of Observations: 87
Average of Differences : 1.37
Sample Standard Deviation of Differences: 0.72
t Statistic: 17.7
Significance Level: 99.999%+

Wilcoxon’s Signed Rank Non-Parametric Test also indicates the pH differences from 2000 with respect to 1999 data to be significant at the 99.9999%+ level. (n=24)

Significant differences from the baseline indicate significant changes in atmospheric chemistry that have occurred since the baseline values were recorded. Significant positive differences indicate a much higher presence of hydroxide ions (OH-) than is expected. Significant differences, as found, warrant a formal investigation into the magnitude and origin of recent changes in atmospheric chemistry.

 

Date (2000)

N

Location

1990
pH

1999
pH

2000
Measured pH

1990
Difference

1999
Difference

Jun 26

1

NM

5.1

5.0

6.6

1.5

1.6

Jun 27

NM

5.1

5.0

6.6

1.5

1.6

Aug 17

NM

5.1

5.0

6.2

1.1

1.2

Aug 18

NM

5.1

5.0

6.3

1.2

1.3

Aug 19

5

NM

5.1

5.0

6.6

1.5

1.6

Sep 10

WA

5.3

5.1

5.3

0.0

0.2

Sep 11

IN

4.4

4.4

7.0

2.6

2.6

Sep 11

Great Lakes

4.4

4.5

6.6

2.2

2.1

Sep 11

Great Lakes

4.4

4.5

7.6+

3.2

3.1

Sep 15

10

OR coast

5.3

5.4

5.6

0.3

0.2

Sept 15

Nor. CA-coast

5.3

5.3

5.0

-0.3

-0.3

Sep 17

MA

4.4

4.5

6.0

1.6

1.5

Sep 15

ND

5.3

6.0

6.0

0.7

0.0

Sep 19

WI

4.7

4.7

6.8

2.1

2.1

Sep 19

15

WI

4.7

4.7

7.0

2.3

2.3

Sep 19

MA

4.4

4.5

6.3

1.9

1.8

Sep 21

KS

5.3

5.1

6.8

1.5

1.7

Sep 21

WA

5.3

5.2

5.3

0.0

-0.1

Sep 19

CO

5.2

4.9

5.7

0.5

0.8

Sep 20

20

CO

5.2

4.9

6.0

0.8

1.1

Sep 20

CO

5.2

4.9

5.9

0.7

1.0

Sep 22

WI

4.7

4.7

6.4

1.7

1.7

Sep 22

WI

4.7

4.7

6.6

1.7

1.7

Sep 23

MI

4.3

4.5

6.2

1.9

1.7

Sep 25

25

CO

5.2

4.9

5.5

0.3

0.6

Sep 25

CO

5.2

4.9

5.9

0.7

1.0

Sep 26

MA

4.4

4.5

6.3

1.9

1.8

Sep 27

TX

5.1

5.1

6.7

1.6

1.6

Oct 5

MA

4.4

4.5

6.2

1.8

1.7

Oct 6

30

IN

4.4

4.4

6.7

2.3

2.3

Oct 6

GA

4.6

4.6

5.2

0.6

0.6

Oct 9

OR coast

5.3

5.4

5.3

0.0

-0.1

Oct 10

CA

5.3

5.5

6.4

1.1

0.9

Oct 9

CA (N.)

5.3

5.4

6.4

1.1

1.0

Oct 10

35

CA (N.)

5.3

5.4

6.4

1.1

1.0

Oct 11

CA (N.)

