Physical retrieval of sea-surface temperature from INSAT-3D imager observations

DescriptionPurposeSpatial resolution (km)10.55–0.75VisibleClouds, surface features1.021.55–1.70Short–wave infraredSnow, ice and water phase in clouds1.033.80–4.00Medium–wave infraredClouds, Fog, Fire4.046.50–7.10Water vapourUpper troposphere water vapour8.0510.30–11.30Long-wave infrared windowCloud top and surface temperature4.0611.50–12.50Long-wave infrared windowLower-tropospheric moisture4.0

Two-dimensional colour histogram plots of the RTTOV simulated and observed INSAT-3D brightness temperature of TIR1 channel, (a) before bias correction and (b) after bias correction.

Two-dimensional colour histogram plots of the RTTOV simulated and observed INSAT-3D brightness temperature of TIR2 channel, (a) before bias correction and (b) after bias correction.

Various steps of the operational INSAT-3D SST retrieval algorithm.

Various steps of the OEM _fsvis SST retrieval algorithm used for INSAT-3D.

Variation of the SST bias and RMSE with respect to (a) iQuam SST, (b) latitude, (c) satellite zenith angle and (d) TCWV.

Table 2.

List of all the SST retrieval algorithms used in this study.

AlgorithmsState vector used,
reduced (R) or full (F)Solution method,
direct (D) or iterative (I)Smoothing applied (Y/N)REG––NOEM _rsvd RDNLSQFDNOEM _fsvd FDNOEM _fsvi FINOEM _fsvis FIY

[i] The algorithms are arranged in order of increasing complexity.

Table 3.

Overall validation statistics of the INSAT-3D imager SST retrievals obtained using OPR, REG, LSQ, OEM _rsvd , OEM _fsvd , OEM _fsvi and OEM _fsvis against iQuam SST.

Bias (K)RMSE (K)SD (K)Median (K)MAD_s (K)

\bar{\frac{\partial SS T_{ret}}{\partial SS T_{true}}}

\bar{DFS}

OPR–0.150.920.91–0.210.89––REG–0.170.700.68–0.190.650.81–LSQ–0.150.980.97–0.150.931.002.00OEM _rsvd 0.140.820.810.140.700.691.23OEM _fsvd 0.230.710.670.210.630.661.13OEM _fsvi 0.190.680.650.160.610.661.13OEM _fsvis 0.060.630.630.030.590.891.69BAK0.480.860.710.470.66––

[i] The statistics of the background SST (BAK) used in these algorithms is also presented. This statistics is obtained from eight months data (October 2017–May 2018). A total of 60314 collocated pairs were used in calculating this statistics.

RMSE and bias plots of SST retrievals, from different methods, as a function of corresponding $χ^{2}$ values. Values of the $χ^{2}$ for the corresponding methods are used for generating these plots. For BAK, $χ^{2}$ values of OEM _rsvd are used.

Table 4.

Validation statistics of the SST retrievals obtained using OPR, REG, LSQ, OEM _rsvd , OEM _fsvd , OEM _fsvi and OEM _fsvis against iQuam SST, for day and night separately.

nDay 36796Night 23518Bias (Median)RMSESD (MAD_s )Bias (Median)RMSESD (MAD_s )(K)(K)(K)(K)(K)(K)OPR–0.26 (–0.32)0.910.87 (0.84)0.02 (–0.01)0.950.95 (0.92)REG–0.19 (–0.20)0.700.67 (0.64)–0.15 (–0.16)0.710.69 (0.67)LSQ–0.21 (–0.20)1.010.99 (0.95)–0.05 (–0.07)0.930.93 (0.90)OEM _rsvd 0.12 (0.10)0.810.80 (0.70)0.17 (0.17)0.840.82 (0.70)OEM _fsvd 0.20 (0.17)0.710.68 (0.64)0.29 (0.25)0.710.65 (0.61)OEM _fsvi 0.16 (0.14)0.680.66 (0.61)0.24 (0.21)0.680.63 (0.60)OEM _fsvis 0.04 (0.0)0.630.63 (0.58)0.01 (0.07)0.640.63 (0.61)BAK0.51 (0.50)0.910.76 (0.72)0.42 (0.43)0.770.64 (0.58)

[i] The statistics of the background SST (BAK) used in these algorithms is also presented. This statistics is obtained from eight months data (October 2017–May 2018).

Spatial maps of the monthly average SST sensitivity, for SST retrieved (a) from OEM _fsvi and (b) OEM _fsvis , for the month of October 2017.

Weekly time series of the error statistics, day and night-time combined, of BAK and OEM _fsvis SST.

Weekly time series of the error statistics, during day-time, of BAK and OEM _fsvis SST.

Weekly time series of the error statistics, during night-time, of BAK and OEM _fsvis SST.

Monthly average SST for the month of October 2017. (a) SST retrieved from INSAT-3D observations using REG, (b) SST retrieved from INSAT-3D observations using OEM _fsvis and (c) OSI-SAF-SST.

Spatial distribution of the improvement parameter, α.

Table 5.

Validation statistics of OEM _rsvd , OEM _fsvd , OEM _fsvi and OEM _fsvis calculated for different values of $χ^{2}$ .

Algorithm

χ^{2}

Bias (K)RMSE (K)SD (K)Median (K)MAD_s (K)N

χ^{2} \leq

10.210.680.650.190.6248024 (∼80%)OEM _rsvd 1

< χ^{2} \leq

20.00.960.96–0.160.877551 (∼12%)2

< χ^{2}

–0.431.541.48–0.711.134739 (∼8%)

χ^{2} \leq

10.240.670.620.230.5949731 (∼82%)OEM _fsvd 1

< χ^{2} \leq

20.210.840.810.110.777836 (∼13%)2

< χ^{2}

0.170.990.98–0.010.812747 (∼5%)

χ^{2} \leq

10.210.640.600.190.5850925 (∼84%)OEM _fsvi 1

< χ^{2} \leq

20.110.800.80–0.020.727005 (∼12%)2

< χ^{2}

0.050.990.99–0.160.752384 (∼4%)

χ^{2} \leq

10.080.560.560.060.5334951 (∼58%)OEM _fsvis 1

< χ^{2} \leq

20.020.660.66–0.040.6215856 (∼26%)2

< χ^{2}

0.060.810.81–0.070.809507 (∼16%)

[i] Values of the $χ^{2}$ for the corresponding methods are used while calculating these statistics.

References

Authors

Metrics

Articles in this issue

DOI: https://doi.org/10.1080/16000870.2019.1657767 | Journal eISSN: 3035-9554

Journal RSS Feed

Language: English

Page range: 1657767 - 1657767

Published on: Jan 1, 2019

Published by: Stockholm University Press

In partnership with: Paradigm Publishing Services

Keywords:

sea-surface temperature,

INSAT-3D,

physical retrieval,

RTTOV

© 2019 Satya P. Ojha, Randhir Singh, published by Stockholm University Press
This work is licensed under the Creative Commons Attribution 4.0 License.

Volume 71 (2019): Issue 1

Physical retrieval of sea-surface temperature from INSAT-3D imager observations

Figures & Tables

Fig. 1.

Fig. 2.

Fig. 3.

Fig. 4.

Fig. 5.

Table 2.

Table 3.

Fig. 6.

Table 4.

Fig. 7

Fig. 8.

Fig. 9.

Fig. 10.

Fig. 11.

Fig. 12.

Table 5.

Paradigm

My account