integration of satellite data and in-situ e. ciancia (1 ... · radiometric measurements (rrs)...
TRANSCRIPT
E. Ciancia (1)*, G. Bernini (2), I. Coviello (1), C. Di Polito (1), M. Faruolo (2), T. Lacava (2), A. Madonia (3), M. Marcelli (3), S. Pascucci (2), R. Paciello (2), A. Palombo (2), N. Pergola(2), V. Piermattei (3), S. Pignatti (2), F. Santini(2), V. Satriano (1), V. Tramutoli (1), and F. Vallianatos (4)
(1) University of Basilicata (UNIBAS), Potenza, Italy; (2) Institute of Methodologies for Environmental Analysis (IMAA), National Research Council
(CNR), Tito Scalo (PZ), Italy; (3) University of Tuscia, Laboratory of Experimental Oceanology and Marine Ecology, Civitavecchia (RM), Italy; (4) Technological Educational Institute of Crete (TEIC) - Chania, Greece
(*) Corresponding author: [email protected]
Dierssen, H.M. (2010). Perspective on empirical approaches for ocean color remote sensing of chlorophyll in changing climate. Proceeding of the National Academy of Sciences of the United States of America 107.
McClain, C.R. (2009). A decade of satellite ocean color observations. Annual Review of Marine Science 1, 19-42.
O'Reilly J. E., Maritorena S., Siegel D., O’Brien M. C., Toole D., Mitchell B. G., et al. (2000). Ocean color chlorophyll a algorithms for SeaWiFS, OC2, and OC4: Version 4. SeaWiFS Postlaunch Technical Report Series, Vol.11. SeaWiFS postlaunch calibration and validation analyses: part 3.
Tramutoli V.: Robust Satellite Techniques (RST) for Natural and Environmental Hazards Monitoring and Mitigation: Theory and Applications. Proceedings of Multitemp 2007, Fourth International Workshop on the Analysis of Multitemporal Remote Sensing Images, Louven, Belgium,18-20 July 2007, DOI: 10.1109/MULTITEMP.2007.4293057, 2007.
NASA Ocean Clor website: http://oceancolor.gsfc.nasa.gov
The first part of this work, focused on the assessment of MODIS Chl-a OC3 algorithm accuracy at local scale, confirms that the standard NASA MODIS algorithm tends to overestimate in situ Chl-a measurements in coastal waters, with an average uncertainty higher than 35%. To improve these results, this algorithm was calibrated at local scale (i.e Ionian sea - South of Italy) using in situ radiometric measurements (e.g. ASD FieldSpec Pro) concurrently with in situ chlorophyll-a data, showing a clear increase of the accuracy of the previous achievements. Obviously, to perform a more robust calibration, a larger number of samples representative of different local bio-optical conditions should be analyzed. The second part of this work, relative to the multi-temporal RST-based analysis, shows that MODIS Chl-a product can be in any case used to describe the general behavior of the tested area in terms of trends and spatiotemporal anomalies. In particular, anomalous areas in terms of Chl-a variations have been detected during the first ten days of February 2011, probably due to an intensive algal bloom.
CONCLUSIONS
INTEGRATION OF SATELLITE DATA AND IN-SITU MEASUREMENTS FOR COASTAL WATER QUALITY MONITORING: THE IONIAN SEA CASE STUDY
1. INTRODUCTION 2. TEST SITE
3. ASSESSMENT OF MODIS Chl-a OC3 ALGORITHM ACCURACY AT LOCAL SCALE
Figure 1. The empirical construction of OC4V4 chlorophyll a algorithm used for the SeaWiFS sensor (Dierssen, 2010)
The testing and assessing of Ocean color (OC) products accuracy at regional scale play a relevant importance in Satellite-Biological Oceanography. In particular chlorophyll-a concentrations derived from satellite
measurements are often useful for primary production estimation; this is why it is essential to provide reliable and accurate data. In such a context, this work is focused on evaluating the potential of MODIS Chlorophyll-a product in terms of:
Assessment of MODIS Chl-a OC3
algorithm accuracy at local scale
Ocx algorithms need a calibration at locale scale
Ocx algorithms are less sensitive to high chl-a
a)
Chl-a spatiotemporal anomalies
Seasonality or long-term trend
b)
Figure 2. Example of historical series of Chl-a maps
+
The investigated area is the Ionian Sea coastal area off Basilicata Region (southern Italy) (Fig. 3). Such an area is a typical transitional environment and has a high historical, cultural and economic value for the region .
Figure 3. Ionian coast off Basilicata region (southern Italy).
To assess the accuracy of NASA Level 2 MODIS chlorophyll-a product (OC3M algorithm) (O’Reilly, 2000), three in-situ measurement campaigns were carried out on 18-19 April 2013 (26 sampling stations), 16 July 2013 (14 sampling stations ) and 1-2 July 2014 (20 sampling stations).
• MODIS Chl-a product (OC3) overstimation
•APD = 83%, well outside the MODIS Chl-a product accuracy goals of 35% (McClain 2009)
II. Match-up analysis
19 April 2013 - 1210 GMT
16 July 2013 - 0940 GMT
02 July 2014 - 1215 GMT
Figure 4. Comparison between MODIS chl-a map and the respective in situ chl-a map.
