Mawrth Vallis, Mars, classified using the NOAH-H deep-learning terrain classification system. Classified mosaics, Manually Mapped Aeolian Bedforms and derrived gridded density statistics.

  

Dataset description: 

This repository contains data pertaining to the manuscript "Mawrth Vallis, Mars, classified using the NOAH-H deep-learning terrain classification system." submitted to Journal of Maps. 

NOAH-H Mosaics: Mawrth_Vallis_NOAHH_Mosaic_DC_IG_25cm4bit_20230121_reclass.zip

This folder contain mosaics of terrain classifications for Mawrth Vallis, Mars, made by the Novelty or Anomaly Hunter - HiRISE (NOAH-H) deep learning convolutional neural network developed for the European Space Agency (ESA) by SCISYS Ltd. In coordination with the Open University Planetary Environments Group.

These folders contain the NOAH-H mosaics, as well as ancillary files needed to display the NOAH-H products in geographic information software (GIS). Included are two large raster datasets, containing the NOAH-H classification for the entire study area. One uses the 14 descriptive classes of the terrain, and the other with the five interpretative groups (Barrett et al., 2022).

· Mawrth_Vallis_NOAHH_Mosaic_DC_25cm4bit_20230121_reclass.tif

Contains the full 14 class “Descriptive Classes” (DC) dataset, reclassified so that pixel values reflect the original NOAH-H ontology, and not the priority rankings described in Wright et al., (2022) and Barrett et al., (2022b). It is accompanied by all auxiliary files required to view the data in GIS.

· Mawrth_Vallis_NOAHH_Mosaic_IG_25cm4bit_20230121_reclass.tif

Contains the 5 class “Interpretive Groups” (IG) dataset, reclassified so that pixel values reflect the original NOAH-H ontology, and not the priority rankings described in Wright et al., (2022) and Barrett et al., (2022b). It is accompanied by all auxiliary files required to view the data in GIS.

Symbology layer files: NOAH-H_Symbology.zip

This folder contains GIS layer file and colour map files for both the Descriptive Classes (DC) and interpretive Groups (IG) versions of the classification. These can be applied to the data using the symbology options in GIS.

Georeferencing Control points: Mawrth_Vallis_Final_Control_Points.zip

This file contains the control points used to georeferenced the 26 individual HiRISE images which make up the mosaic. These allow publicly available HiRISE images to be aligned to the terrain in Mawrth Vallis, and thus the NOAH-H Mosaic. 

Twenty-six 25 cm/pixel HiRISE images of Mawrth Vallis were used as input for NOAH-H. These are: 

  

PSP_002140_2025_RED

 

PSP_002074_2025_RED

 

ESP_057351_2020_RED

 

ESP_053909_2025_RED

 

ESP_053698_2025_RED

 

ESP_052274_2025_RED

 

ESP_051931_2025_RED

 

ESP_051351_2025_RED

 

ESP_051219_2030_RED

 

ESP_050217_2025_RED

 

ESP_046960_2025_RED

 

ESP_046670_2025_RED

 

ESP_046525_2025_RED

 

ESP_046459_2025_RED

 

ESP_046314_2025_RED

 

ESP_045536_2025_RED

 

ESP_045114_2025_RED

 

ESP_044903_2025_RED

 

ESP_043782_2025_RED

 

ESP_043637_2025_RED

 

ESP_038758_2025_RED

 

ESP_037795_2025_RED

 

ESP_037294_2025_RED

 

ESP_036872_2025_RED

 

ESP_036582_2025_RED

 

ESP_035804_2025_RED

NOAH-H produced corresponding 25 cm/pixel rasters where each pixel is assigned a terrain class based on the corresponding pixels in the input HiRISE image. To mosaic the NOAH-H rasters together, first the input HiRISE images were georeferenced to the HRSC basemap (HMC_11E10_co5) tile, using CTX images as an intermediate step. High order (spline, in ArcGIS Pro 3.0) transformations were used to make the HiRISE images georeference closely onto the target layers. 

Once the HiRISE images were georeferenced, the same control points and transformations were applied to the corresponding NOAH-H rasters. To mosaic the georeferenced NOAH-H rasters the pixel values for the classes needed to be changed so that more confidently identified, and more dangerous, classes made it into the mosaic (see dataset manuscript for details. 

To produce a HiRISE layer which fits the NOAH-H classification, download one of the listed HiRISE images from https://www.uahirise.org/,  Select the corresponding control point file from this archive and apply a spline transformation through the GIS georeferencing toolbar. 

