Ch 3. Environmental Chamber

OVERVIEW

    • In previous chapters we have learned how to load KPP file (generated from MCM website) into the model. We also briefly looked at the photochemistry of CH4, the simplest volatile organic compound (VOC). There are thousands of chemical species even in the cleanest air, perhaps more.
    • In this chapter, we will take a closer look at one chemical species that is emitted from trees, isoprene (C5H8). The global biogenic emission of isoprene is huge. This thing is very reactive in the air, producing a wide range of compounds. Hundreds of species and thousands of reactions are involved!
    • One way to study the atmospheric chemistry is to use environment chambers, because the conditions can be well controlled. In this chapter, we will try to model the OH oxidation of isoprene under high and low NOx conditions, and compare the results to the data reported in Paulot et al Atmos. Chem. Phys. 2009 and Paulot et al Science 2009.

1. Go to the MCM website

  • Browse or search isoprene (or C5H8) and mark it. Then EXTRACT the KPP file. Remember to include inorganic reactions, and generic rate coefficients.

2. Set up the model

  • Open the blank template (pxp). Click the load new KPP button.
  • Then click the ODE_ScratchBoard botton, to check there is an existing ODE_ScratchBoard.ipf. If yes, kill it.
  • Then click the generate new ODE_ScratchBoard button, guess which one it is?
  • Click the make kr button to generate the wave that contains all rate coefficients. Copy the entire wave, CTRL+M, paste inside the FOR loop (see snapshot).
  • Now the model is almost setup!

3. Chamber condition

  • The chamber experiment (Paulot et al Atoms. Phys. Chem. 2009) was operated under this condition:

Temperature: about 296.5 K, RH < 6%.

Initial isoprene: 94 ppbv; NO: 500 ppbv; H2O2: 2.1 ppmv

H2O2 was used to generate OH radicals; RH was maintained low to minimize H2O2 loss. NO was used because the major focus of that study was high NOx chemistry, also to suppress O3. Set these initial conditions in the model!

3. Mimic chamber photolysis

  • Photolysis is a bit tricky. In theory, photolysis frequency (j-value) can be calculated using (spectually resolved) actinic flux, absorption cross section, and quantum yield. But we don't have access to the spectually resolved actinic flux data during that chamber study (there is a chance that this may be reported in other CalTech chamber configuration papers though). If you have this, you may calculate j-values by yourself.
  • What we can do, in a quick and dirty way, is to use the j-value parameterized by MCM, (as a function of SZA). Then we can test a few SZA values until the decay of isoprene looks reasonable, compared to observations. Obviously this is not the best way, as the MCM photolysis parameterization is for ambient conditions, while the chamber was using blacklights (276 GE350BL) emitting 300-400 nm, with a maximum at 354 nm.
  • It appears SZA = 66 looks pretty good? Let's set SZA to 66 degrees, and get the model running.

4. Modeled vs measured:

  • Moment of truth: now let's compare a few more modeled species to the measured!
  • I have included the time series of a number of species in the IGOR experiments. Note that you probably need to convert the units accordingly, especially concentration_matrix[][] is in molec/cm3, and you probably wanna convert to mixing ratios (ppm, ppb, ppt, etc) for comparison. Here's a plot that I made.
  • Overall I'd say the agreement is not bad at all! Usually measurements would have uncertainties. Many of the species in Paulot et al were measured using CIMS, which would have a measurement uncertainty of, let's say, tens of percent. Really depends on how you run the instrument and how you calibrate it.
  • Note that model would have uncertainties too! One important source of model uncertainty is from the kinetics, e.g. uncertainties of measured rate coefficient may span vastly from a few % to hundreds of %. Yield / branching ratio estimated based on measurements would have uncertainties too, which are essentially translated / propagated from individual measurements.

5. Let's have more fun:

  • In the previous sections, we set up the model for Paulot et al Atmos. Chem. Phys. 2009, which was under high NOx condition. With the model, we can also test what's going on under low NOx condition, and we can compare to Paulot et al Science 2009.
  • You can use the same IGOR file. Condition are kinda similar: initial isoprene 94 ppbv, initial H2O2 1.660 ppmv, and let'ts set initial NO to 0. Tune SZA until you get reasonable decay of isoprene (SZA = 70 degree in my case). Then plot the time series of IEPOX and ISOPOOH. Not bad huh ;)

6. Further notes:

In this chapter, we set up the model in a way that is only remotely comparable to chamber condition, therefore the comparison to chamber measurements, although not bad, is not apple-to-apple. It's natural to use 0-D model to model chambers, but there are things you should pay close attention to. I'm not an expert in this area so I'll only list a few in my experience:

    • Wall loss: all molecules in the chamber are constantly colliding with the walls, and once they collide there is a chance they may stick there "forever". Some molecules are more "sticky" than others, e.g., HNO3, H2O2, NH3, and compounds with a lot of functional groups, wall loss of these things could be quite substantial. Not only gases, particles undergo wall loss as well! Depending on sizes and properties, the wall loss characteristic time could vary a lot. For example, NH4NO3 wall loss is pretty fast. Usually humidity would speed up the wall loss; and wall loss is generally a bigger concern for smaller chambers. One way to account for the wall loss in the model, is to add this loss mechanism using the characteristic time, which may be estimated based on measurements. Let's say you injected some H2O2 into a particle-free chamber at very low RH, and you watch the decay under dark. Using the data you can probably estimate a "wall loss characteristic time scale". Then you can add this into the model: open the ODE_ScratchBoard.ipf, scroll all the way to the bottom, then add this line (before the return statement):

 YDOT[x] = YDOT[x] - Conc[x] / CharTime_SpeciesX_sec

where x is the species index, CharTime_SpeciesX_sec is the wall loss characteristic time of this species (unit: second). YDOT[] is just dC/dt (molec/cm3/s), and Conc[] is the concentration (molec/cm3) at each iteration step.

Note that sometimes people would report "wall loss corrected data", or normalize the measurements of a reactive constituent to a non-reactive one. In these cases, you probably don't need to worry too much about wall loss. Bottom line is, you need to know what exactly you're modeling and what exactly you're comparing to.

    • Photolysis: in this demo we used the MCM built-in parameterization for photolysis, as a function solar zenith angle. This is for ambient conditions (clear sky), and obviously doesn't work for most chamber conditions. Photolysis is a chamber-specific thing. You need to figure out a proper way to estimate the photolysis in your chamber. Sometimes people would estimate a "background" 1st-order photolytical decay rate for certain species and take it as the effective photolysis frequency of this species for the chamber experiments (watch out for wall loss, OH-propagation, etc). You can load this into the model, and scale the j-values of other species to this species.
    • Condensation / evaporation particles: we will not discuss this in detail, but there are equations out there.
    • Heterogeneous reactions: we will briefly touch this topic in Chapter 5. One thing unique about chamber heterogeneous reactions is that, you may need to correct for the diffusion limitations, and this is especially the case for big particles and very sticky compounds.