 SUBSURFACE WATER ICE MAPPING (SWIM) ON MARS: THERMAL AND NEUTRON DATASETS.  H. G. Sizemore1, A. Pathare1, R. H. Hoover2, N. E. Putzig1, Z. Bain1, C. M. Dundas3, M. T. Mellon4, and the SWIM Team. 1Planetary Science Institute (sizemore@psi.edu), 2Southwest Research Institute, 3U. S. Geologic Survey, 4Cornell University.  Introduction:  The Subsurface Water Ice Mapping (SWIM) project supports an effort by NASA's Mars Exploration Program to determine in situ resource availability [1-2]. We are performing global reconnaissance mapping as well as focused multi-dataset mapping to characterize the distribution of water ice from 60ºS to 60ºN latitude. In 2019, we produced ice consistency maps for the northern hemisphere (0 to 60ºN) from 0-225ºE and 290-360ºE longitude. In 2020, we are extending our mapping into the southern hemisphere (0 to 60ºS) and from 225-290ºE longitude in the northern hemisphere at elevations <+1km. Our maps are being made available on the SWIM Project website (https://swim.psi.edu), and we intend to complete our global mapping by the summer of 2020. Follow us on Twitter @RedPlanetSWIM for project news and product release information. The SWIM Datasets: To search for and assess the presence of shallow ice across our study regions, we are integrating multiple datasets to provide a holistic view of the upper 10s of m of the Martian subsurface. The individual datasets and methods we employ include neutron-detected hydrogen maps (MONS), thermal behavior (TES, THEMIS, MCS), multiscale geomorphology (HiRISE, CTX, HRSC and MOLA), and SHARAD surface and subsurface radar echoes. Consistency Mapping:  For the SWIM 2019 maps, we used the SWIM equation [2-4] to provide a quantitative assessment of how consistent (or inconsistent) the various remote sensing datasets are with the presence of shallow (<5 m) ice. The SWIM Equation yields consistency values ranging between +1 and -1, where +1 means that the data are consistent with the presence of ice, 0 means that the data give no indications of the presence or absence of ice, and -1 means that the data are inconsistent with the presence of ice. Here, we focus on our mapping of ice consistency values for the thermal and neutron data sets, and specifically compare ice table depths derived from TES [5] and MONS [6] to depths independently derived from Mars Climate Sounder (MCS) data [7]. We also discuss new strategies for incorporating thermal datasets into the SWIM framework. For more information on the SWIM project and its techniques and datasets, visit our website and associated presentations at this LPSC: Putzig et al. (summary), Perry et al. (SWIM methods), Morgan et al. (surface radar reflectivity), Petersen et al. (radar subsurface mapping), Baker et al. (geomorphic mapping), and Bain et al. (focus regions). SWIM 2019 Approach & Results: Apparent thermal inertia (ATI) exhibits diurnal and seasonal variations as a result of heterogeneity within the field of view of a given bolometer or thermal camera [8, 9]. Observed seasonal variations in ATI can be compared to predictive thermal models to distinguish vertical layering and horizontal mixing of varying types of materials (e.g., dust, sand, duricrust, and 'rock,' where rock is thermally indistinguishable from ice and icecemented soil). Using global maps of TES ATI and model comparison techniques developed by [9, 10], we created a global heterogeneity map that identified the best fit two-component layer model of the shallow subsurface for each map pixel (Fig. 1). At 1.25° per pixel, this global layering map constituted a substantial improvement in spatial resolution over previous heterogeneity maps [9] and it formed the basis for the northern hemisphere thermal ice consistency map currently available on the SWIM website [5]. SWIM 2020 Methods: An unexpected challenge in the 2019 SWIM project was the failure of the TES layering analysis to detect extended ice deposits in the latitude range 45°-60° N. Widespread ice is expected in this region based on predictive ice-stability simulations [e.g. 11], analysis of leakage neutron spectra [e.g. 6], and recent efforts to identify thermal signatures of ice in MCS data [7]. MONS data were included in the  Figure 1. Two-layer model matches derived from TES ATI [Hoover et al., 2019].   