Attribution model for watercolours assigned to the Costa Rican painter Fausto Pacheco: a chemical and antimicrobial assessment

Introduction

The current research systematically evaluates Fausto Pacheco’s watercolours to provide objective scientific insights^1,2 into his colour palette and sketching underdrawing techniques. Furthermore a deterioration map was created and the main microorganisms colonising the paintings were identified. The ongoing debate surrounding the scientific understanding of his paintings—marked by their irregular quality and lack of dating—creates an ideal context for establishing a coherent attribution process applicable to other similar works of art. While several studies have aimed to differentiate authentic paintings from forgeries through pigment analysis and material assignment^3,4,5,6,7,8, no research has concurrently examined watercolours from a key diagnostic information alongside probabilistic indicators for authorship attribution. To illuminate this area and forge a new path using a multidisciplinary approach, we concentrated on developing a comprehensive database containing technical images and interpretations related to the artist. This resource aims to assist stylistic connoisseurs, conservators, cultural heritage scientists, and materials analysts in addressing attribution questions regarding Pacheco’s watercolours in the future.

To the best of our knowledge, the attribution process in prestigious art objects generally takes place in painting media such as oil paintings, frescoes, and acrylics⁹, and therefore it is a complex task to imagine a direct path towards the authentication of artworks^10,11,12,13. Techniques such as multivariate analysis and machine learning have been employed to evaluate material compositions, stylistic patterns, and other measurable features of artworks, enabling experts to assign probabilistic or numerical weights to different characteristics. These approaches often integrate spectroscopic data and computational tools to identify patterns of authenticity or forgery. However, most existing methods rely on statistical clustering or predictive models rather than explicitly weighting specific features based on their relative importance^14,15,16,17. The method used in this study, which assigns cumulative authenticity probability percentages to attributes such as pigments, brushstrokes, and paper characteristics, represents an innovative approach to systematically quantify attribution probabilities. This experimental framework not only enhances the reproducibility of the results but also bridges the gap between subjective expertise and data-driven validation, making it a novel contribution to the field.

Nevertheless, we conducted a multi-analytical study in watercolours which includes multispectral imaging (MSI) analyses¹⁸, computational tools, along with spectroscopic technologies that are commonly and effectively applied in the field of art conservation and analysis—and microbiological techniques^{19,20,21,22,23,24,25,26}. It is important to bear in mind that this selection of techniques was chosen due to the nature and characteristics found in the objects of study. Using these complementary technologies, we identified materials and the pictorial palette of four artworks, and used this information to compare with a highly probable artwork of the artist (Paisaje con árbol). In other words, we focused on reproducibility patterns and assigned probabilistic indicators to each painting. Moreover, we studied the antimicrobial properties of some pigments (ultramarine blue and yellow ochre) in these artworks.

In our investigation, we analysed five watercolour paintings: Paisaje con árbol, Paisaje con casa campesina and El puente belonging to the Foundation for the Administration of Museums of the Central Bank of Costa Rica (MBCCR, from the Spanish Museos del Banco Central de Costa Rica) and, Paisaje A and Paisaje B belonging to the Museum of Costa Rican Art (MAC, from the Spanish Museo de Arte Costarricense), attributed to the Costa Rican artist Fausto Pacheco Hernández (1899–1966). These artworks were selected for their distinct visual characteristics and historical significance, as mentioned below.

Here we used MSI, a non-destructive approach^5,6,27, in order to create a systematic baseline for comparing these artworks and to be able to monitor changes over time^10,28,29. Nowadays, MSI is integrated with computational tools primarily to detect, analyse, and enhance specific features of artworks, such as underlying layers, hidden details, or material alterations, contributing to both scientific analysis and the simulation of potential restoration interventions^3,30,31. Specifically, with the computational tool called RegionsOfInterest which was developed by our group, it is possible to accurately calculate the light intensity within the regions selected by the user in a high-resolution photograph of the object under study¹⁹. Using the aforementioned software, it was possible to propose pigments used by the artist for later confirmation with spectroscopic techniques.

To assess the systematic reproducibility of the artist’s techniques, our research group developed a software tool called GraphiteComparison in order to compare the amount of graphite in each painting using an IR photograph. It is possible to determine the pictorial palette with spectroscopic techniques in situ, such as X-ray fluorescence (XRF) and Fiber optic reflectance spectroscopy (FORS). XRF detects the chemical elements within a determined material, aiding in the characterisation of specific pigments^29,32,33. Likewise, with FORS the area of interest is illuminated and the backscattered light is collected, obtaining a reflectance graph. The data measured are statistically processed^{29,32,34,35,36,37}.

To learn more about the nature of paper and pigments, invasive techniques such as scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDX)^17,19, Raman spectroscopy³⁸ and Fourier transform infrared-attenuated total reflectance spectroscopy (FTIR-ATR)¹⁹ were used. A signature review based on the qualitative and quantitative analysis of graphical feature, was carried out. The qualitative comparison makes use of a grid for form and spatial relationships. A structural similarity index measure (SSIM) is used for the relative comparison of morphological graphical features (explained in more detail below).

Finally, the biodeterioration capability of seven microorganisms (Aspergillus section Versicolores, Cladosporium tenuissimum, Curvularia asiatica, Candida tropicalis, Clavispora lusitana and Bacillus cereus) isolated from the paintings El Puente, Paisaje A and Paisaje B was determined. The isolates were freely placed in different pigments (ultramarine blue and yellow ochre) to find out what kind of damage they could be causing in the artworks and what kind of nutrients attract them. Preventing the growth of various microorganisms in works of art is a challenge for their conservation, and hence, the importance of determining which elements could be affecting historical objects³⁹.

Historical context

The art movements of the 1920s and 1930s were characterised by bold and transformative approaches that significantly influenced the visual arts. During this period, several new European art movements emerged, including Art Deco, Cubism, and Surrealism⁴⁰. In Costa Rica, a notable artistic movement known as the Nationalist Generation also emerged^41,42. This movement aimed to shift the direction of painting by integrating modern European trends with local artistic expressions. The artists of this movement were dedicated to capturing the national landscape and cultural identity in their work.

Two key figures from the Nationalist Generation were instrumental in shaping the art scene of their time. Teodorico Quirós Alvarado (1897–1977), a Costa Rican architect and painter, led the group and was a prominent figure in the movement⁴¹. Fausto Pacheco Hernández, another significant painter of this era, was renowned for his mastery of watercolour. Pacheco’s transition from oil painting to watercolour reflects his deep connection with the Costa Rican landscape, a theme that pervades his work. His paintings often depict natural scenery and traditional adobe houses^41,42.

