ESA CNN - CCI LAKES + Consistency

CCN BACKGROUND AND OBJECTIVES Concise presentation of the background and objectives of the CCN This CCN is one of the Options proposed for the CCI+ Lakes project. Its objective is to investigate and improve the Lake’s inter-ECV consistency, by statistical tests, cross ECV time series analysis, identification of the sources in case of inconsistency and proposing improvements for ECVs where possible. Inconsistent information between ECVs can lead to miss-interpretation and / or to reduced acceptance by the user community. This option CCN will provide recommendations for the production centres for improving improve the quality and consistency of the products and therefore will support to deliver consistent ECVs to the climate modelling group as well as providing improved products to users. The baseline activities will provide prototype products for a number of ECVs, primarily applying existing algorithms individually per variable. The consistency between them needs to be assessed in detail and inconsistencies are possible in many respects, e.g. between lake ice melting time and LWST, or ice covered / free water colour products. Furthermore, some algorithms rely on heritage from other domains, such as the ocean colour based algorithms for lake optical products (CDOM, Turbidity, Chlorophyll). Given the large variability of specific inherent optical properties of lakes we have to expect limitations in accuracy that may lead to inconsistencies. This option CCN will provide deeper analyses for identifying the issues that need to be addressed. The assessment will be done at a global scale, by investigating the time series of the first generated set of Lake ECV products. Feedback will be provided in due time for the final processing.
An important starting activity for this option CCN will be used to consolidate the definition of indicators to assess the consistency. An initial list is provided in this proposal. These indicators will be analysed at the level of grouped Lake Water Types, in order to manage the large number of lakes and to be able to synthesise. The Lake Water Types will be defined by clustering of the Lakes using the calculated ECVs. The indicators/tests for consistency fall into different categories, which concern either algorithmic, limnological or logical inconsistency. While some indicators will be independent of the Lake Water Type, others will strongly depend on it and will need individual definition of expectations. In this initial task we will work on the definition of the Lake Water Types and the Indicators. The UR will be a driving force for these definitions. The methods to be applied for the analysis/tests will be based on time series analysis of the defined indicators/tests which will cover one or more ECV. The work currently performed by Iestyn Woolway will be a valuable input for defining our indicators. In completion to the work performed by Iestyn, we will specifically look into the reasons for the inconsistencies and we will investigate them depending on different aspects, such as lake types or geographical distribution. Besides the investigation of time series for different locations in lakes, we will also look into inconsistencies within lakes and the spatial patterns of consistencies.
Another aspect will be the intercomparison of spatial patterns of the ECVs and geographical distribution. The following activities under this option CCN include the preparation of the data to be analysed, the implementation of the tools for the Water Type Classification and the Indicator calculation, and the application of the those. The analysis of the Indicators will lead to probably a complicated and diverse picture. If we assume ~ 10 different Lake Water Types and 5 different indicators, we will generate ~50 results, whereby each of those will be a time series of a more or less complicated/abstract variable. The interpretation of this dataset is a critical and not easy, but the key task in this optionCCN. It will give indications to issues in one or more Lake ECVs, probably in connection with certain conditions, such as geographic area or timing of the year. We will study exactly those potential issue in order to be able to provide veary indications to the algorithm developers where algorithm need improvements and where gaps might have occured. Where possible, we will even go further and elaborate algorithm improvements. Iterations of the indicator calculation on small subsets of the datasets in order to test hypothesis as well as algorithmic improvements are expected. Another important information to be included are the uncertainties provided with each ECV. If possible, we will provide proposals for the production lines for improvements, but at least the development of an improved quality flagging for inconsistent pixels shall be given. The final recommendations will be provided to the baseline activities in due time to be considered for the final production. Indicators which are relevant for the expected improvement will be calculated for final dataset, in order to close the feedback loop.