5.3

5.4

6.4

1.1

1.0

Oct 13

WI

4.8

4.8

6.6

1.8

1.8

Oct 16

MA

4.4

4.5

6.1

1.7

1.6

Oct 18

MA

4.4

4.5

6.2

1.8

1.7

Oct 23

40

WI

4.8

4.8

6.8

2.0

2.0

Oct 23

WI

4.8

4.8

6.6

1.8

1.8

Oct 22

CO

5.2

5.0

7.0

1.8

2.0

Oct 23

WI

4.8

4.8

6.8

2.0

2.0

Oct 8

TX

5.1

5.1

6.5

1.4

1.4

Oct 15

45

TX

5.1

5.1

6.8

1.7

1.7

Oct 23

TX

5.1

5.1

7.0

1.9

1.9

Oct 25

CA

5.3

5.4

6.3

1.0

0.9

Oct 26

WI

4.8

4.8

6.4

1.6

1.6

Oct 27

WI

4.8

4.8

6.6

1.8

1.8

Oct 27

50

CA

5.3

5.5

6.2

1.9

1.7

Nov 1

ND

5.3

6.0

6.3

1.0

0.3

Nov 2

WI

4.8

4.8

7.0

2.2

2.2

Nov 5

MA

4.4

4.5

6.2

1.8

1.7

Nov 6

WI

4.8

4.8

6.2

1.4

1.4

Nov 6

55

WI

4.8

4.8

6.4

1.6

1.6

Nov 7

IN

4.4

4.4

6.8

2.4

2.4

Nov 9

GA

4.6

4.6

5.7

1.1

1.1

Nov 14

MA

4.4

4.5

6.2

1.8

1.7

Nov 12

WI

4.8

4.8

6.4

1.6

1.6

Nov 13

60

OR

5.3

5.4

4.9

-0.4

-0.5

Nov 14

OR

5.3

5.4

4.9

-0.4

-0.5

Nov 20

NC

4.5

4.7

6.0

1.5

1.3

Nov 20

NC

4.5

4.7

6.5

2.0

1.8

Nov 21

IL

4.5

4.5

6.0

1.5

1.5

Jan 16

65

MA

4.4

4.5

6.3

1.9

1.8

Jan 19

WA

5.3

5.2

5.5

0.2

0.3

Jan 20

MA

4.4

4.5

6.2

1.8

1.7

Jan 21

ID

5.3

5.2

6.3

0.9

1.0

Jan 21

MA

4.4

4.5

6.2

1.8

1.7

Jan 21

70

ID

5.3

5.2

6.0

0.7

0.8

Jan 23

CA

5.3

5.4

6.8

1.5

1.4

Jan 23

CA

5.3

5.4

6.2

0.9

0.8

Feb 06

MA

4.4

4.5

6.1

1.7

1.6

Feb 09

OR

5.3

5.4

5.1

-0.2

-0.3

Feb 27

75

AR

4.7

4.8

5.9

1.2

1.1

Feb 25

AR

4.7

4.8

6.4

1.7

1.6

Feb 25

NC

4.5

4.9

6.4

1.9

1.5

Mar 08

OH

4.4

4.5

6.1

1.7

1.6

Mar 22

MA

4.4

4.5

6.2

1.8

1.7

Apr 20

80

WI

4.8

4.8

7.0

2.2

2.2

May 02

WI

4.8

4.8

6.9

2.1

2.1

May 05

NM

5.1

5.1

6.0

0.9

0.9

May 06

ND

5.3

6.0

6.5

1.2

0.5

May 04

WI

4.8

4.8

7.0

2.2

2.2

May 07

85

WI

4.8

4.8

6.8

2.0

2.0

May 10

WI

4.8

4.8

6.8

2.0

2.0

May 16

ME

4.5

4.8

6.1

1.6

1.3

It is emphasized once again that:

“The single most important chemical species in clouds and
precipitation is the hydrogen ion (H+), whose concentration can be
indicated by specifying the solution’s acidity, or pH value. You may
recall from high school chemistry that the pH scale ranges from 0 to
14, low pH values indicating high acidity (high concentrations of H+)
and high pH values indicating high alkalinity (low concentrations of H+)”

from Atmosphere, Climate, and Change by Graedel and Crutzen,
Scientific American, 1997.


pH map
Lines of equal pH in the United States 1990
from Atmosphere, Climate and Change by Graedel and Crutzen 1997
(“The levels below 5.0 east of the Mississippi River are the result of
anthropogenic [man-made] emissions of sulfur and nitrogen oxides
.“)

