search

ktoddbrown / decomPower

A soil carbon model power study
MIT License
3 stars 0 forks source link

Cap vs measurement time #3

Open ktoddbrown opened 7 years ago

ktoddbrown commented 7 years ago

In general the difference between cap time and measurement time is dictated by the detection limit of the instrument. From what I can recall this is generally an hour or so for high carbon soils in the beginning (days) of the experiment and can be as long as a month for low carbon soils late (1yr) into the incubation.

We can go off of the data for now however I think that Sean's studies are low carbon soils so I would expect them to be on the long cap time end of things.

Sean might have better insight then I on this one.

milkha commented 7 years ago

@SeanSchaeffer,

We are trying to simulate data with different sizes to examine the effects of sample size on estimated parameters (uncertainty interval, identification, etc.). Could you please let me know what you think about any of the following?

(i) To have different sizes of data, I am thinking of having a set with many cap/measurement times, and then subsample this at different frequencies. To be precise, in R, we can have a variable t_meas which is, say, of length 100. Then, we can subsample this as t_meas[seq(1,length(t_meas),k)] to keep every k sample. (Same for t_cap, the cap times). Then we can compare the inference results for different subsamples. (@bob-carpenter, is this a sensible way to examine the effects of sample size?)

(ii) For now, we can limit the measurement time to one year and vary the number of data points within the year. Later, we can expand the horizon to 10 years.

(iii) Now, we have to think about defining t_meas. @ktoddbrown mentioned we have more measurements in the beginning and then the measurements are spaced out as experiments continue. What do you think about the following measurement pattern: every half day for the first two weeks, every day for the next month, every two days for the next 3 months, and every week for the rest of the year. This is a somewhat arbitrary choice I made, please let me know if that makes sense or somewhat agrees with how measurements are done in real experiments.

(iv) Finally, we can define t_cap to be t_meas minus a small amount: one hour in the beginning to one day after a few months.

Thanks!

SeanSchaeffer commented 7 years ago

@milkha

The measurement makes sense to me on a biogeochemical perspective. We would want to capture those easily accessible C substrates that get gobbled up early on (soluble compounds, microbial residues). At some point in time the microbial community should come into equilibrium with rate of substrate supply, and presumably recycling of microbial biomass, and things would slow down quite a bit. A bit of an aside, but the d13C data could suggest when these changes in substrate utilization may be occurring. Obviously, your measurement times are more frequent than we can realistically sample by hand (would be nice though).

I'm not sure what you mean by cap times. I may have missed part of the conversation on this.

milkha commented 7 years ago

I'm not sure what you mean by cap times. I may have missed part of the conversation on this.

@ktoddbrown suggested we fit the model to the CO2 flux, which is the rate of change of CO2 at t_meas. Since it is not possible to measure the rate of change exactly, it is approximated by the change of CO2 between t_cap and t_meas, i.e., (CO2[t_meas] - CO2[t_cap]) / (t_meas - t_cap). The cap time is slightly earlier than measurement time (1 hour to 1 day).

bob-carpenter commented 7 years ago

The cap time is the amount of time before the measurement at which a cap was put on the sample jar (or whatever---they always seem to be mason jars in the photos) to collect carbon. So when there's less being released, you need to put cap on earlier. Then the measurement is carbon collected over the period of time between the capping and the measurement.

On Jan 24, 2017, at 11:17 AM, Milad Kharratzadeh notifications@github.com wrote:

I'm not sure what you mean by cap times. I may have missed part of the conversation on this.

@ktoddbrown suggested we fit the model to the CO2 flux, which is the rate of change of CO2 at t_meas. Since it is not possible to measure the rate of change exactly, it is approximated by the change of CO2 between t_cap and t_meas, i.e., (CO2[t_meas] - CO2[t_cap]) / (t_meas - t_cap). The cap time is slightly earlier than measurement time (1 hour to 1 day).

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.