combined 2019 SWIM ice confidence map specifically to capture the geographic extent of the shallow midlatitude ice in the final product. In 2020, we are adding a new layer to the SWIM products to incorporate the MCS results. We are also undertaking a variety of activities to quantify systematic differences between the TES, MONS, and MCS datasets and improve the predictive capability of the combined ice confidence map at latitudes equatorward of 60°.  Ice-table depth comparisons: TES, MCS, and MONS analysis all produce independent estimates of the depth of the ice-table, or the top of the ice. Figure 2 shows dry layer thicknesses derived from Pathare et al. [6] based on an assumption of uniform soil porosity and grain density. Variations on this map can be generated by making soil property assumptions guided by the TES layer matching map or other maps of thermal inertia, for self-consistent comparisons. In the lead up to the Phoenix landing, ice-table depth estimates from a variety of thermal techniques were compared to neutron-derived depths to aid landing site selection [12]. Subsequent trenching at the Phoenix site allowed for ground truth comparison at one location [13], and there were a variety of efforts to understand depth heterogeneity at different lateral length scales [14, 15, 16]. We are carrying out a similar comparison of depths derived from TES, MCS, and MONS, over a much wider geographic region, using ice-exposing impacts and ice cliffs [17, 18] as multiple proxies for "ground truth." Understanding differences between the thermal and neutron datasets is critical at low latitudes, where lateral heterogeneity in the burial depth and occurrence of subsurface ice is expected to be much more extreme than at Phoenix. 3-layer thermal simulations: The TES layering map (Fig. 1) was produced by matching observed seasonal ATI variations to predictive thermal models that assumed a two-layer subsurface. Over much of the 45°60° N region where TES did not return layering matches consistent with shallow ice, the data indicated the occurrence of a shallow duricrust. We are currently testing a multi-layer thermal model that allows us to predict the thermal behavior of three-layer subsurfaces and explore scenarios in which shallow duricrusts and dust layers partially or fully mask evidence of underlying ice or ice-cemented sand. Preliminary results suggest that veneers of both dust and duricrust can have a masking effect, potentially explaining some differences between datasets, and opening new analysis pathways for TES and THEMIS data. See Perry et al. (SWIM methods) at this LPSC for more detail on improvements to the Marstherm code base. Future Work: We are currently generating new thermal ice consistency maps spanning both hemispheres of Mars (-60° to 60° N), along with global comparisons of ice table depths derived from SWIM  TES analysis [5], independent MCS analysis [7], and MONS analysis [6]. We are also carrying out focused parameter studies comparing the thermal behavior of specific 2- and 3-layer surfaces. These simulations are aimed at elucidating the role of non-ice heterogeneities in analysis of thermal datasets and will be incorporated into map interpretations presented at the 51st LPSC. Acknowledgments: The SWIM project is supported by NASA through JPL Subcontracts 1611855/1639821. References: [1] Morgan et al., Mapping Water Ice on Mars: Human Mission Resources and Climatic Implications, submitted to Nature Astronomy. [2] Putzig et al. (2019) Ninth Int. Conf. Mars, no. 6427. [3] Perry et al. (2019) LPSC 50, no. 3083. [4] Putzig et al. (2019) LPSC 50, no 2087. [5] Hoover et al. (2019) LPSC 50, no. 1679. [6] Pathare et al. (2018), Icarus 301, 97-116. [7] Piqueux et al. (2019), Geophys. Res. Let., doi:10.1029/2019GL083947 [8] Mellon et al. (2008), The thermal inertia of the surface of Mars, in The Martian Surface, Cambridge University Press, p. 399. [9] Mellon, M. T., & Putzig, N. E. (2007). LPSC 38, no 2184. [10] Putzig et al. (2014), Icarus 230, 64-76. [11] Mellon et al. (2004), Icarus 169, 324-340. [12] Mellon et al. (2008), JGR Planets, doi: 10.1029/2007JE003067. [13] Mellon et al. (2009), JGR Planets, doi: 10.1029/2009JE003417. [14] Bandfield, Nature 447, 64-67. [15] Sizemore et al. (2009), Icarus 199, 303-309. [16] Sizemore et al. (2010), JGR Planets, doi:10.1029/2009JE003414. [17] Dundas et al. (2014) JGR Planets 119, 109-127. [18] Dundas et al. (2018), Science 359, 199-201.  Figure 2. Dry layer thickness (ice table depth) derived from  Pathare et al. (2018) assuming globally uniform soil porosity and grain density.    
