MODIS/Aqua Validation Analysis

The effect of the proposed MODIS/Aqua reprocessing changes on the comparison with in situ data are presented here. A description of the methods employed for these analyses, with minor modifications, can be found in Volume 10, Chapter 7 of the SeaWiFS Post Launch Technical Memorandum Series.
There are approximately 55 radiance and 200 chlorophyll matchups which pass all exclusion criteria.
The matchup analysis was performed three ways, using:

the current operational parameters
the proposed reprocessing parameters with out the f/Q correction applied
the proposed reprocessing parameters with the f/Q correction applied

Statistics Summary:

Current Operational Processing

Proposed Reprocessing (without f/Q)

Proposed Reprocessing (with f/Q)

Band	N	Mean Ratio^*	% Difference^**	Bias	rms⁺
nLw_412	54	0.742	26.005	-0.271	0.453
nLw_443	65	0.807	20.418	-0.225	0.388
nLw_488	46	0.986	15.757	-0.058	0.268
nLw_531	15	0.963	13.933	-0.074	0.264
nLw_551	54	0.993	11.128	0.044	0.194
nLw_667	46	0.815	27.702	-0.012	0.039
chl	200	1.268	48.235	0.299	0.763

N	Mean Ratio^*	% Difference^**	Bias	rms⁺
55	0.637	37.088	-0.393	0.565
66	0.828	19.427	-0.191	0.375
47	0.934	18.852	-0.111	0.285
14	0.955	13.404	-0.105	0.272
55	0.989	10.599	-0.008	0.185
46	0.689	37.449	-0.018	0.042
197	1.331	48.614	0.345	0.819

N	Mean Ratio^*	% Difference^**	Bias	rms⁺
53	0.633	39.311	-0.398	0.568
64	0.804	20.428	-0.230	0.395
45	0.823	21.873	-0.189	0.315
14	0.822	19.017	-0.218	0.321
53	0.874	17.135	-0.123	0.217
45	0.621	41.341	-0.025	0.045
197	1.139	45.339	0.219	0.775

* arithmetic mean of all satellite / in situ ratios
** median absolute percent difference
+ sqrt ( &sum(satellite - in situ)^2 / n )
chlorophyll data were log transformed prior to statistics calculation

The Plots

Operational Radiances Chlorophyll
Proposed Reprocessing without f/Q Radiances Chlorophyll
Proposed Reprocessing with f/Q Radiances Chlorophyll

Discussion

Unfortunately, the quantity of validation data available for MODIS-Aqua is sparse and dominated by regions with challenging aerosol conditions. In fact, the radiance matchups are dominated (30 of the 55) by cruises in the Adriatic Sea. Given these limitations, it is difficult to draw solid conclusions from the available dataset. However, there are some general trends which are worth noting:

There is a noticeable negative bias in the validation results with the f/Q correction applied vs. results without the correction
The overall median percent difference and rms values are coming close to those achieved by SeaWiFS
As was the case with SeaWiFS, the 'green' bands are closer to in situ values than the 'blue' bands. This points to an atmospheric correction problem and not necessarily a sensor calibration issue, although sensor calibration cannot be ruled out.

Curator: OceanColor Webmaster

Authorized by: gene carl feldman

Updated: 19 June 2009