Prelim. outline for Data Analysis

Monthly scale – refers to 5 weeks of data plotted 1-2-3-4…as days

Plot: CO2 continuous data from picarro, 5-min running mean

Analyses:

-       regression line

o   8 oct à 27 oct, fitted to 2 weeks of data from window

o   28 oct à 14 nov, fitted to the other 3 weeks of data (from the roof)

o   slope from regression line is the ‘quantifiable’ increase in [CO2]/time

-       periodicity – interpolate onto graph to see how data of individual days relates to ‘average’ periodicity of the dataset

o   are our hypothesized timescales of interest reflected in the periodicity of the data?

-       Fourier Transform – see how noisy the data is, look at frequency domain

o   Substantial differences between the roof and window data?

o   Is data useful or just a mess (extremely noisy)?

-       regression and correlation for each factor to [CO2] and see which is dominant (has the greatest ‘relationship’ to the pattern of CO2

o   wind speed and (direction), ppn, T

§  do you have to transform CO2 to the same timescale to analyze or do you interpolate the factor data

-       do Principle Component Analysis?

-       regression between δ13C month-long plot & [CO2] (negative)

-       Mass balance and keeling plot

Weekly scale (M, T, W, Th, F, Sa, Su stacked on top of each other – average)

x2, for 2 weeks of window measurements and 3 weeks of roof measurements

                                                                                               

Analyses:

Just for traffic and CO2

-       must test which days of the calendar week best represent our hypothesis for ‘weekday’ traffic pattern being different from ‘weekend’ traffic pattern

o   i.e. how do we define a weekday and a weekend? (aside from solely based on the calendar)

1^st test:

plot Mà F vs Sa, Su

                  2^nd test:

plot T à Th vs F, Sa, Su, M

  

From these tests:

- Is there a significant different between the traffic and CO2 patterns on weekdays (defined by tests 1 & 2) and weekends?

Analyzing all factors:

Now, take out a ‘test week’ from roof and window datasets (picarro) and look at these two datasets in relation to averaged out weekly CO2

-       again correlation, regression etc

o   do the values change as compared to those calculated for monthly?

o   If so, can we infer something about the degree of relationship (between each factor and CO2, or for traffic and biogenic - δ13C) changing over time or according to timescale being analysed

§  Is our method of looking at these relationships skewing the results or do they change over time

-       for CO2, look at variance/std deviation etc in relation to averaged out week to see if there is some sort of representativeness in the average values

-       pick weeks with 2+ days of traffic data from counting cars (‘real-time’ instead of averaged out google data)

Diurnal – Averaged out so that it represents more than a single day i.e. is representative of all days stacked into 1 24h period

-       there are many options for how we construct this 24 hour period

o   Based on weekly analysis of the CO2 pattern

1)   2 plots, one for a weekday (defined as Mà F or T à Th) and one for a weekend (Sa, Su or F, Sa, Su, M)

o   x2 for the 2 weeks of window data and the 3 weeks of roof data

2)   1 plot using data from weekends and weekdays

o   x2 for window and roof

3)   1 plot using data from weekends and weekdays for all 5 weeks

Analyses: will compare this ‘ideal’ or averaged out day to actual days where the influence of one factor is predominant (e.g. Oct 22, very heavy winds)

-       see if some factors have  a relatively constant value for correlation despite not being ‘dominant’ on that day

-       will want to look at amplitude of peaks, [CO2] values, variation (fluctuations), etc

o   basically, characterize or describe the CO2 pattern, δ13C as an ideal/representative pattern, and then see how it changes on a case-by-case basis

Case-by-case analysis – Has been  expanded and will be explained in class and then added.

Update on CO2 data

Here are some graphs of the recent Picarro data, which Gregor gave me last week. The fist graph shows part of the data we looked at on last Tuesday (this time only form 17th to 26th). The second graph is the most recent CO2 data that i have. It ranges from October 26th to November 4th.

I thought it is a good idea to show this now, such that everyone can get an idea what the recent data looks like.

