Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluating Terrain-Dependent Performance for Martian Frost Detection in Visible Satellite Observations

Gary Doran, Serina Diniega, Steven Lu, Mark Wronkiewicz, Kiri L. Wagstaff

Abstract

Seasonal frosting and defrosting on the surface of Mars is hypothesized to drive both climate processes and the formation and evolution of geomorphological features such as gullies. Past studies have focused on manually analyzing the behavior of the frost cycle in the northern mid-latitude region of Mars using high-resolution visible observations from orbit. Extending these studies globally requires automating the detection of frost using data science techniques such as convolutional neural networks. However, visible indications of frost presence can vary significantly depending on the geologic context on which the frost is superimposed. In this study, we (1) present a novel approach for spatially partitioning data to reduce biases in model performance estimation, (2) illustrate how geologic context affects automated frost detection, and (3) propose mitigations to observed biases in automated frost detection.

Evaluating Terrain-Dependent Performance for Martian Frost Detection in Visible Satellite Observations

Abstract

Seasonal frosting and defrosting on the surface of Mars is hypothesized to drive both climate processes and the formation and evolution of geomorphological features such as gullies. Past studies have focused on manually analyzing the behavior of the frost cycle in the northern mid-latitude region of Mars using high-resolution visible observations from orbit. Extending these studies globally requires automating the detection of frost using data science techniques such as convolutional neural networks. However, visible indications of frost presence can vary significantly depending on the geologic context on which the frost is superimposed. In this study, we (1) present a novel approach for spatially partitioning data to reduce biases in model performance estimation, (2) illustrate how geologic context affects automated frost detection, and (3) propose mitigations to observed biases in automated frost detection.
Paper Structure (8 sections, 6 figures, 1 table)

This paper contains 8 sections, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background
  3. Methodology
  4. Creating a Machine Learning Ready Dataset
  5. Spatial Validation
  6. Model Training and Evaluation
  7. Results
  8. Conclusion and Future Work

Figures (6)

  • Figure 1: Previously studied northern mid-latitude frost sites used for training: A: $64.550^{\circ}\mathrm{N}$, $315.907^{\circ}\mathrm{E}$, B: $58.236^{\circ}\mathrm{N}$, $89.607^{\circ}\mathrm{E}$, C: $63.738^{\circ}\mathrm{N}$, $11.035^{\circ}\mathrm{E}$, D: $42.572^{\circ}\mathrm{N}$, $67.332^{\circ}\mathrm{E}$, E: $56.847^{\circ}\mathrm{N}$, $350.401^{\circ}\mathrm{E}$, F: $59.839^{\circ}\mathrm{N}$, $135.999^{\circ}\mathrm{E}$, G: $64.829^{\circ}\mathrm{N}$, $209.406^{\circ}\mathrm{E}$.
  • Figure 2: Visible indications of frost, including uniform albedo (top left), polygonal features (top right), halos (bottom left), and defrosting marks on dunes (bottom right).
  • Figure 3: Site locations within the HEALPix partition (white grid) of the surface (North pole orthographic projection).
  • Figure 4: Comparison of pixel intensity distributions across all train and test tile pixels.
  • Figure 5: Frost detection recall scores as a function classification threshold for train (left) and test (right) sets.
  • ...and 1 more figures