Although Pacheco’s work contributes significantly to Costa Rican art, the irregular quality of some of his pieces, coupled with his practice of not dating his paintings, complicates the process of periodising and authenticating his artwork⁴³. Our study addresses these challenges by focusing on consistent technical and material features to support a probabilistic approach to authorship attribution. Nonetheless, his innovative approach and dedication to capturing the essence of Costa Rican nature make his work an essential part of the Nationalist Generation’s legacy. Understanding Pacheco’s techniques and materials offers valuable insights into this pivotal period in Costa Rican art, including a clearer grasp of his technical mastery and artistic innovation, the cultural and historical context of his works, and his influence on subsequent generations of artists both regionally and internationally. Additionally, this understanding provides important information for the conservation and preservation of his paintings, ensuring that future generations can continue to appreciate his contributions.

The artworks of Fausto Pacheco are held in several important collections both in Costa Rica and abroad. In Costa Rica, his paintings are featured in the art collection of the Central Bank of Costa Rica (MBCCR), and the Museum of Costa Rican Art (MCA), among others. Internationally, some of his pieces have been acquired by collectors and galleries in countries such as the United States and Mexico, reflecting the recognition and appreciation of his art beyond Costa Rican borders.

Although over the years various specialists in artistic matters have expressed reservations about the authenticity of some of these artworks, their criteria for attribution has so far been unclear⁴³. This established the need to complement these technical criteria, with systematic evidence for more conclusive analysis. Our investigation had the opportunity to adopt a multidisciplinary approach to the study of these artworks, to find out more about the artist and the materials he used, and to establish parameters to determine the authenticity of the artworks. The subjects in study (see Fig. 1) are named in Spanish, following the conventions of the collections department of the MBCCR and MAC, who have identified them as: Paisaje con árbol (Landscape with tree) (57.0 cm (w), 65.0 cm (h)), Paisaje con casa campesina (Landscape with peasant house) (43.3 cm (w), 56.7 cm (h)), El puente (The bridge) (21.6 cm (w), 30.3 cm (h)), Paisaje A (Landscape A) (24.5 cm (w), 30 cm (h)) and Paisaje B (Landscape B) (24 cm (w), 30 cm (h)).

Attribution model for watercolours assigned to the Costa Rican painter Fausto Pacheco: a chemical and antimicrobial assessment — **Fig. 1: Multispectral imaging of *Paisaje con árbol* (top panel), *Paisaje con casa campesina* (second panel), *El puente* (third panel), *Paisaje A* (fourth panel) and *Paisaje B* (bottom panel).**

Results and discussion

Paisaje con árbol

Paisaje con árbol features a serene landscape dominated by a robust tree in the foreground on the left side of the composition, with an adobe house on the right, surrounded by vegetation (see Fig. 1). The artist used a warm colour palette with a slight contrast of cool colours in the mountains and more pronounced hues in the foreground, alongside a masterful choice of complementary harmony. The treatment of foliage and sky creates a peaceful and balanced atmosphere. The delicate brushwork and ‘spots’ of colour are indicative of the artist’s watercolour technique, which often aims to evoke a sense of calm and harmony in its representations.

During the technical examination of the artwork, a stamp was identified on the back with the name of a person (Arnoldo Castro Jenkins) who, according to art historians, bought paintings directly from the artist. As was common among collectors, he placed a stamp on the reverse to mark ownership. Based on this information and as well as on the artistic composition of the artwork, it was determined that this work is highly likely to have been painted by Fausto Pacheco, thus the other paintings will be compared to this one. According to the above, the point of comparison for the other artworks under study corresponds to Paisaje con árbol. The indicators to be compared for each one correspond to the composition of the work (depth plans, sketches), pigments and paper used.

By employing Multispectral imaging (MSI) which is a non-invasive technique, a preliminary diagnostic in the electromagnetic spectrum of the artwork is established in order to create high resolution images. With images from the visible (VIS) and IR region, it is possible to create an infrared false colour image (IRFC) in order to obtain a preliminary qualitative identification of a pigment. An additional benefit of this non-invasive approach is that it is also possible to observe previous restorations attempts. Figure 1a.1–a.4. show multispectral images of Paisaje con árbol. Four spectral regions are shown: VIS in Fig. 1a.1, ultraviolet (UV) in Fig. 1a.2, infrared (IR) in Fig. 1a.3 and infrared false colour (IRFC) in Fig. 1a.4. The VIS photograph (Fig. 1a.1) establishes a systematic baseline that allows for future comparisons and to track changes in the painting over time. Furthermore, in terms of composition we identified depth planes usually used by the artist: a dark foreground with mountains, an adobe house followed by other illuminated in which the main element was found, being in this case the tree. VIS also allowed us to know the pictorial palette that the artist used in the painting. In the region of the tree, shades of yellow and brown are perceived that provide volume information and a realistic effect to the painting. Near the house, vegetation also shows different shades of green and yellow; however, it is difficult to differentiate the possible amount of pigments used in this section and it is necessary to use other diagnostic tools. In this case, IRFC (Fig. 1a.4) showed us the regions of the painting that have the same pigment, in order to identify the areas to carry out spectroscopic sampling. For example, the yellow used for the tree is the same one as used in many regions of the grass.

After visual analysis of the painting and colour palettes with MSI, we hypothesised that the artist used blue, green, yellow and brown to create Paisaje con árbol. It is important to clarify that in watercolour technique, white pigment is not used to create light effects; instead, the paper itself is left exposed to provide brightness. Pigments are applied on the paper and blended to achieve the desired tones⁴⁴. Therefore, in all the proposed hypotheses, the colour white is not considered. Using RegionsOfInterest we were able to identify the composition of colours of the painting. This software analyzes image pixels in two steps. First, users draw a region of interest on the image. Then, the software isolates and processes pixels within a specific intensity range, using OpenCV to convert colour images to grayscale. It creates histograms showing the distribution of pixel intensities and generates a colour-highlighted image showing pixels within the selected range. The prevalent colours of the painting according to the hypothesis (made as a simple visual evaluation) were yellow and blue with 25% each. The least used colours in this case, corresponded to brown with 20% and green with 18%. These colour intensities constitute about 90% of the artwork. The remaining percentages are mixtures of pigments, something very common in watercolourists⁴⁴. This information helped us establish the areas of interest for analysis with spectroscopic techniques, and get an idea if it was a probable area of mixture or pure pigment.