 

MICROSCOPIC PARTICLE COUNT STUDY NEW MEXICO 1996-1999

MICROSCOPIC
PARTICLE COUNT STUDY
NEW MEXICO 1996-1999
Clifford E Carnicom March 23 2000

Clifford E Carnicom March 23 2000

A study of the airborne microscopic particle count data from the State of New Mexico covering the years 1996-1999 has been conducted. Approximately 175,000 observations of hourly monitored data from five stations in the state have been analyzed; this represents a sizable statistical sample. The statistical test that has been designed questions the difference between the data of 1999 vs. the combined data of the three previous years: 1996, 1997 and 1998. The data which has been measured is the airborne particles which measure less than or equal to 10 microns in size (PM10) ( a human hair is approximately 60-100 microns in thickness). The design of this test results from the fact that aerial spraying over the United States has been repeatedly observed and documented with photographs and personal testimony through the course of 1999 and 2000. Records of such spray activity are not available on any widespread basis for the years 1996-1998, and that lack forms the basis for statistical investigation of the relevant data.

The results show that there is a significant statistical difference between the magnitudes, averages, and variances of the two data sets in the state of New Mexico. There is an increase of 16% in the magnitude of the average microscopic particle count data of 1999 vs. the average of the combined years of 1996-1998. The differences between the averages of these two data sets is significant at the 99.9% plus level. Lastly, the differences between the variances (indicative of data distribution) are also equally significant at the 99.9% plus level. Normal distributions are assumed in the analysis.

The conclusion to be reached from this study is that the microscopic air particle count in the state of New Mexico in 1999 is significantly different from that of the preceding three years, and that this difference is directly correlated with the observations of aerial spraying that have taken place during this same time period. The results of this study form a further basis for criminal investigation of the documented spray activity, and for congressional hearings on this subject. This initial study further substantiates those findings that have been presented which document the crimes of aerial spraying against the American people without their informed consent. These findings include numerous telephotos of offending aircraft with extensive spray methods easily visible, cloud progression photographic series, meteorological studies in an arid environment that defy any basis for cloud formation, a certified receipt ground sample which to this date has not been identified by Carol M. Browner, Administrator, United States Environmental Protection Agency, repeated letters of denial and discount by official agencies of the United States government, refusal of or inadequate response by William Jefferson Clinton, William Cohen (DoD), Janet Reno (Attorney General), Carol M. Browner (EPA) , and Jane Garvey (FAA), repeated refusal of response by the various state senators and governors, as well as hundreds of thousands of eye witness accounts and testimony recorded from the beginning of 1999 to the present day.


Clifford E Carnicom
March 23, 2000

 

APPENDIX:

Source of data : New Mexico Environment Department – Air Quality
No. of observations from five monitoring stations 1996-1998 : 129410
No. of observations from five monitoring stations 1999 : 43449
Measured quantity : PM10(<=10microns)
Mean of observations 1996-1998 : 39.42 micrograms/cubic meter
Mean of observations 1999 : 45.70 micrograms/cubic meter
Standard deviation of observations 1996-1998 : 111.69micrograms/cubic meter
Standard deviation of observations 1999 : 134.57micrograms/cubic meter
Zm Statistic : 11.65
F Statistic : 1.45

Readers from other states are encouraged to request similiar data for further analysis. Please feel free to contact me (Clifford Carnicom) for additional assistance in this matter.

PRELIMINARY METEOROLOGICAL STUDY

PRELIMINARY METEOROLOGICAL STUDY

 UNNATURAL CLOUD FORMATIONS IN SANTA FE NM

contrail cloud 2d contrail cloud 2c contrail cloud 2a

February 14, 1999 Santa Fe 0930-1000

Summary: The argument given that upper atmospheric conditions in Santa Fe NM allow for persistent contrails and subsequent cloud formation is refuted with an analysis of upper altitude relative humidity data. The more plausible explanation is direct chemical spraying from aircraft active over this same area, as documented with abundant photographic evidence on www.carnicom.com.


