Rewrite the report in my own words for better understanding (personal use).

Chapter 1 Introduction

Phytoplankton plays an important role in the global carbon cycle. It is responsible for approximately half of the global photosynthetic uptake  of carbon (Field et al., 1998). Chl concentration, a proxy for phytoplankton biomass, can be detected remotely by satellite ocean-color sensors, thus making it possible to observe the Chl distribution on synoptic scale. In response to the importance of phytoplankton as well as the lack of comprehensive data, global ocean-color satellite missions, namely SeaWiFS, MERIS, MODIS-Aqua, POLDER-2/GLI, POLDER/PARASOL, VIIRS-NPP, S-GLI, VIIRS-NPOESS and Sentinel-3, all of which in polar-orbiting configuration, have been designed to acquire and produce high quality global ocean-color data. Q: What is the lifetime of each sensor?
With the proliferation of  global missions, the temporal and spatial coverage of Chl distributions are greatly expanded, while a single polar-orbiting ocean-color satellite does a poor job of sampling the ocean on short time scales due to the effects of clouds, inter-orbit gaps, sun glint or thick aerosols. # proliferate: to increase a lot and suddenly in number. Considering that phytoplankton populations can increase their biomass more than double in a single day under favorable circumstances, it is important to have a better coverage of Chl distributions in order to understand phytoplankton dynamics and their relationship with natural variability.
An important way to improve the daily coverage of the global ocean is to combine or merge data from coincident multiple satellites (e.g. Gregg et al., 1998; Gregg and Woodward, 1998) Data merging is both technically and politically challenging because the ocean color sensors are very complex, with a very small signal relative to noise sources, and  require a massive amount of effort to get high-quality observations.
This report describes the basis for ocean-color data merging. Here data merging refers to the process of combing coincident data from more than one satellite sensor. The main goals are to improve temporal resolution, coverage and, to a lesser extent, improve accuracy.
Note that data merging does not mean the construction of ocean-color time series. Antonie et al.(2005) and Gregg et al.(2002) have argued that time series construction requires that all sensors involved have similar atmospheric and bio-optical algorithms in order to avoid confusing trends with methodological differences. However, data merging does not require similar methods, but rather seeks to improve an overall data set where coincident observations occur.

Chapter 2 Benefits of Merging

The goal of data merging is to merge multi-instrument and multi-year observations of the oceans into consistent daily, global, high-resolution data records based on accurate and uniform calibration and validation over the lifetime of the missions. That is to say, the merger is a natural progression from multi-sensor data inter-comparisons, cross-validations and calibrations.  Q: What are the differences between inter-comparisons, cross-validations and calibrations? The inter-comparisons, cross-validations and calibrations are aimed at identifying and eliminating sensor or algorithm-originated biases and trends (systematic errors) in data, which are impossible to detect solely by comparison with sparse in situ measurements. In other words, they bring multi-instrument time series to a consistent ocean-color baseline. Subsequently, the merger can:
· increase the daily ocean-color global coverage, which facilitates enhanced spatial and temporal resolution of ocean processes; # facilitate: to make something possible or easier.

· expand sampling rate for each location, thus reduce the random errors and improve the statistical confidence when extracting bio-optical parameters;
· allow the data users to access the multi-sensor ocean-color data with a single access point instead of the data holdings among different space agencies;
· provide consistent quality dataset for the users, which helps them to advance science and make educated decisions based on the dataset. Subsequent findings and operational applications derived from the multiple data sources is more convincing compared to those employing the data from only a single mission.
A generic baseline for ocean-color merged products has 3 characteristics:
· Daily global coverage.
· High accuracy – consistent and seamless in space and time.
· High spatial resolution.
Q: What does "binned datasets" mean?
Different applications of the user community have their specific requirements of data accuracy, spatial and temporal resolution, time series length, grid type, output products and operational data delivery. Q: What is the grid type? Therefore, data merge methodology have to match with the user's requirements (See Page 17, Table 2.2). For earth system science, more observations with better signal to noise is of great importance in order to understand marine biogeochemical cycles.

Chapter 3 Coincident Global Ocean-Color Missions

# coincident: happening at the same time.
Global missions as well as smaller scale missions Level-3 product are recommended for data merging. Level-3 is defined as products mapped onto an Earth projection, developed following standard guidelines described in IOCCG report4. # project: to cause a film, image, or light to appear on a screen or other surface: Laser images were projected onto a screen. Level-3 is the preference because:
·  it is easy to use.
·  it is available.
·  one the main advantages of merging is improved spatial coverage, which is inherently defined on an Earth grid. Q: do not understand!

The Level-3 data product information for global missions that overlap in time with at least one other mission at some point during their lifetime are listed in Table 3.1. A listing of non-overlapping global missions as well as all the smaller scale missions is provided in the Appendix. Further information can be obtained from http://www.ioccg.org/sensors_ioccg.html

Chapter 4 Survey of Ocean-Color Data Merging Methods

This chapter contains a list of data merging methodologies used in ocean-color research. These methodologies are all very different, however, they share some features in common:
·  all but one of the methods used Level-3 data and involved only Chl.
·  all methods are error-correcting in nature, and many are bias-correcting. Most of them are statistical methods.
- Statistical Methods
Random error-correcting:
· Binning
· Averaging
· Error-weighted averaging
Bias-correcting:
· Subjective Analysis                                                                                                        

· Blended Analysis
· Optimal Interpolation
· Objective Analysis
· Wavelet Analysis
· Machine Learning Analysis
- Bio-optical Methods
· Spectral bio-ptical modelling
- Numerical Model-Based Methods
· Data assimilation into numerical models

4.1 Binning
The binning method starts with Level-2 data and producing Level-3 data by placing within pre-defined bins. Although all data are treated equally, the Level-3 merged result will be biased in favor of data with the highest native Level-2 resolution. Q:do not understand!

4.2 Averaging

$C_{ij}=\frac{\sum_s C_{ijs}}{\sum_s n_{ijs}}$  --Fomular 4.1

C stands for Chl from sensor s, n is the number of observations from sensor s, ij represents the Level-3 grid point and the summations are over the sensors. Note that any Level-3 product can use this formular by replacing itself with Cij.
This method is simple, total objective, i.e., no sensor data are preferred over others, and computationally fast. If Level-3 grid locations are common among the different sensor products, the method is straightforward. If not, interpolation may be required. Q: do not understand! Also, uncorrected or unrecognized bias in the data can be propagated into the merged field and a poor quality data set can be produced.