Infrared (IR) photography (Fig. 1a.3) showed regions of the painting that contained graphite. To determine the amount of graphite we developed the software GraphiteComparison and selected common areas between the three paintings to be studied (sky, vegetation, signature and the main element of each one). These chosen areas correspond to those where it is highly probable that the artist has made a sketch. After processing the areas with the computational tool (in accordance with the methodology described later), it was possible to determine that the work is composed of approximately 36.4% graphite.

As described above, once the colours that the artist could have used were determined, two spectroscopic techniques were implemented to identify the pigments to which each colour corresponds. The selection of colours was made with the help of VIS photography where the most characteristic colours of each artwork and those most associated with the theme of art were identified. Likewise, the pigments determined with the spectroscopic techniques were compared with commercial ones from the brand Winsor & Newton®, which corresponds to the probable brand used by the artist^41,42.

Measurements made with Fiber optic reflectance spectroscopy (FORS) and X-ray fluorescence spectroscopy (XRF) are shown in Fig. 2. For each measurement, the Cultural Heritage Open Source Pigments Checker (CHSOS) was used to identify the pigments. FORS spectra (see Supplementary Figs. S3, S5 and S7) were inconclusive; therefore, a principal component analysis (PCA) model was created for each painting (see Supplementary Information Fig. S4), and three components (PCs) were considered from FORS spectra. All the XRF spectra (see Supplementary Fig. S10) presented lines of the same elements (S, Ca, Ba and Fe), which, as will be seen later, are related to chemicals used in the paper industry. XRF of blue areas as the sky (A1, A2) and details in the house and mountains (A3, A4, A5) reveals lines of Fe and Cu suggesting the presence of ultramarine (Na₈₋₁₀Al₆Si₆O₂₄S₂₋₄), Prussian blue (C₁₈Fe₇N₁₈) and phthalo blue (C₃₂H₁₆CuN₈)^29,32. XRF is not, however, the most suitable technique to identify and map ultramarine. Besides the elements described above, the XRF emission of the other elements characteristic of ultramarine (Na, Al, Si) are easily absorbed—by paint materials, the air, the detector window—and are hardly detectable with a scanning setup⁴⁵. Instead PCA results of sky and mountains samples showed a trend towards ultramarine and blue-painted areas in the house were close to phthalo blue and ultramarine.

**Fig. 2: Fiber optics reflectance spectroscopy (FORS) and X-ray fluorescence spectroscopy (XRF).**

For blue with a tint of green areas (A20–A24) like the lake and bushes, XRF spectra showed lines of Fe and Cu. Those lines suggest the presence of ultramarine, Prussian blue, phthalo blue, yellow ochre (Fe₂O₃ ⋅ n H₂O) and raw sienna (α-FeO(OH))^29,32,37. Because the apparent colour includes a green undertone, the option that the artist used blue and yellow to generate that undertone cannot be omitted. PCA showed closeness to Prussian blue, yellow ochre and raw sienna. The sampling area is close to a dark section (brown-reddish), for this reason the PCA shows a closeness to red ochre (Fe₂O₃).

Yellow pigments identified with XRF for this painting were different, spectra showed peaks of Fe, corresponding to a mixture of yellow ochre and raw sienna for lighter areas (A7, A8, A11, A12). For darker, yellow-painted areas (A9, A10) a mixture of yellow ochre and raw sienna was identified but with a tint of red ochre³². These pigments are located in the principal element of the painting, the tree. For lighter yellow, PCA was close to gamboge (C₃₈H₄₄O₈), an organic compound, so it does not present a characteristic line in XRF. For darker yellow, PCA showed closeness to gamboge and alizarin (C₁₄H₈O₄), also an organic compound.

XRF spectra of rocks and trunk of the tree (A13–A16) showed the presence of Fe, corresponding of a mixture of red ochre and raw sienna, which matches PCA results. In watercolours, the use of pigment mixtures is common, which is why in the case of colours such as green, a mixture of primary colours (blue and yellow) is proposed in order to obtain the desired colour. XRF spectra of green areas such as trees showed Fe and Cu peaks³². A mixture of yellow ochre, raw sienna, ultramarine, Prussian blue and phthalo blue is suggested. PCA showed closeness with phthalo blue and gamboge. Pigments identified with both techniques are shown in Table 1 for FORS and Table 2 for XRF.

Table 1 Suggested pigments for each painting identified with FORS

Full size table

Table 2 Suggested pigments for each painting identified with XRF

Full size table

In order to compare the other paintings analysed to Paisaje con árbol and to determine the likelihood of attribution to Fausto Pacheco, a numerical model based on cumulative authenticity probability (CAP) was designed. The model consisted of visual, technical and material indicators, that were weighted (expressed as a percentage) to reflect their relative importance in verifying authenticity according to similar studies and literature^17,46: type of pigments used (30% CAP), composition and brushstroke style (25% CAP), support materials (paper, canvas, etc.) (15% CAP), recurring motifs and themes (20% CAP) and aging and conservation analysis (10% CAP). These percentages provide a structured approach to probabilistically assess authorship, where each characteristic aligns closely with established qualities in Pacheco’s works. In the ‘Materials and methods’ section, Quantifying the probability of attribution, we fabricate a more concise description of these aspects.

In attributing Paisaje con árbol to Fausto Pacheco, we conducted a detailed analysis of its defining characteristics, as to stablish the baseline for comparison. The analysis of pigments is fundamental for understanding the material choices in Fausto Pacheco’s artworks. As mentioned before, the brand traditionally employed by Pacheco was Winsor & Newton®. Therefore, an investigation was conducted into the availability of pigments that this brand offered during the time the artist was active, as well as their accessibility in the region. This approach ensures that the colours found in the artwork are representative of the palette Pacheco used, thereby reinforcing the authenticity and historical value of the painting.

Regarding composition and brushstroke style, Pacheco is known for his loose and expressive technique, which infuses movement and fluidity into his works. His mastery of water in watercolour allowed him to mix colours in a way that generated smooth transitions and subtle nuances, characteristics observable in Paisaje con árbol. Additionally, the organisation of elements within the artwork reflects careful balance, often employing the rule of thirds. In this painting, the tree is presented as the central element, while a farmhouse is depicted in the background, adding context and depth, emphasising his focus on light and shadow.

Fausto Pacheco’s works, as shown in Paisaje con árbol, are deeply rooted in the celebration of Costa Rica’s natural beauty. The primary motifs and themes presented in his paintings reflect a profound appreciation for the country’s lush landscapes, capturing vibrant scenes that showcase its rich flora and fauna. Through his depictions of rural life, Pacheco pays homage to Costa Rican traditions and daily life, often highlighting agricultural and pastoral work. In Paisaje con árbol, the artist presents a traditional landscape scene that embodies the essence of rural Costa Rica during his time. The composition not only reflects the physical beauty of the environment but also evokes a sense of nostalgia and connection to the land and its cultural heritage.