Manually Mapped Aeolian Bedforms: Mawrth_Manual_TARs.zip

The manually mapped data was produced by Fawdon, independently of the NOAH-H project, as an assessment of “Aeolian Hazard” at Mawrth Vallis. This was done to inform the ExoMars landing site selection process. 

This file contains two GIS shape files, containing the manually mapped bedforms for both the entire mapping area, and the HiRISE image ESP_046459_2025_RED where the two datasets were compared on a pixel scale. The full manual map is offset slightly from the NOAH-H, since it was digitised from bespoke HiRISE orthomosaics, rather than from the publicly available HiRISE Red band images. It is suitable for comparison to the NOAH-H data with 100m-1km aggregation as in figure 8 of the associated paper. It is not suitable for pixel scale comparison. 

The map of ESP_046459_2025_RED was manually georeferenced to the NOAH-H mosaic, allowing for direct pixel to pixel comparisons, as presented in figure 6 of the associated paper. 

Two GIS shape files are included: 

· Mawrth_Manual_TARs_ESP_046459_2025.shp

· Mawrth_Manual_TARs_all.shp

Containing the high fidelity data for ESP_046459_2025, and the medium fidelity data for the entire area respectively. The are accompanied by ancillary files needed to view them in GIS.

Gridded Density Statistics 

This dataset contains gridded density maps of Transverse Aeolian Ridges and Boulders, as classified by the Novelty or Anomaly Hunter – HiRISE (NOAH-H). The area covered is the runner up candidate ExoMars landing site in Mawrth Vallis, Mars. These are the data shown in figures; 7, 8, and S1. Files are presented for every classified ripple and boulder class, as well as for thematic groups. These are presented as .shp GIS shapefiles, along with all auxiliary files required to view them in GIS. Gridded Density stats are available in two zip folders, one for NOAH-H predicted density, and one for manually mapped density. 

NOAH-H Predicted Density: Mawrth_NOAHH_1km_Grid_TAR_Boulder_Density.zip

Individual classes are found in the files: 

· Mawrth_NOAHH_1km_Grid_8TARs.shp

· Mawrth_NOAHH_1km_Grid_9TARs.shp

· Mawrth_NOAHH_1km_Grid_11TARs.shp

· Mawrth_NOAHH_1km_Grid_12TARs.shp

· Mawrth_NOAHH_1km_Grid_13TARs.shp

· Mawrth_NOAHH_1km_Grid_Boulders.shp

Where the text following Grid denotes the NOAH-H classes represented, and the landform classified. E.g. 8TARs = NOAH-H TAR class 8.

The following thematic groups are also included: 

· Mawrth_NOAHH_1km_Grid_8_11continuousTARs.shp

· Mawrth_NOAHH_1km_Grid_12_13discontinuousTARs

· Mawrth_NOAHH_1km_Grid_8_10largeTARs.shp

· Mawrth_NOAHH_1km_Grid_11_13smallTARs.shp

· Mawrth_NOAHH_1km_Grid_8_13AllTARs.shp

When the numbers denote the range of NOAH-H classes which were aggregated to produce the map, followed by a description of the thematic group: “continuous”, “discontinuous”, “large”, “small”, “all”. 

Manually Mapped Density Plots: Mawrth_Manual_1km_Grid.zip

These GIS shapefiles have the same format as the NOAH-H classified ones. Three datasets are presented for all TARs (“_allTARs”), Continuous TARs (“_con”) and Discontinuous TARs (“_dis”)

· Mawrth_Manual_1km_Grid_AllTARs.shp

· Mawrth_Manual_1km_Grid_Con.shp

· Mawrth_Manual_1km_Grid_Dis.shp

Related public datasets:

The HiRISE images discussed in this work are publicly available from https://www.uahirise.org/. and are credited to NASA/JPL/University of Arizona. 

HRSC images are credited to the European Space Agency; Mars Express mission team, German Aerospace Center (DLR), and the Freie Universität Berlin (FUB). They are available at the ESA Planetary Science Archive (PSA) https://www.cosmos.esa.int/web/psa/mars-express and are used under the Creative Commons CC BY-SA 3.0 IGO licence.

SPATIAL DATA COORDINATE SYSTEM INFORMATION

All NOAH-H files and derivative density plots have the same projected coordinate system:

“Equirectangular Mars”

- Projection: Plate Carree

- Sphere radius: 3393833.2607584 m

SOFTWARE INFORMATION

All GIS workflows (georeferencing, mosaicking) were conducted in ArcGIS Pro 3.0. NOAH-H is a deep learning semantic segmentation software developed by SciSys Ltd for the European Space Agency to aid preparation for the ExoMars rover mission.