卫星遥感反演叶绿素(Chla)产品在海洋学研究中具有极其重要的意义,但现有不同的卫星遥感反演算法获得Chla产品的区域适用性仍有诸多不确定性,尤其是在光学特性复杂的近岸水体。
本研究比较了目前三种主流算法(OCM3、GSM、GIOP)的精度及其在南海北部水体的适用性,结果表明,在SEATS(Southeast Asian Time-series Study station,18oE,116oE)站、夏季珠江口水华水体,三种算法产品的时空分布具有很好的一致性,但在近岸水体,三种算法产品存在明显差异,与实测结果相比,OC3M算法的对数均方根误差RMSE约0.4,平均百分比误差约115%(Chla范围0.05-10.41mg /m3,n=114); GIOP算法(采用算法默认参数)平均百分比较误差较高,为329%,调整区域化参数后,尤其对于近岸水体,三种算法反演精度均得到了较大程度的提高(Table 1;Fig. 1)。
该研究成果已发表于Biogeosciences, 111, 269–280, 2014。
(Shang, et al., On the consistency of MODIS chlorophyll a products in the northern South China Sea, Biogeosciences , 111, 269–280, 2014)
表1 Error statistics between derived
and in situ Chl.
Algorithm |
R2 |
e (%) |
RMSE |
|
N |
n |
MODIS Rrs derived (in situ Chl=0.05-10.41 mg/m3, mean=0.87, std=1.36) |
OC3M |
0.39 |
118 |
0.380 |
0.145 |
114 |
114 |
OC3M* |
0.40 |
63 |
0.290 |
0.015 |
114 |
114 |
GSM |
0.35 |
113 |
0.400 |
0.069 |
114 |
112 |
GSM* |
0.29 |
76 |
0.329 |
0.029 |
114 |
112 |
GIOP |
0.23 |
329 |
0.600 |
0.340 |
114 |
111 |
GIOP* |
0.32 |
111 |
0.396 |
0.081 |
114 |
110 |
MODIS Rrs derived (<50 m) (chl="0.09-10.41" mg/m3, mean=1.27, std=1.70) |
OC3M |
0.32 |
155 |
0.433 |
0.225 |
64 |
64 |
OC3M* |
0.32 |
75 |
0.330 |
0.012 |
64 |
64 |
GSM |
0.28 |
147 |
0.438 |
0.163 |
64 |
63 |
GSM* |
0.24 |
80 |
0.356 |
-0.002 |
64 |
63 |
GIOP |
0.16 |
441 |
0.692 |
0.464 |
64 |
61 |
GIOP* |
0.26 |
131 |
0.431 |
0.108 |
64 |
63 |
In situ Rrs derived (Chl=0.03-51.15mg/m3, mean=2.89,std=6.63) |
OC3M |
0.81 |
111 |
0.363 |
0.132 |
192 |
192 |
GSM |
0.85 |
94 |
0.342 |
0.086 |
192 |
174 |
GIOP |
0.13 |
256 |
0.548 |
0.163 |
192 |
160 |
*
N is the number of
R
rs data input, while
n is the number of valid retrievals.
图
1 Climatological monthly mean Chl in the northern South China Sea in March, July, October and December from three algorithms: (top) C_OC3M; (middle) C_GSM; (bottom) C_GIOP.