The choice of support also plays a crucial role in the quality of the watercolour. To create works like Paisaje con árbol, Pacheco likely used special watercolour paper, which is thicker and more resistant to water than conventional art papers, allowing for multiple layers of paint without deformation. Although the specific type of paper he used is unknown, experts have suggested, based on visual analysis, that the artist employed high-quality papers with substantial weight. Regarding aging and conservation, it is important to note that over time, materials can undergo chemical changes that affect pigments and binders, resulting in discoloration and loss of vibrancy. In Paisaje con árbol, traditional signs of aging, e.g. colour degradation, humidity and pollution, have been observed.

Paisaje con casa campesina

In Paisaje con casa campesina, a rural scene is depicted with a typical house of the era, emphasising the simplicity and rustic charm of the Costa Rican countryside. The composition features earthy tones and vibrant colours, with the focal point on the house. The handling of light and shadow in this piece suggests Pacheco’s interest in capturing the warm natural light typical of the region, enhancing the painting’s authenticity and its connection to the place.

Similar to Paisaje con árbol, Fig. 1b.1–b.4 shows multispectral images of Paisaje con casa campesina. Four spectral regions are shown: VIS in Fig. 1b.1, UV in Fig. 1b.2, IR in Fig. 1b.3 and IRFC in Fig. 1b.4. In terms of the composition of the painting, in the VIS photograph (Fig. 1b.1), depth planes usually used by the artist were identified: a dark foreground with the vegetation followed by other illuminated in which the main element was found, being in this case the adobe house. In the case of Paisaje con árbol depth planes were also identified. With UV photography we identified dark spots as mentioned before mostly in the adobe house. With that information, four microorganisms were identified in our previous study⁴³.

Given the landscape theme, lots of greenery and blue skies were expected. As with Paisaje con árbol, with the computational tools mentioned above, it was possible to identify that the majority of colours present in the painting were blue with 25%, followed by green with 20%. On the other hand, the minor colours, yellow and brown were determined with 15% and 10% respectively. These colours make up approximately 70% of the artwork, which corresponds to approximately 20% less than Paisaje con árbol. Like Paisaje con árbol the remaining percentages correspond probably to mixtures of pigments. With IR photography (Fig. 1b.3) of Paisaje con casa campesina and accompanied by GraphiteComparisson we identified 28.9% of graphite present in the painting, almost 8% less than Paisaje con árbol.

XRF spectra of Paisaje con casa campesina are shown in Supplementary Fig. S10. The XRF inspection of blue-painted areas (see Fig. 2d) presented in the house (PPC1, PCC2), mountains (PCC3, PCC5) and sky (PCC4) reveals signals of Fe and Cu. Those lines are characteristic for ultramarine (Na₈₋₁₀Al₆Si₆O₂₄S₂₋₄), Prussian blue (C₁₈Fe₇N₁₈) and phthalo blue (C₃₂H₁₆CuN₈)^29,32. This is held by the PCA results obtained for different blue-painted areas. The dependence of the first, second and third principal components, PC2 vs. PC1 and PC2 vs. PC3, are shown in Supplementary Fig. S6.

The XRF spectra acquired for the yellow-painted details were separated in two groups, lighter (PCC6–PCC8) and darker (PCC9–PCC11) yellow. Darker yellow painted areas in trees are characterised by a strong peak of Fe, which corresponds to yellow ochre (Fe₂O₃ ⋅ n H₂O) and raw sienna (α-FeO(OH)). Both pigments yellow ochre and raw sienna are chemically similar and therefore there is no difference in XRF spectra. Lighter yellow sections in trees and bushes are characterised by peaks of Pb and Fe. Those lines are characteristic for raw sienna, yellow ochre and naples yellow (Pb₂Sb₂O₇)^29,32,37.

PCA for the yellow-painted areas, showed two different groups using XRF. Darker yellow areas suggested the presence of red ochre (Fe₂O₃), alizarin (C₁₄H₈O₄) and yellow ochre. It suggests that the yellow ochre of those parts was coloured with some red pigment chemically similar, therefore in XRF no differences could be observed. In the case of alizarin, it corresponds to an organic compound, so it does not present a characteristic line in XRF. Accordingly, in the dark yellow–painted sections, the presence of yellow ochre, red ochre and alizarin is suggested. Something similar occurs for the light areas, PCA suggests the presence of naples yellow, yellow ochre, gamboge (C₃₈H₄₄O₈) and red lead (Pb₃O₄). Those yellow-mixture-painted areas were tinted with some red pigment containing lead (Pb).

XRF spectra of green areas (PCC12–PCC14) such as grass showed Fe, Cu, Cr and Pb peaks. Due to the presence of Cr, the appearance of chrome oxide green (Cr₂O₇) and viridian (CrO₃ ⋅ 2 H₂O) is suggested (chemically both pigments are similar) and a mixture (related to the other lines of the spectrum) of yellow ochre, naples yellow, raw sienna, ultramarine, Prussian blue and phthalo blue³². The use of Prussian blue in these regions was confirmed by the Raman spectrum (see Fig. 3e). PCA showed closeness with Prussian blue, phthalo blue, ultramarine, yellow ochre, raw sienna. Because these zones are mixed with reddish rocks, those greener areas were probably tinted with some red pigment based on Pb. Pigments identified with both techniques are shown in Table 1 for FORS and Table 2 for XRF.

**Fig. 3: Composition of paper and pigments.**

Additionally, samples from this painting were analysed with other techniques in order to obtain more information about the paper and the pigments as shown in Fig. 3. Figure 3c shows that the sample (see Fig. 3b) is composed of two layers identified as follows: (1) top layer of paint and (2) paper. To know about the composition of the paper, a sample was analysed using FTIR–ATR.

The spectrum shown in Fig. 3d corresponds to: cellulose and calcium carbonate, CaCO₃, an inorganic filler^47,48,49. The presence of the cellulose hydroxyl group is confirmed by the stretching of the OH by approximately 3300 cm⁻¹. The peaks between 1100–1200 cm⁻¹ correspond to asymmetric ring respiration, and to the C-O-C glycosidic stretch vibration. The bands between 1100–1000 cm⁻¹ are due to C-OH stretching vibrations of the primary and secondary alcoholic groups of cellulose^47,50. The presence of lignin is confirmed by the vibrations of the aromatic ring between 1650–1600 cm⁻¹.