Figure 5. Scatter plot of Chl-a situ and Chl-a sat for the three measurement campaigns
Figure 6. Scatter plot in logarithmic scale of Chl-a situ and MBRASD for the second measurement campaign (16 July 2013)
III. Calibration
To improve the satellite chl-a product accuracy, its calibration at local scale was carried out. In a first step, only in situ radiometric measurements (Rrs) obtained by an ASD FieldSpec Pro concurrently with in situ chlorophyll-a data were compared. The Maximum Band Ratio at the basis of the OC3M algorithm was implemented in this analysis, using Rrs measurements achieved during the 16 July 2013 campaign, at specific wavelengths, to compute it:
max(Rrs 443; Rrs 488)
Rrs 547 MBR ASD =
Then, carrying out a regression analysis, a linear best fit was chosen (R2=0.73) as representative of the relationship between MBRASD and Chl-a in situ measurements (Figure 11)
where R=LOG10(MBR ASD)
(-2.51*R -0,26) Chl-a situ = 10
Finally the calibration procedure was completed, carrying out a regression analysis between Rrs situ and Rrs MODIS. It should be stressed that this kind of analysis was implemented on the respective maximum band ratios (MBRASD – MBRMODIS) rather than considering the single Rrs band, to reduce errors at 443 nm, due to atmospheric effects and thus to maximize the coefficient of determination R2 0.24*MBRMODIS +0,98 MBRASD =
•Chl-a algorithm calibration
•Rrs calibration
MODIS Chl-a map In situ Chl-a map
Touristic village Marinagri-Policoro Metaponto beaches
Archaeological Park Policoro Archaeological Park Policoro
Multi-temporal analysis of MODIS
Chl-a product at regional scale
III. Validation
•Chl-a algorithm validation
The calibration equation was applied on data acquired during the third measurement campaign (01-02 July 2014), in seasonal conditions (stratified waters) similar to the ones of the calibration dataset (figure 6).
02 July 2014 – 12:15 GMT
OC3M Chl-a algorithm
02 July 2014 – 12:15 GMT
OC3M Chl-a algorithm - calibrated
Ionian algorithm
OC3M algorithm
Figure 8. Scatter plot of Chl-a algorithm validation for the third measurement campaign (01-02 July 2014)
Figure 7. Chl-a satellite maps on 02 July 2014: comparison between OC3M and Ionian chl-a algorithm
APD ≈ 25% < 35% (MODIS goal)
MODIS Aqua Ocean Color Product (L2_LAC) from 2003 to 2013 covering the regions of interest have been acquired (by NASA OC web portal); in particular, for each month, Chl-a temporal mean and standard deviation have been computed. MODIS OC L2 data associated with the following quality flags were discarded: atmospheric correction warning, large viewing angle, large sun angle, clouds, low water-leaving radiance and Chl-a algorithm failure
Mean Chl-a FEB 2003 - 2013 Mean Chl-a MAY 2003 - 2013
Mean Chl-a AUG 2003 - 2013 Mean Chl-a NOV 2003 - 2013
3. MULTI-TEMPORAL ANALYSIS OF MODIS Chl-a PRODUCT AT REGIONAL SCALE
I. Chl-a seasonality analysis
Figure 9. Seasonal monthly mean (2003-2013) Chl-a in Ionian Sea off Bsilicata region
• chl-a highest values from November to April due to run-off and nutrients increase;
• most stratified waters from May to October
Figure 10. Seasonal Chl-a trend performed for case study area
II. RST approach
Historical series of MODIS chl-a product were analyzed by a general multi-temporal approch, the: Robust Satellite Techniques RST (Tramutoli, 2007)
y
t
x
From HOMOEGENEOUS (same time of the day, months of the year, etc.) historical chl-a data-set
),(
),(),,(),,(
yx
yxVtyxVtyx
V
REFV
ALICE (Absolutely Local Index of Change of the Environment)
Example of chl-a mean
VREF(x,y)
σV(x,y)
Example of chl-a standard deviation
III. From long-term interannual variability to short-term investigations: chl-a anomalies
In order to identify when chl-a anomalous values occured in the selected area an interannual chl-a variability analysis (at monthly scale) was carried out:
Figure 11. Chl-a interannual variability (2003 – 2013) at monthly scale for the analyzed area (39 – 40.7 N, 16.4 – 18.6 E)
Figure 12. Chl-a spatial average trend during Febraury 2011. The days characterized with more than 50% of cloudy pixels were discarded.
Following the RST approach, the first eleven days of the month are considered statistically anomalous, because the chl-a spatial average is two times higher than the unperturbed value.
In order to understand what days of February 2011 produced these high values, Chl-a daily spatial average trend was performed and compared with RST monthly values.
Focus on Febraury 2011
Figure 13. Chl-a ALICE Maps of some rapresentive days during Febraury 2011, depicted by red circles in the figure 12
anomalous areas due to an intensive algal bloom affected the first ten days of February 2011
Feb (2003 – 2013) interannual spatial average
Feb (2003 – 2013) interannual spatial average + 2σ
Daily spatial average
(*) Corresponding author: [email protected]
cloud or no data Chl-a ALICE > 5 Chl-a ALICE > 4 Chl-a ALICE > 3
05 – 02- 2011 12:30 GMT 07 – 02- 2011 12:20 GMT
08– 02- 2011 11:25 GMT 27 – 02- 2011 11:55 GMT