Remote Sensing for Climate Analysis

NASA’s new Surface Biology and Geology (SBG) mission will launch in late 2026 and carry a hyperspectral imager to observe Earth’s surface at high resolution (~30 meter) in the visible and thermal regions of the electromagnetic spectrum. Daily data volume is expected to be 2.5 to 5 petabytes. The mission’s science objectives include understanding active surface changes, snow and ice accumulation, hazard risks, changing land use, plant physiology, and terrestrial and aquatic ecosystems. To meet these objectives, geophysical properties of Earth’s surface must be inferred from observed spectra. Spectra are related to surface states via physical forward models embedded within inference algorithms. These forward models are computationally demanding, and will require emulation in order to keep up with data flow. In this talk we introduce a forward model emulator for SBG using a new method for fitting covariance parameters of Gaussian Processes, called Kernel Flows (KF; Owhadi and Yoo, 2019). KF uses mini-batch stochastic gradient descent and cross-validation to achieve robust estimates in a computationally efficient manner. KF has deep connections to neural networks when viewed through the lens of decision theory. The innovation of this work is in the computational implementation of the algorithm, and its application in the remote sensing context.

Sea ice involves a wide range of scales making measuring the evolution and implementing models to predict its future state very complex. In this talk, I will give a brief summary of remote sensing techniques for sea ice, the state of the Arctic sea ice, and the importance of measuring small-scale ice features. New opportunities for Arctic-wide observations of evolving summer sea ice conditions are available with the launch of earth observing satellites with higher-resolution capabilities and continuous measurements. Analysis of ICESat-2 and Sentinel-2 data reveal a three dimensional view of the evolving ice cover, giving us insights on sea ice summer melt on a basin scale. We also explore new techniques for observing and tracking individual ice floes to describe a relationship between dynamical processes and floe size distributions for the first time. I will end by discussing the need to synthesize sea ice observations from varying scales, instruments, methodologies, and disciplines for an inclusive view of the sea ice cover.

Two-dimensional sea surface height (SSH) reconstructions from satellite altimeter observations are an important dataset for studying upper ocean dynamics. Reconstructing the full SSH field is challenging since observations are currently only made along one-dimensional tracks widely spaced in time and space, leaving large gaps unobserved. To date, optimal interpolation has been used, yielding a SSH reconstruction accurate on large scales across much of the ocean. However, in regions with energetic mesoscale turbulence, this reconstruction shows significant errors, with some mesoscale eddies missed entirely and others distorted. These errors could impact our physical interpretations of ocean dynamics affecting, for example, such higher-order statistical metrics as the oceanic kinetic energy spectra and spectral energy fluxes.

A successful reconstruction method must account for the non-linear dynamics governing SSH evolution. We attempt to incorporate non-linear dynamics into the interpolation using deep learning. 

Why deep learning? Deep learning models can learn non-linear mappings from input data to a desired output given sufficient training examples, so it is plausible that such a model could be trained to learn the dynamics of mesoscale ocean turbulence. We thus hypothesize that a neural network can be trained to recognize and reconstruct patterns of mesoscale turbulence even from relatively sparse observations.

Here, we design and train a neural network for reconstructing SSH from real-world satellite observations with active mesoscale turbulence. We augment the SSH data with observations of other physical quantities which, from the physics, we expect to influence SSH dynamics.

Our results are promising, with a lower RMS error than optimal interpolation when tested against independent observations, a higher effective spatial resolution, and qualitatively more physically realistic mesoscale features. Preliminary results from applying our method to a region of the ocean with energetic mesoscale turbulence suggest our new SSH map could lead to a significant re-estimation of the kinetic energy associated with surface ocean currents when compared to previous satellite altimetry studies, a result that would have implications for global climate if borne out on a global scale. We discuss the limitations and challenges of our methodology, its advantages relative to traditional interpolation and data assimilation approaches, and our ideas to improve upon and apply this work in the future.

Over the past three decades, observations have shown that both the Antarctic and Greenland Ice Sheets have been losing mass at a fast pace. Glaciers and ice sheets have become today the largest contributors to sea level rise, but their contribution over the next century remains a key uncertainty in sea level rise projections. Understanding and reducing these uncertainties to improve the representation of past and future behavior of the ice sheets and their interactions with the other components of the Earth system remains scientifically and technically challenging.

The number of remote-sensing observations of polar ice sheets has followed an exponential growth over the past decade. These new observations offer a great opportunity for ice flow models to calibrate model parameters, validate simulation results and investigate physical processes for which only a limited number of direct observations exist. I will show a few examples of how such observations can be used in ice flow models to better understand processes poorly represented, better capture the observed trend in ice mass loss and ultimately improve ice sheet projections, using data assimilation and parameter estimate. The talk will conclude with some future research directions and new challenges.

While the early 21^st century has seen dramatic changes in climate, it has also been the golden age of satellite observations of the Earth’s climate. During this period, climate has been observed from space with revolutionary detail and, in particular, climate change has been monitored in unprecedented ways. Satellite data has played a key role in weather and climate science by, among other contributions, helping to dramatically improve the accuracy of numerical weather prediction, leading to the discovery of new phenomena such as new source regions of atmospheric gravity waves, and monitoring key trends in our climate such as the increase of sea level and carbon dioxide, and the decrease of Arctic sea ice. In this presentation, some of the most significant aspects of the global climate in the 21^st century, focusing on the dramatic changes that we have been witnessing, will be briefly discussed. Three critical issues will be presented in detail as examples of areas where modern statistics and data science can potentially play a key role.