In the case of CaCO₃, a band is produced near 1400 cm⁻¹ and a small one at 880 cm⁻¹. These correspond to the C-O asymmetric stretching mode, C-O symmetric stretching mode of calcite, one of the CaCO₃ polymorphs. However, in the main peak of CaCO₃ (1400 cm⁻¹) they overlap with the flexural band CH₂ of cellulose^47,48. Every vibration that is active in the IR spectrum is also active in Raman spectroscopy⁵¹. In the FTIR spectrum, cellulose bands from the paper sample typically mirror those in the Raman spectrum, except for the 1650–1600 cm⁻¹ range where the absorption by water molecules in the IR spectra overlaps with the lignin peak at 1600 cm⁻¹, resulting in a strong absorption in the Raman spectrum^51,52.

To confirm the presence of fillers in watercolour paper, the sample mentioned above was analysed with SEM-EDX. For this a pigment-free zone was identified (see Fig. 3g) and its elemental composition was analysed, finding a large amount of calcium as shown in the spectrum of Fig. 3h, which was not surprising. Calcium in paper could be originated from various sources, including additives (fillers, coatings, and bleaches)⁵³. The high amount of sulfur present is due to the bleaching process in the paper industry where sulfur dioxide (SO₂) is used to remove excess chlorine to prevent reversion of colour in the paper⁴⁹.

A section of the sample with pigment was also analysed with SEM-EDX. In the spectrum shown in Fig. 3f some characteristic elements of the composition of the paper were identified, as well as lead, which confirms the presence of red lead (Pb₃O₄) and supports the other techniques. Likewise, a large amount of barium was identified, a peak that appeared in all XRF spectra, as described before. Barium sulfate (BaSO₄) is a classic filler mineral used in paint production as a binder. According to the reports of the state of conservation made in 2011 by ConArte^54,55, there is no presence of varnish or binders in this painting. However, based on observational evidence of BaSO₄, we hypothesise that Pacheco probably used this material to create a shiny effect in his artworks. Due to its extremely low oil absorption values, it is predestined for paint formulations. It acts as a texturizing agent without taking the risk of damaging its splendour⁵⁶.

Similar to Paisaje con árbol, the percentages for each indicator were determined as follows: in comparison to the pigments identified in Paisaje con árbol, four different pigments were identified in Paisaje con casa campesina, resulting in an assigned 27% CAP for this characteristic; however, we cannot rule out the possibility that the artist may have had access to different pigments or used some from other artists of the time, as it was common for artists to gather and paint together in the countryside. In terms of composition, the main plane prominently features the peasant house, yet the other planes of depth are less easily distinguished. As previously mentioned, the characteristic use of light and shadow by the artist is also evident in this work, which accounts for an assigned 23% CAP in this category. Regarding materials, they are similar, although not identical, to those used in Paisaje con árbol in terms of paper weight, thus receiving a 14% CAP attribution.

Thematically, Paisaje con casa campesina reflects a deep interest in rural life and Costa Rican traditions. This work showcases natural landscapes in which Pacheco emphasises elements of rural life, such as simple wooden houses with tiled roofs, set within natural surroundings that capture the essence of Costa Rican identity. Due to this alignment, a 20% attribution is given to this feature. Finally, this piece shows signs of restoration, as noted earlier, and similar to Paisaje con árbol, it exhibits typical aging marks. Therefore, aging and conservation are assigned a 9%. This results in a total authenticity score of 93%, a difference of nearly 7% compared to Paisaje con árbol.

Paisaje A and Paisaje B

Both works bear significant resemblance to others executed by Pacheco, with the central element being the adobe farmhouse. They depict a panoramic view of a Costa Rican landscape from the early twentieth century, characterised by bright colours and brushstrokes. Multispectral images of Paisaje A and Paisaje B are shown in Fig. 1d.1–e.4. Four spectral regions are shown: VIS in Fig. 1d.1 and e.1, UV in Fig. 1d.2 and e.2, IR in Fig. 1d.3 and e.3 and IRFC in Fig. 1d.4 and e.4, respectively. With UV photography we identified dark spots as in the other paintings, mostly in the sky for Paisaje A. With that information one microorganism was identified. In the case of Paisaje B, dark spots are not as visible as in the other paintings, with an exhaustive review of the photograph, some areas were identified and because of that a total of three microorganisms were identified (see Supplementary Fig. 16).

Such as Paisaje con casa campesina, with the computational tools mentioned above, it was possible to identify that the majority of colours present in Paisaje A and Paisaje B were blue with 15% and 22%, followed by green with 20% and 25% respectively. On the other hand, the minority colours, yellow and brown were determined with 5% and 12% for Paisaje A and 10% and 12% for Paisaje B. Like the other paintings the remaining percentages correspond probably to mixtures of pigments. XRF spectra of Paisaje A and Paisaje B are shown in Supplementary Figs. S14 and S15, respectively. The XRF inspection of blue-painted areas presented in the houses of each painting (PA6, PB5, PB6), mountains (PA2) and sky (PA1, PB1) reveals signals of Fe and Cu. Those lines are characteristic for ultramarine (Na₈₋₁₀Al₆Si₆O₂₄S₂₋₄), Prussian blue (C₁₈Fe₇N₁₈) and phthalo blue (C₃₂H₁₆CuN₈)^29,32. The XRF spectra acquired for the yellow-painted details (PB4) are characterised by peaks of Pb and Fe. Those lines are characteristic for raw sienna (α-FeO(OH)), yellow ochre (Fe₂O₃ ⋅ n H₂O) and naples yellow (Pb₂Sb₂O₇)^29,32,37. Both pigments yellow ochre and raw sienna are chemically similar and therefore there is no difference in XRF spectra.

For green areas (PA4, PA7, PB3, PB7, PB8) such as grass, XRF spectra showed Fe, Cu, Cr and Pb peaks. Due to the presence of Cr, the use of chrome oxide green (Cr₂O₇) and viridian (CrO₃ ⋅ 2 H₂O) is suggested (chemically both pigments are similar) and a mixture (related to the other lines of the spectrum) of yellow ochre, naples yellow, raw sienna, ultramarine, Prussian blue and phthalo blue³². Because these zones are mixed with reddish rocks, those greener areas were probably tinted with some red pigment based in Pb. Brownish-red areas of the trees, rocks and signature (PA3, PA8, PB9, PB10), suggested pigments such as red lead (Pb₃O₄) and red ochre (Fe₂O₃) according to lines of Fe and Pb in XRF spectra. The XRF examination of orange areas located in bushes, trees and house in those paintings (PA5, PB2). Those spectra reveal signals of Cr, Pb and Fe that corresponds to chrome yellow (PbCrO₄), probably mixed with a red that could be red ochre or red lead. Pigments identified with XRF are shown in Table 2.

