PI: Toru Hirawake, Hokkaido Univ. CI: Sei Ichi Saitoh, Hokkaido Univ. Amane Fujiwara, NIPR

Transcription

1 Progress report in 2013FY Estimation of net primary productivity (NPP) and discrimination of phytoplankton functional types (PFTs) (GCOM C1 RA4 No.304) PI: Toru Hirawake, Hokkaido Univ. CI: Sei Ichi Saitoh, Hokkaido Univ. Amane Fujiwara, NIPR

2 Tasks in 2013FY 1. Reconstruction and validation of NPP and PFT algorithms 2. Collection of new data 3. Construction of database for in situ data 4. Writing of protocols for in situ measurements

3 Ocean Net Primary Productivity algorithm Absorption Based Primary Productivity Model (ABPM)

4 PP eu B Pf [ã ph Chl (0 )] opt P B Absorption based primary productivity model (ABPM) surf Z eu E E Function of SSTin VGPM 0 DL P B opt Chl surf VGPM (Behrenfeld & Falkowski, 1997) Assuming chl.a is homogeneous within water column = Popt Depth Chl a and SST were not used. ã ph (0 ) Absorption coefficient at the sea surface Hirawake et al Polar Biology 4

5 Reconstruction of Algorithm Prototype (Hirawake et al., 2011; Hirawake et al., 2012) a ph (443, 0 ) or ā ph (0 ) New version a ph (443, 0 ) x E 0 / DL or ā ph (0 ) x E 0 / DL (mol photons m 3 h 1 ) P opt =P B opt C surf P opt =P B opt C surf

6 Map of sampling stations for NPP : Development : validation

7 Dataset for NPP

8 Estimated P opt (P B opt C surf ) [mg C m-3 h -1 ] Estimated P opt In In situ P opt P opt In-situ P opt (P B opt C surf ) [mg C m-3 h -1 ] using in situ data Latitude RMSE=0.34 n= Estimated P opt 443 (P B opt C surf ) [mg C m-3 h -1 ] Estimated P opt In situ P opt In-situ P opt 443 (P B opt C surf ) [mg C m-3 h -1 ] Latitude RMSE=0.34 n= Validation of Popt=PBopt Csurf ā ph (0 ) x E 0 / DL a ph (443, 0 ) x E 0 / DL

9 In-situPPeu[mgCm-2d-1]EstimatedPPeu[mgCm-2d-1]RMSE=0.35 n= EstimatedPPeu(443)[mgCm-2d-1]In-situPPeu(443)[mgCm-2d-1]RMSE=0.33 n= validation of NPP (PPeu) using in situ data Estimated PP eu Estimated PP eu Latitude In situ PP eu In situ PP eu Latitude a ph (443, 0 ) x E 0 / DL ā ph (0 ) x E 0 / DL

10 RMSE=0.34 n=37in-situppeu(443)[mgcm-2d-1]on dayestimatedppeu(443)[mgcm-2d-1] RMSE=0.34 n=100 ±3daysIn-situPPeu(443)[mgCm-2d-1]EstimatedPPeu(443)[mgCm-2d-1] Validation of NPP (PPeu) using MODIS data Estimated PP eu In situ PP eu Estimated PP eu In situ PP eu On the day +/ 3 days MODIS a ph (443, 0 ) x E 0 / DL

11 Phytoplankton Functional Types (PFTs) Phytoplankton groups categorized by biogeochemical and ecological function Phytoplankton taxonomy Phytoplankton size class (this study) determines number of trophic levels in food web

12 Particle size distribution (PSD) N/V(D) = N 0 /V (D/D 0 ) ξ (Reynolds et al., 2001) Particle number concentration at diameter D (N/V(D)) is a power law function of Junge Slope ξ (Junge, 1963) Advantage Deriving the Junge slope (ξ)and N 0 /V make it possible to estimate particle number at any diameter Disadvantage It derives particle number, not phytoplankton directly

13 Calculation of the fraction of phytoplankton size classes Chla: Chlorophyll a concentration (mg m 3 ) within a size range from D min (μm) to D max (μm) nchla: Chl a concentration normalized by the size bin width D 0 : Reference diameter D: Size of phytoplankton η: Slope of size distribution Size fraction could be calculated without deriving nchla(d 0 ), D 0. Only η is essential parameter. Size fraction at any size can be derived, by substituting Dmax and Dmin (e.g. ultraplankton < 5 m) We can calculate the size fraction using the following equation

14 Dataset In situ data at 106 stations Size fractionated Chla Phytoplankton absorption coeff. (a ph ) Particle backscattering coeff. (b bx ) Remote sensing reflectance (R rs ) Development: 74 stations (70%) Validation: 32 stations (30%)

15 Reconstruction of algorithm In 2012 FY (RA2) Derived the slope η directly from phytoplankton absorption (a ph ) and particle backscattering coefficients (b bp ). Value of b bp is very small and particle size distribution is sensitive to b bp. Therefore, small change (error) in b bp induce large error. In 2013 FY (RA4) Derived the slope η through Fmicro from only aph. Roy et al. (2013) is almost same way but they used aph(676) which is difficult to estimate from satellite correctly. Therefore, shorter wavelengths were chosen.

16 Algorithm η=a 0 + A 1 x F micro + A 2 x F micro 2 + A 3 x F micro 3 A 0 = 2.53, A 1 = , A 2 = , A 3 = r 2 = 0.99 F micro =1/[1+exp(B 0 xa ph (412)+B 1 xa ph (443)+B 2 xa ph (510))]x100 B 0 = , B 1 = , B 2 = , B 3 = Estimation of F micro, F nano, F pico

17 Estimation of η y = 0.58 x 0.42 R² = 0.47 Estimated η in situ η

18 Estimated η y = 0.52 x 0.67 R² = F nano (%) y = 0.34 x R² = 0.34 RMSE = % N = 32 Validation using in situ R rs F micro (%) y = 0.49 x R² = 0.38 RMSE = % N = F pico (%) y = 0.36 x R² = 0.23 RMSE = % N = 32 RMSE (FY2012 FY2013) F micro 35.94% 25.15% F nano14.81% 11.34% F pico26.31% 22.29% in situ

19 Problem in F nano Fraction of each size サイズ class (%) (%) Fmicro 40 Fnano 30 Fpico η η

20 Database for in situ data Dataset Spectral radiation, pigments, absorption, backscattering, PAR, NPP, etc... Interface Web based interface (upload, search and download) Output Surface data as a table and profile/spectrum data as files (excel, csv, hdf, etc.)

21 Chl.a NPP HPLC absorption Instruments Station scattering Calibration Cruise Radiation

22 Interface (Data upload) Just put formatted Excel files on this window.

23 Interface 1. Search the data based on date, region of interest, and parameters. 2. Push Download 3. CSV formatted data file is created and automatically downloaded on local PC.

24 Summary: Progress in 2013FY Collected new data (> 50 stations) from the Bering, Chukchi and Japan Seas NPP and PFT algorithms were reconstructed and validated using new data. Some improvements are required (nanoplankton for PFT, vertical distribution for NPP Prototype of database for in situ data was created