There are a couple of gaps in the readings which corresponds to the time when flask measurements were taken. Except the large one between Tuesday 2^nd and Wednesday 3^rd, which is due to the power shut down.

The black crosses correspond to the raw 12-CO2 data. We can see that the CO2 signal becomes much more spread starting at mid -Thursday (28^th of October). I’m guessing that this corresponds to the event when the tube was set up on the roof.

I also attached a 3rd graph showing the standard deviation of the 2nd CO2 time series (the one between October 26th and November 4th). You can see that the error increases significantly after the 28th, and then even more after the 29th of October.

The red line is a running mean with a window of 100 time units, corresponding to about 15 minutes.  The running mean shows a single curve with a detectable trend, but after the 28th of October it has a much larger deviation associated with it.

In addition, it seems that the "spread of the Co2 signal" decreases during night time, but is prevalent through the day until late night.

That is what I can say now. Let me know if you have some questions and feel free to comment on this.

Even more things to look at

Comparison between our first Panorama (18th of Oct) and the one on Nov.1. 

-g

Something to look at!

Hi, This is the graph showing the assembled 12-CO2 Picarro measurements from the 9th to the 26th of October.

Enjoy!

- Jan-Erik

Transportation Study in Montreal

This is from a report (following a transportation conference) by Chapleau and Morency (from the Civil Engineering Department (Transportation) at Ecole Polytechnique de Montreal): "Generic in nature but specific to the methodological procedures undertaken in the Greater Montreal Area (GMA), the CATI (Computer-Assisted Telephone Interview) household survey is conducted about every five years over a 5% sample. Typically, it represents about 160,000 people belonging to 65,000 households declaring some 400,000 individual trip records for an average weekday. Individual trips are geo-referenced for the residence, trip origin and destination, modal junction points (kiss-and-ride and park-and-ride locations), and are described for their household and personal characteristics (age, gender, car license, car ownership, income) in addition to the trip attributes (purpose, mode, departure time, train-subway-bus routes taken if traveling by transit, bridges and highway taken if traveling by car)."

They have a lot of GIS analysis done on transportation ranging from buses, cars to subway trains. A nice map they made was the percentage of motorized trips in Montreal am rush hour:

We could contact them for the data if we find it is useful!

-Angela.

Mount-Royal Flasks

This morning at 7am (it was still dark outside), Pogo and myself took our first flask measurment! I placed a stick with a Yellow Flag in the ground and I tied an Orange tape across a branch in the air at the spot where i took the measurments. It is directly facing you when you reach the top of the stairs.

-Atleast it is warm today, Angela.

Principles of Urban Meteorology (re atmospheric CO2)

(by Andrew)

The generalization of fundamental meteorological principles over urban environments is more problematic than to other land surfaces. This is for three reasons. First, the heterogeneity of roughness elements within the complex 3-D geometry of the urban “canopy” creates spatial variability in turbulence patterns. Second, the multiple sources and sinks of momentum, heat, moisture, and emissions creates spatial variability in fluxes and concentrations. Third, the impact of human activities to continually re-shape and alter, in new and distinct ways, the urban environment itself limits the long-term validity of any findings. In short, there is considerable uncertainty concerning the dynamics of the unique microclimates found within previously understudied and increasingly complex urban areas.

Both across cities and between cities, significant spatial and temporal variability in CO2 concentrations and fluxes can be expected as a consequence of the distribution of anthropogenic sources (mobile and fixed), processes of urban vegetation (including irrigation, and patterns of atmospheric convection and advection. Previous studies have shown that there is a marked and distinct diurnal cycle in the concentration of CO₂ with a morning peak attributable to anthropogenic (largely traffic), biogenic (nocturnal respiration), and meteorological (atmospheric stability) factors. In contrast, a mid-afternoon minimum can be attributed to vegetative photosynthesis and strong convective turbulence; concentrations then begin to rise again during the evening “rush-hour” traffic-flow.

....