As in the previous cases, the percentages for each indicator were determined as follows: in comparison to the pigments identified in Paisaje con árbol, four different pigments were identified in Paisaje A and Paisaje B, resulting in an assigned 25% CAP for this characteristic. In terms of composition, such as Paisaje con casa campesina, we identified the principal element of the painting in the main plain, however the other planes of depth are difficult to identify. The use of light and shadow is less than the other paintings, which counts for an assigned 20% CAP in this category. Regarding materials, they are quite different compared to Paisaje con árbol and Paisaje con casa campesina, in terms of weight. Paper used in Paisaje A and Paisaje B have less texture that could indicate a lower weight or a special finish that provides a more uniform surface, for this reason, a 10% CAP attribution is given to this aspect.

The themes of the works Paisaje A and Paisaje B focus on the representation of the natural environment and everyday life in Costa Rica, which are distinctive characteristics of Fausto Pacheco’s work. Both pieces highlight the beauty of the Costa Rican landscape, capturing elements such as mountains, lakes, and the lush vegetation that surrounds rural areas. Due to this alignment, a 20% CAP attribution is given to this feature. Finally, such as the other painting, they exhibit typical aging marks and no restoration report. Therefore, aging and conservation are assigned 10% CAP. This results in a total of 85% CAP, a difference of nearly 15% CAP compared to Paisaje con árbol.

El puente

This work depicts a bridge as the central subject, surrounded by a natural setting. There is a focus on perspective and meticulous treatment to represent the structure as accurately as possible. The execution reflects the artist’s interest in architectural elements within a natural context. The colours are vibrant, with an emphasis on yellowish greens and some warm colour accents. Multispectral images of El puente are shown in Fig. 1c.1–c.4. Four spectral regions are shown: VIS in Fig. 1c.1, UV in Fig. 1c.2, IR in Fig. 1c.3 and IRFC in Fig. 1c.4. With UV photography we identified dark spots as in Paisaje con casa campesina, mostly in the sky. With that information five microorganisms were identified in a previous study⁴³.

After analysis with MSI, we hypothesised that the artist used blue, green, yellow, red, orange and purple to make El puente. Using RegionsOfInterest we were able to identify the prevalent colour of the painting according to the hypothesis, in this case green corresponded to the most used colour with a 25% CAP, followed by blue and purple with 10% CAP each, while red, yellow and orange were the least with 5% CAP each. These colour intensities constitute about 60% CAP of the work, 10% less than Paisaje con casa campesina. The remaining percentages are mixtures of pigments like in the other cases. As the other cases with IR photography (Fig. 1c.3) of El puente and in the company of GraphiteComparisson we identified 24.3% CAP of graphite in the painting, almost 5% CAP less than Paisaje con casa campesina and 12% CAP less than Paisaje con árbol.

The XRF (see Supplementary Fig. S12) examination of the blue colour presented in the sky (P1) revealed signals of Fe and Cu such as Paisaje con casa campesina and Paisaje con árbol. Those lines are characteristic for ultramarine (Na₈₋₁₀Al₆Si₆O₂₄S₂₋₄), Prussian blue (C₁₈Fe₇N₁₈) and phthalo blue (C₃₂H₁₆CuN₈)^29,32. PCA (see Supplementary Fig. S8) showed proximity to phthalo blue. For blue with a tint of green areas like the lake and bushes (P2–P5), XRF spectra showed lines of Fe, Cu, Cr and Pb. Those lines suggest the presence of ultramarine, Prussian blue, phthalo blue, chrome oxide green (Cr₂O₇) and viridian (CrO₃ ⋅ 2 H₂O)^29,32,37. Because the apparent colour includes a green undertone, the option that the artist used blue and yellow to generate that undertone cannot be omitted. That is why the PCA shows closeness to Prussian blue, phthalo blue, yellow ochre (Fe₂O₃ ⋅ n H₂O) and raw sienna (α-FeO(OH)). The sampling area is close to a section of the bridge and reddish stones, very likely for this reason the PCA shows a closeness to red lead (Pb₃O₄).

Red areas of the bridge and rocks (P7), suggested pigments such as red lead and red ochre according to lines of Fe and Pb in XRF spectra. PCA suggested the artist used alizarin (C₁₄H₈O₄), red lead and since the area is close to trees, chrome oxide green. The XRF examination of yellow areas included a mixture with orange located in bushes and trees in the painting (P10–P12). Those spectra reveal signals of Cr, Pb and Fe that correspond to chrome yellow (PbCrO₄), probably mixed with a red that could be red ochre (Fe₂O₃) or red lead. Yellow ochre and raw sienna could not be possibly identified since this requires the presence of k_α and k_β of Fe³⁷. PCA results suggested the presence of naples yellow (Pb₂Sb₂O₇), gamboge (C₃₈H₄₄O₈), alizarin and red lead.

The XRF spectra acquired for the green-painted areas like grass and trees (P13, P14) showed signals of Fe and Cu, and other ones of Pb. Those lines are characteristic of all the blues mentioned above and yellow ochre and raw sienna^32,37. PCA results confirmed the presence of yellow ochre, raw sienna, phthalo blue and Prussian blue. It also showed a closeness to red ochre since some areas were adjacent to reddish zones of the painting. Pigments used in the principal element of the painting, the bridge, could not be identified with these techniques. Nevertheless, with these results, we were able to determine some of the possible pigments used by the artist. Pigments identified with both techniques are shown in Table 1 for FORS and Table 2 for XRF.

The percentages for each indicator were determined as follows: in terms of pigments, two were identified that differ from those found in Paisaje con árbol. As a result, this aspect has been assigned a weight of 27% CAP. Regarding the composition of the piece, the brushstrokes are not as delicate as those in the other works in the study, and there is a lack of the traditional colour transitions typically observed in his watercolours. The main element, the bridge, occupies the foreground, but the other planes of depth are almost non-existent. Consequently, this characteristic has been assigned a weight of 15% CAP.

Regarding the material, the paper used is quite different from that of the previous works; it has very little texture and, like Paisaje A and Paisaje B, has a low weight, making it highly flexible. This characteristic affects how the brushstrokes are perceived and how the pigments interact with the surface. Therefore, this aspect has been assigned a weight of 10% CAP.

