Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2022 Nov 1;12(1):18395.
doi: 10.1038/s41598-022-22087-2.

Gap-filling of ocean color over the tropical Indian Ocean using Monte-Carlo method

Aditi Modi  1   2 , M K Roxy  3 , Subimal Ghosh  4   5
Affiliations

Gap-filling of ocean color over the tropical Indian Ocean using Monte-Carlo method

Aditi Modi et al. Sci Rep. .

Abstract

Continuous remote-sensed daily fields of ocean color now span over two decades; however, it still remains a challenge to examine the ocean ecosystem processes, e.g., phenology, at temporal frequencies of less than a month. This is due to the presence of significantly large gaps in satellite data caused by clouds, sun-glint, and hardware failure; thus, making gap-filling a prerequisite. Commonly used techniques of gap-filling are limited to single value imputation, thus ignoring the error estimates. Though convenient for datasets with fewer missing pixels, these techniques introduce potential biases in datasets having a higher percentage of gaps, such as in the tropical Indian Ocean during the summer monsoon, the satellite coverage is reduced up to 40% due to the seasonally varying cloud cover. In this study, we fill the missing values in the tropical Indian Ocean with a set of plausible values (here, 10,000) using the classical Monte-Carlo method and prepare 10,000 gap-filled datasets of ocean color. Using the Monte-Carlo method for gap-filling provides the advantage to estimate the phenological indicators with an uncertainty range, to indicate the likelihood of estimates. Quantification of uncertainty arising due to missing values is critical to address the importance of underlying datasets and hence, motivating future observations.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1
Figure 1
Missing values in ESA chlorophyll and mean OLR in the tropics during boreal summer. (a) Number of missing pixels (in percentage) in 8-day composites of ESA OC-CCI chlorophyll data from 1998–2019 over the tropical oceans during boreal summer (June–September). Gap-free pixels are indicated in white. The pixels having more than 25% of missing values are shown in black; and (b) Climatological map of OLR (in W/m2) during summer (June–September) for the tropical oceans for the period 1998–2019. The regions in dark blue are associated with a weaker convection and those represented in yellow represent strong convection. This figure is created using Python 3.9.10 software (https://docs.python.org/release/3.9.10/).
Figure 2
Figure 2
Annual and seasonal maps showing the count of missing pixels (in percentage) in the ESA merged chlorophyll data over the tropical Indian Ocean. Number of missing pixels (in percentage) in the 8-day composites of 22 years of ESA OC-CCI chlorophyll data from 1998–2019 in the Indian Ocean during the months (a) January–December; (b) June–September; (c) December–February. The seasonal maps of (b) and (c) indicate the seasonal contribution to the total number of observed pixels in (a). Gap-free pixels are indicated in white. The pixels exceeding 25% of missing values are marked as black. This figure is created using Python 3.9.10 software (https://docs.python.org/release/3.9.10/).
Figure 3
Figure 3
Schematic of the algorithm to fill the gaps of missing chlorophyll concentrations. The daily chlorophyll fields are available at a spatial scale of 4 km × 4 km for the period 1998–2019 by ESA OC-CCI. 8-day composites are prepared from the daily fields and re-gridded to 1 degree to reduce gaps. Then the gaps are filled in two steps: (a) Linear Interpolation; (b) Monte-Carlo Multiple Imputation. The arrows depict the sequence of the algorithm (refer to “Methods” section for details). Only pixels with missing data are reconstructed. The schematic is adapted from Racault et al..
Figure 4
Figure 4
Spatial maps of missing pixels (in percentage) before and after applying the gap-filled methodology. Spatial maps showing missing number of pixels (in percentage) in the 8-day composites of chlorophyll during 1998–2019 in the (a) original chlorophyll data from ESA OC-CCI, (b) after the first step of interpolation, and (c) after the final step of Monte-Carlo multiple imputation. The regions in white indicate gap-free pixels. This figure is created using Python 3.9.10 software (https://docs.python.org/release/3.9.10/).
Figure 5
Figure 5
Annual cycle of chlorophyll and its spatial distribution in the Arabian Sea and Bay of Bengal for the original and gap-filled datasets. Climatological annual cycle of reconstructed chlorophyll (8-day composites) for the period 1998–2019 in the (a) Arabian Sea [60°E–70°E, 8°N–16°N], and (b) Bay of Bengal [85°E–95°E, 8°N–16°N]. Light Blue line indicates ESA v4.2 satellite chlorophyll (original data); pink line represents the mean of gap-filled chlorophyll datasets (reconstructed data); and violet line represents the climatology of the gap-filled annual cycle reconstructed by Levy. The boxplot overlaid on the time series represents the range of values between the 25th and the 75th percentile. The black dots represents the outliers. Spatial distribution of chlorophyll concentration (in mg/m3) in the tropical Indian Ocean for the period 1998–2019 in (c) satellite chlorophyll, (d) reconstructed chlorophyll using our proposed methodology, and (e) Levy’s reconstructed dataset. This figure is created using Python 3.9.10 software (https://docs.python.org/release/3.9.10/).
Figure 6
Figure 6
Phenological Indicators derived for the annual cycle of chlorophyll in the Arabian Sea. Bloom initiation and bloom peak estimated for the 8-day composites of the reconstructed datasets showing (a) for the year 2018, (b) chlorophyll climatology during 1998–2019. The phenological indices shown here are computed for a grid location in the central Arabian Sea [64°E, 11°N]. Light Blue line indicates ESA v4.2 satellite chlorophyll (original data); dotted pink line represents the mean of gap-filled chlorophyll datasets (reconstructed data). The bloom initiation is indicated by solid circles and bloom peak by solid squares in same color as the data. The uncertainty in bloom initiation and peak timings as derived from our gap-filled datasets are represented by a horizontal solid line (pink). The horizontal dashed lines in (b) represents the annual median value of both the datasets. This figure is created using Python 3.9.10 software (https://docs.python.org/release/3.9.10/).
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Banse K. Seasonality of phytoplankton chlorophyll in the central and northern Arabian sea. Deep Sea Res. Part A Oceanogr. Res. Papers. 1987;34:713–723. doi: 10.1016/0198-0149(87)90032-X. - DOI
    1. Kumar, S. P., Narvekar, J., Nuncio, M., Gauns, M. & Sardesai, S. What Drives the Biological Productivity of the Northern Indian Ocean? in Indian Ocean Biogeochemical Processes and Ecological Variability 33–56 (American Geophysical Union (AGU), 2013). 10.1029/2008GM000757.
    1. Kumar SP, et al. Physical forcing of biological productivity in the Northern Arabian Sea during the Northeast Monsoon. Deep Sea Res. Part II. 2001;48:1115–1126. doi: 10.1016/S0967-0645(00)00133-8. - DOI
    1. Schott FA, Mccreary JP. The monsoon circulation of the Indian Ocean. Prog. Oceanogr. 2001;51:1–123. doi: 10.1016/S0079-6611(01)00083-0. - DOI
    1. Shankar D, Vinayachandran PN, Unnikrishnan AS. The monsoon currents in the north Indian Ocean. Prog. Oceanogr. 2002;52:63–120. doi: 10.1016/S0079-6611(02)00024-1. - DOI

Publication types