SeaWiFS and MODIS PAR

[peland]

Hi,

We've recently downloaded SeaWiFS and MODIS daily PAR data, and when we plot data from west of Portugal, we notice a small step change in the annual maximum PAR between SeaWiFS (~62 E/m2/day) and MODIS (~64 E/m2/day). Is this expected?

Cheers, Peter

[gnwiii]

I see nobody has responded, so I'll suggest you start with Photosynthetically Available Radiation (PAR) which has links to comparisons. 

It is worth noting that comparisons of extremes are problematic:

"the measurement of uncertainty has often been ignored in extreme value applications. There is some irony in this, as an analysis of extreme values is likely to have more sources of uncertainty than most other statistical analyses" -- An Introduction to Statistical Modeling of Extreme Values

[peland]

Hi Gnwiii,

Thanks for the info - I couldn't get the comparison pages to load, but they're comparisons with in situ data rather than intersensor. The extremes I'm talking about are not quite as extreme as you might think, the annual maximum is very consistent in the SeaWiFS data (this is Portugal after all) and similarly consistent in the MODIS data, but with a step change between the two much larger than the interannual variation. The question is how to bring the two into line to generate a continuous time series. One 'quick and dirty' idea is to use the long overlap, treating MODIS as 'truth' and correlating the two, then 'correcting' pre-MODIS SeaWiFS to make it consistent with MODIS. A more rigorous idea is to compare both with the Gregg & Carder 1990 clear-sky model, which should form an upper envelope to the data. Thoughts, anyone?

Cheers, Peter

[OB.DAAC - SeanBailey]

Peter,

The PAR algorithm is producing integrated daily averages from instantaneous measurements.  There are a LOT of differences in the between SeaWiFS and MODIS  (e.g. spatial resolution, nominal equatorial crossing time, tilt, band suite, etc.) so getting the level of consistency between the two that has been achieved is quite frankly remarkable.
Your proposed approach to normalize out the slight bias between the two is reasonable to me :smile: 

Sean

[gnwiii]

In Portugal you are free of one of the big issues -- seasonal ice cover.   Careful investigation of differences between sensors using means often reveals that the differences between sensors are small under ideal conditions but that systematic differences in missing values dominate differences in the means.  You don't have ice to consider, but land might have a systematic impact due to differences in sensor resolution, stray light response, etc.  We use PAR in production calculations but because production ins integrated over depth, shallow water (e.g., where light reaches the bottom) is masked out.  It would be interesting to see if the means for the two sensors are more alike if you stay away from land.

For your area I would guess that cloud could cause differences: a) diurnal patterns on cloud combined with different pass time distributions could introduce sensor-dependent biases, b) for cloudy pixels, sensor responses to bright pixels may differ.  

Another way to get an envelope is to take the maximum for all longitudes (-180,..., 180 deg) at each latitude in your region of interest.