In terms of theme, the work focuses on representing the connection between the landscape and the interaction between nature and human infrastructure, a hallmark of the artist’s style. For this reason, it has been assigned a weight of 20% CAP. Finally, the piece shows signs of having undergone restorations, as previously mentioned, and, similar to Paisaje con árbol, characteristic aging stains were found. Thus, a weight of 9% CAP has been assigned to the category of aging and conservation. This results in a total of 81% CAP, a difference of nearly 20% CAP compared to Paisaje con árbol.

Signature review

Signatures were analysed using visible images of the corresponding section for each painting: Paisaje con árbol labelled as A, Paisaje con casa campesina labelled as B, and El puente labelled as C. Distinctive features in the signatures can be seen letter by letter and globally (see Supplementary Fig. S1). The uppercase ‘F’ is very similar between signatures, signature C is slightly different because of the greater separation between horizontal and jagged lines, that is why a structural similarity index measure (SSIM) is greater when comparing A and B. The case of uppercase ‘P’ is very close to the last, here C has a sharper non-closed in the left looking semicircle, this feature makes the SSIM lower when compared to A or B. For readers interested in this ideal technology for image comparison, we recommend the following references^57,58.

In the case of the lowercase ‘a’ in C, the upper overhang does not leave space over the circle a feature not shared by the rest, nevertheless, the SSIM detects more similarity between B and C, this may be due to greater intersection between the two owing to the line width, still, qualitatively A and B are closer. Similarly, the first lowercase ‘c’ in B is the only one with an empty space and the SSIM showed more similarity between B and C, the reasons are the same as before as well as a more closed script in A. The lower overhang in lowercase ‘h’ for A, B, and C is an embellishment unique to the author’s signature, nonetheless, is significantly longer in C, which may be the reason why the SSIM ranks better the comparison between A and B. The pressure spot at the top of the letter h for B is unique but does not affect the overall similarity with A.

For the lowercase ‘e’, C script is not closed, a feature that may explain the lower SSIM, again a pressure spot is in the letter ‘c’ for B. For the second lowercase ‘c’, C presents a hooklike structure that may explain again the lower SSIM for C. Lowercase ‘o’ yields lower SSIM when compared to A, since this script does not have any special feature, the lower similarity can be due to the thicker line. For uppercase ‘H’, two embellishment overhangs for A and B, while this feature is not shared by C, the larger overhang in A seems to yield lower SSIM, nevertheless qualitatively A and B seem to be closer. In general, most of the letters of A and B are similar, the baseline is close between these two. All the last commented factors substantiate the global SSIM of 0.652 between A and B, and lower SSIM of 0.647 for B vs. C and 0.640 for C vs A. Reported values of SSIM values of real/real and real/forgery pen–drawn signatures can intersect and on average have a difference of 3.8%, higher than 1.8%, the maximum calculated in our case for the whole signature.

While our SSIM values are not conclusive to point directly to a forgery, it is important to consider that the analysed signatures were applied with watercolour and therefore, they fundamentally constitute very different objects from signatures written with a pen on paper since strong pressure should not be applied to make the signature. Additionally, parameters such as thickness can vary between artworks and the transparency does not allow us to analyse pressure spots in detail. A forgery cannot then be concluded only by SSIM, especially with three samples, yet they provide a relative comparison consistent with qualitative results to point out the outlier C.

Analysis of biological factors

A microbiological sampling of the watercolours was carried out to isolate possible microorganisms responsible for the paintings biodeterioration⁴³. The sites of interest for the sampling were decolouration patches or dark stains, observed either by VIS or UV light. From the painting Paisaje con casa campesina, one bacterium (Dermacoccus nishinomiyaensis) and two non-sporulated septate filamentous fungi were isolated from previous interventions done over parts of the white and blue walls of the adobe house and the fungi Diaporthe sp. was isolated from a zone of decolouration in the sky. In the case of El Puente, the bacterium, Bacillus cereus, was isolated from a dark spot, observed under UV light, in a cloud. As for the fungal isolates, two different strains of Aspergillus section Versicolores were retrieved from the artist’s signature, Cladosporium tenuissimum came from a dark stain in the back side of the painting (behind the dark spot where B. cereus was isolated) and a dematiaceous non-sporulated fungi from some writing in the back of the painting (see Supplementary Table S3). These two paintings are often used for exhibitions or to decorate the offices of the bank’s personnel; hence are exposed to different environmental conditions. Unlike these two watercolours, Paisaje A and B, are kept in a vault that has a low fungal aerial concentration of 96 spores/m³. Curvularia asiatica was the only microorganism isolated from Paisaje A. It came from a dark spot (visible under UV light) in the adobe house’s white part of the wall. Finally, from Paisaje B, the mould Schizophyllum commune was also isolated from a dark stain in the white part of the adobe house’s wall, the yeasts Candida tropicalis and Clavispora lusitaniae (Candida lusitaniae) were obtained from a dark spot in the sky and another isolate of C. lusitaniae was obtained from an area of decolouration, also in the sky (see Supplementary Fig. S16).

To evaluate the effect of the pigments used by the artist on the spread of microorganisms, two strains of B. cereus (M1P isolated from the painting El Puente and M211 isolated from food), two Aspergillus section Versicolores (MP5.1 and MP5.2) and C. tenuissimum (A1P) were selected for the in vitro experiments. The results of the inhibitory effect of watercolours on B. cereus are shown in Fig. 4. It was observed that the inhibitory halo was larger in the presence of yellow ochre when compared with ultramarine blue, for both strains and all the dilutions. However, that inhibitory effect was significant (p < 0.05) just in the case of larger dilutions (1/8, 1/16 and 1/32); it is possible that some of the lack of inhibition in the case of the original dilution (1/4), is related with a reduced mobility of the pigment on the agar surface⁵⁹. When comparing both strains, it is observed that both isolates behaved essentially the same in the presence of the two pigments. Interestingly, the M1P strain was more susceptible to both pigments in the same dilution (1/8). Regarding the comparison of both watercolours, the yellow ochre was more inhibitory (p < 0.05) than ultramarine blue, for most of the dilutions and both strains. These results show no evidence of a greater resistance to the pigments in the case of the isolate obtained from the painting, meaning that effective bacterial colonisation is not related to this trait. Also, given that the inhibitory effect is intimately related with the mobility of the pigment, it is very unlikely that watercolours on the painting may represent an important barrier for microbial colonisation. In fact, efficient antibacterial surfaces must function to control both attachment and biofilm formation of bacteria⁶⁰ and watercolours may not provide this dual-functionality.

