top of page
A multi-modal approach for mapping and identification of moisture and salts
-Fort Brockhurst-


Salt weathering is a major issue for heritage buildings, causing chemical and mechanical degradation. Salts are dissolved in rain and ground water, entering cracks and pores where they crystallise as the solution dries. Further wetting and drying of these deposited salt crystals, plus the accumulation of further salt crystal deposits, can cause a continuous cycle of damage.  


The traditional approach to monitoring salt formation involves physical removal of salt deposits for identification. However, this can be challenging for many reasons. Access to out of reach areas may be limited without special safety and sampling equipment, and analysis is restricted to only the sampled area. Additionally, for the precise examination of the hydration state of the collected salt samples, it is necessary to stabilise the hydration environment of the sample. The aim of this study was to assess a completely new approach to the identification of salts on historic buildings. Our approach allows in situ and non-invasive salt identification on historic buildings- and surveying of large areas- by coupling remote spectral imaging and sensing with machine learning.  

Salt blooms on a building causing paint damage

Addressing the challenge 


Short wave infrared (SWIR) remote spectral imaging and Raman remote sensing systems were used to conduct in situ analysis at Fort Brockhurst, an English Heritage property in Hampshire. The walls of the west-side caponiers were of particular concern as they are underground and against the water filled moat on their outside surface. 

Areas which were scanned using the SWIR imaging system

First, we scanned large areas using our long-range standoff short-wave infrared (SWIR) spectral imaging system. We then used machine learning to automatically “cluster” pixels with similar spectra into groups. The cluster map can be visualised as a false colour map with each cluster represented by a unique colour. Since spectral similarities suggest material similarities, this step offers the ability to monitor salt and moisture spatial distributions.


(left) Colour image of a wall at Fort Brockhurst where remote multi-modal examination was undertaken. 

(right) SWIR cluster map showing the moisture and salts variation and distribution across the examined area. 

Raman spectra were taken from areas which were representative of each cluster, offering precise identification of the salts in their different hydration states. This confirmed the presence of highly damaging sodium sulphate in two hydration states- mirabilite and thenardite. 

SWIR false colour cluster maps (left images) alongside the corresponding SWIR mean cluster spectral information (middle) and Raman spectra (right images). Images a, b, and c show only the presence of moisture and calcite (used to whitewash the walls). Images d, e and f confirm the presence of mirabilite and thenardite. 

Making a difference 

The results show that this new method of remote sensing coupled with machine learning works well for the in situ and non-invasive monitoring and identification of salts in historic buildings. The identification of salts in various hydration states can be performed, whilst avoiding the obstacles that sampling introduces. In addition, monitoring of salts and moisture spatial distribution can be applied across the walls of heritage buildings, to show areas which are at particular risk of damage. Monitoring can also be conducted over different time periods. Data can be collected during return visits throughout the year to assess and monitor hydration stability in different weather conditions and seasons.


Kogou, S., Li, Y., Cheung, S., Liang, H., Thickett, D, Liggins, F. and Butler, L. 2021. Multimodal remote sensing for the monitoring and identification of moisture and salts on heritage buildings. In: H. Liang and R. Groves, eds., Optics for Arts, Architecture, and Archaeology Viii. Proceedings of SPIE, 11784. Washington: SPIE. ISBN 9781510644021

bottom of page