The goal of our proposal is to assist the development of long-term satellite data products suitable for climate research. The need for high quality climate data originates from the assessment that the Earth's global temperature trend due to increase in CO2 is about 0.1K/decade. However larger changes can be seen on decadal scales in satellite data. These changes may be caused by low-frequency natural variability, such as ocean or solar variability. A single point measurement from a satellite infrared instrument typically has an accuracy of about 1K. Thus, large samples and statistical methods must be brought to bear. Currently, most climate data productions are reduced to means and standard deviations within predefined grid cells. However climate change is driven by many physical processes and the resulting climate variables have complex probability distributions, in contrast to the Gaussian distributed standard errors of measurements, which are fully characterized by the means and standard deviations. Given this, to what extent is the use of means and standard deviations useful or correct? What information is lost? On the user end, concern arises from the way the data are used to compare models output to observations. Commonly used methods of identification of climate variability based on linear and parametric approaches such as regression, correlation analyses and band-pass filtering do not properly take into account non-linear and non-stationary nature of climate variability and impose strong, unnecessary demands on accuracy and precision of data. Our proposal addresses the following problem formulated in the AO: "Estimating, validating, and conveying systematic errors in long-term Earth system science data records". We investigate the uncertainties and errors in the construction and use of climate data records. First, we will examine the validity of means and standard deviations as a basis for climate data products. We will explore the conditions under which these two simple statistics are inadequate summaries of the underlying empirical probability distributions by contrasting them with a nonparametric, method called Entropy-Constrained Vector Quantization technique designed to preserve the statistical properties of the underlying data and to provide a better way to summarize large volumes of remote sensing data for climate studies. Second, we will carry out in-depth analyses of the properties of long-term data sets using data-adaptive methods, with a focus on detecting systematic errors caused by the use of linear and parametric assumptions. We will investigate applications of 1) Empirical Mode Decomposition, which accounts for non-linearity and non-stationarity without parametric or functional assumptions, and 2) Recurrent Synchronization to investigate how averaging and filtering influence phase relationships between two data sets. We will use the Level 1B data and investigate the procedure of production of the Level 3 out of Level 2 (retrieved) data based on the measurements by the Atmospheric Infrared Sounder and Atmospheric Microwave Sounding Unit (AIRS/AMSU-A), a set of two instruments operating in the infrared and microwave regions on Aqua satellite. The instruments have been in operation about eight years and are projected for 14 years of operation thus providing a sufficiently long time series of data. The proposed research and resulting technical recomendations will pave the way to improve data products that preserve the statistical information obtained in the measurements. It will significantly contribute to answering the critical questions on accuracy and time span of data that are sufficient to reliably evaluate climate variability using spacecraft data. This will be beneficial for the NASA climate change research overall, and specifically for formulation and evaluation of science goals for NASA missions, such as the CLARREO mission recommended by the Decadal Survey.
Project PI: Alexander Ruzmaikin/Jet Propulsion Laboratory
Jet Propulsion Laboratory M/S 169-506 4800 Oak Grove Drive Pasadena, CA 91109
Phone: (818) 393-3953
Fax: (818) 354-8895
Nothing to see here at the moment. Check back later.
Log in to start a discussion.