The following preliminary meteorological study has been made on August 26, 1999:

Please note the following information from http://www.weather.unisys.com/model/details.html :

“The relative humidity field is a good predictor of cloud location and thickness. Areas of relative humidity less than 60% generally are clear or have partly cloudy skies. Areas of 60-80% are generally overcast or mostly cloudy. Areas greater than 80% are overcast with a high likelihood of precipitation as relative humidity approaches 100%.”

The following monthly averages of relative humidity in Santa Fe NM at 30,000 ft. above mean sea level have been obtained from the Climate Diagnostics Center of the National Oceanic and Atmospheric Administration:

Jan 1999 38%
Feb 1999 34%
Mar 1999 33%
Apr 1999 28%
May 1999 32%
June 1999 26%
July 1999 32%
Aug 1999 33% (Aug 1-Aug 26)

The average humidity for this eight month period is 32%, with a sample standard deviation of 3.7%.

The previous relative humidity field description would lead one to conclude that Santa Fe would generally have skies at 30,000 ft. that are often very clear. This is, in fact, the meteorological norm for Santa Fe.

Now introduce 21 days of aerosol activity as documented on www.carnicom.com. These days represent the most obvious and blatant examples of spraying, of which numerous cloud progression photos and telephotos of offending aircraft have been presented. These days are:

Feb 14, 16, 17, 25, 28 (27.5, 38, 60, 50, 32.5% respectively)
Mar 2, 6, 24 (55, 48, 20%)
April 7, 11, 12, 18 (25, 35, 28, 32.5%)
May 7, 17, 28, 29 (33, 35, 42.5, 50%)
July 9, 13, 14 (35, 25, 55%)
Aug 14, 26 (21, 40%)

The average relative humidity at the same 30,000 ft. above mean sea level of these 21 days is 37.5%, with a sample standard deviation of 11.7%. This data was also obtained from the Climate Diagnostics Center of the National Oceanic and Atmospheric Administration. By the same reasoning, one would have expected these days to be generally clear at that altitude, but that is not the case.

On each of these days, almost without exception, the morning sky began as clear, and as the planes progressed with the spraying, a cirrus cloud layer was created which often transformed itself into a cirro-stratus layer. In short, a clear day became generally cloudy or hazy.

Please also note the following expectation from http://cimss.ssec.wisc.edu/wxwise/class/contrail.html :

“How long a contrail remains intact, depends on the humidity structure and winds of the upper troposphere. If the atmosphere is near saturation, the contrail may exist for sometime. On the other hand, if the atmosphere is dry then as the contrail mixes with the environment it dissipates.”

By any standards, the average relative humidity of 37.5% at 30000 ft. MSL determined on observable spray days in Santa Fe NM would be considered a dry environment, and according to the previous reference any contrails should dissipate. Instead, the evidence clearly shows that clouds have formed from these trails.

Another source is quoted on http://www.politicalwomen.com/debate.htm as stating that :

“If a contrail is formed in air of low humidity, the ice crystals will rapidly evaporate. But if the air is saturated
(humid) enough, the crystals will persist for some time.”

This statement reiterates for the third time the foundation that low humidity conditions are not conducive to persistent contrails, or subsequent cloud formation. Such events do occur, however, on all 21 documented spray days in Santa Fe itemized above. An average relative humidity of 37.5% is to be considered low, and therefore by all references one would expect the ‘contrails’ to rapidly dissipate. The photographs on this page and elsewhere on this web site provide ample evidence that this is not the case.

One should ask, why are there clouds in our skies on these days if meteorological conditions do not normally support their formation? One reasonable explanation is that there are aircraft leaving aerosol trails. Photographic evidence shows cloud formations progressing in direct correlation with aircraft activity, ground sample photographs show materials that have originated from the sky during aerosol activity, and aircraft telephotos of spray configurations have been captured. These photographs are available at www.carnicom.com.

albuquerque 1 albuquerque 2 albuquerque 3 albuquerque 4
albuquerque 5 albuquerque 6

Albuquerque July 1999