The main roughness elements in urban environments are trees and buildings. Other than being large, trees and buildings share few characteristics of meteorological significance. Aerodynamically, buildings are true “bluff” bodies because of their impermeability, inflexibility, and sharp edges. When exposed to airflow they create strong positive and negative pressure differences over their surface, leading to flow separation and vortex shedding. Trees are also good generators of mechanical turbulence but buildings have to be judged as more effective roughness elements. The effects of smaller roughness elements, such as cars or paved surfaces, are minimum in comparison.

Three spatial scales are commonly utilized for studying urban environments:

-       the micro-scale (10¹–10²m) involves spatial differences in response to individual roughness elements (variability in building/canyon dimensions, trees) and proximity to localized emissions sources (e.g. roads, vegetation);

-       the local-scale (10²–10⁴ m) represents the integrated response of an array of roughness elements with spatial variability reflecting the unique characteristics of different neighborhoods/land-uses;

-       and the meso-scale (10⁴–10⁵ m) considers the city in its entirely, and differentiated from its surroundings, areas of forest, agriculture, etc.

The urban canopy layer (UCL) is defined as being from the ground to the mean height of the roughness elements, usually just below roof –level, where micro-scale effects of the site characteristics are dominated. The UCL is most clearly delineated in areas of high building density; it may be discontinuous or absent in less densely developed suburban areas.

The layer extending from the top of the UCL, to a height where urban surface influences are no longer perceptible, is defined the urban boundary layer (UBL). It includes the roughness sub-layer immediately affected by the individual roughness elements, the turbulent surface layer (local-scale), and the outer mixed layer (meso-scale).

....

Of those studies employing atmospheric-based measurement methods to study CO2 concentrations in urban environments to date, virtually all, with a few exceptions, have focused on the micro-scale, considering processes and patterns within the UCL. Inadequate attention has as yet focused on how micro-scale results can be extrapolated to larger scales and on how to accurately study the local-scale using atmospheric-based measurement methods. 

In regards to the latter, current debates focus on determining the height (or “depth”) of the roughness sub-layer, in which the perturbations caused by individual roughness elements are “blended” together due to atmospheric turbulence. It is as this height that instruments are to be placed in order to study at the local-scale that is spatially representative of a distinct urban neighbourhood/land-use. To be sure, placing instruments are greater heights than this leads to increased risk of incurring errors due to advection from dissimilar upwind surfaces and storage changes below the measurement level due to vertical flux divergence.

It is known that the height of the roughness sub-layer is a function of both the length/height of roughness elements (z_H) and their horizontal spacing. More recent research suggests that the latter factor may in fact be the primary determinant. It has been estimated that, as a general “rule-of-thumb”, instruments must be mounted at a height at least twice the mean height of the roughness elements (approximately 20-90m) to ensure that they are above the influence of individual roughness elements and, therefore, that the measurements represent an integrated response at the local-scale.

....

References

Grimmond, C.S.B., et al. (2006) Progress in measuring and observing the urban atmosphere.

Theoretical and Applied Climatology 84, 3-22.

Grimmond, C.S.B., et al. (2002) Local-scale fluxes of carbon dioxide in urban environments:

methodological challenges and results from Chicago. Environmental Pollution 116,

243-254.

Kanada, Manabu. (2007) Progress in Urban Meteorology: A Review. Journal of the

Meteorological Society of Japan 85B, 363-383.

Koerner, B. and J. Klopatek. (2002) Anthropogenic and natural CO2 emission sources in an

arid urban environment. Environmental Pollution 116, 45-51.

Nemitz, E., K. J. Hargreaves, A. G. McDonald, J. R. Dorsey, and D. Fowler (2002) Micrometerological Measurements of the Urban Heat Budget and CO₂ Emissions on a City Scale. Environ. Sci. Technology 36, 3139-3146.

Oke, T.R., et al. (1988) The urban energy balance. Progress in Physical Geography 12, 471-

483.

Oke, T.R., et al. (1989) The Micrometeorology of the Urban Forest. Philosophical

Transactions of the Royal Society of London 324, 335-349.

Wentz, Elizabeth A., et al. (2002) Spatial Patterns and Determinants of Winter Atmospheric

Carbon Dioxide Concentrations in an Urban Environment. Annals of the Association of American Geographers 99(1), 15-28.