**Fig. 4: Effect of two watercolour pigments on *Bacillus cereus* growth on Tryptic Soy Agar (TSA).**

Yellow ochre pigments are mostly extracted from organic components, such as soil, wood, and volcanic rocks⁶¹. In fact, yellow receives its characteristic ochre pigmentation from iron oxides (Fe₂O₃) as described above⁶². Larger antimicrobial effect of yellow ochre may be associated with microbial toxicity due to a series of interactions with iron oxides, including membrane depolarisation, production of reactive oxygen species (ROS) with lipid peroxidation and DNA damage⁶³. Iron oxide has been used as a main component of antimicrobial nanoparticles with biomedical applications. The effect of iron oxide nanoparticles against B. cereus has been reported before⁶⁴.

In addition, the minimal inhibitory concentration (MIC) to the watercolour ultramarine blue and yellow ochre pigments was also determined for three fungi as shown in Fig. 5, previously isolated from altered areas of the watercolour El Puente, Paisaje A and Paisaje B. Five of the isolates were retrieved from the front side of the painting (A. section Versicolores, isolates MP5.1 and MP5.2, C. asiatica, C. tropicalis and C. lusitaniae), and one from the back side (C. tenuissimum, isolate A1P)⁴³. The MICs for ultramarine blue and yellow ochre are presented in Table 3.

**Fig. 5: Effect of two watercolours on fungal growth.**

Table 3 Minimal inhibitory concentration to ultramarine blue and yellow ochre pigments of six fungi isolated from the watercolours El puente, Paisaje 1 and Paisaje 2 from the Costa Rican artist Fausto Pacheco

Full size table

Similar to with bacteria, the pigment that resulted most permissible for fungal growth was ultramarine, which gets its bright blue colour mainly by sulfur compounds⁶⁵. Ultramarine pigment was originally made from extracts of the rock Lapis lazuli, which contains high quantities of the mineral Lazurite ((Na,Ca)₈(AlSiO₄)₆(SO₄,S,Cl)₂)⁶⁵. However, given newest mining regulations and ecological impact concerns, nowadays the pigment is mainly produced synthetically⁶². This finding is in coherence with the isolation areas of most of the microorganisms, since they came from either the blue sky or the white and blue walls of the adobe houses.

On the other hand, yellow ochre has antifungal properties because it inhibits spore germination⁶⁶. A previous study conducted in Évora University (Évora, Portugal) analysed, via in vitro simulation assays, the biodeterioration capability of Aspergillus niger against ultramarine, red ochre and yellow ochre pigments. The fungal isolate was able to proliferate at high concentrations of these pigments, remaining metabolically active even after 30 days in contact with them, results which are in concordance with our data. Also, they determined that A. niger was able to induce high discoloration of the pigments, evidencing its biodeteriorative ability⁶¹.

Microorganisms can have a detrimental impact on paintings by degrading pigments, leading to discoloration and fading, particularly in organic pigments. For example, the fungal isolates Diaporthe and C. lusitaniae were isolated from such areas. They can also weaken the paper or canvas, causing structural damage like thinning or tearing. Additionally, microorganism growth and metabolism can produce stains, like the ones observed where B. cereus, S. commune, C. tropicalis and C. lusitaniae were obtained. Surface damage, including pits and crumbling, can affect the visual quality of the artwork. The presence of microorganisms often indicates poor environmental conditions, such as high humidity, which can accelerate deterioration. Addressing these issues involves controlling environmental factors, carefully cleaning affected areas, and applying conservation treatments to stabilise and protect the artwork^17,67.

Conclusions

This research aimed to identify the materials and pictorial palette used by the Costa Rican artist Fausto Pacheco Hernández through a multi-analytical approach. By analysing a known work by the artist, i.e. Paisaje con casa árbol, we developed a model that consisted of visual, technical and material indicators: pigments used (30% CAP), composition and brushstroke style (25% CAP), support materials (15% CAP), recurring motifs and themes (20% CAP), and aging and conservation analysis (10% CAP), which can help identify paintings likely created by Pacheco or others in the future. However, we are aware that these findings have to be carefully interpreted by a group of experts of different areas (curation and museology, library and archival science, preservation science, archaeology, cultural artifacts material scientists), in order to give context to the results; since this study marks the beginning of a comprehensive artists’ material database and his historical context, enhancing our understanding of his creative techniques and palette. When comparing the paintings to Paisaje con árbol, it can be concluded that Paisaje con casa campesina was the most similar with a cumulative attribute percentage of 93%, followed by Paisaje A and Paisaje B with 85%, respectively. The least similar was El puente with an 81%. The methodologies and findings from this study offer valuable insights that extend beyond the specific artworks examined. The multi-analytical approaches used here can be applied to other artworks, including different types of watercolours, to gain similar insights into material composition, degradation patterns, and conservation needs. This research provides a foundational framework for analysing artworks in various contexts, potentially informing conservation practices and scholarly research across different media.

Our findings also revealed that some of the watercolours used in Pacheco’s paintings possess antibacterial and antifungal properties. However, the presence of microbial contaminants on the surface of these artworks indicates that the pigment concentrations used are insufficient to prevent microorganism colonisation and subsequent deterioration. This highlights the need for preventive measures to preserve such artworks, since the pigment concentrations in actual paintings are probably too small. Therefore, further studies are essential to explore the effects of microbial growth and metabolism on pigment discoloration, and to refine preventive measures for the preservation of watercolour paintings. By expanding the application of these analytical techniques, we can better understand and protect a broader range of artistic works, contributing to the field of art conservation and the preservation of cultural heritage.

Materials and methods

Multispectral imaging

Photographs in different regions of the electromagnetic spectrum were produced: visible (VIS), ultraviolet (UV), infrared (IR), and infrared false colour (IRFC), which was computationally generated using VIS and IR images. VIS photograph shows possible superficial damage and allows us to visualise the insinuate the colours used by the artist, which due to their chemical properties may have changed over time or restoration⁶⁸. IR imaging reveals information under the surface of the painting, in order to visualise preparatory sketches or corrections made by the artist^69,70. The UV image shows materials that are sensitive to UV light and therefore fluoresce, such as some pigments, binders and glues^70,71.

To create each of the photographs, 15 individual images of each of the artworks were taken for each spectral region: VIS, IR, UV. All photographs were acquired using a Nikon D5300 camera with an AF-S DX Micro-NIKKOR 40 mm f/2.8G lens and was adjusted in order to obtain high resolution photos (24 megapixels)^{69,70,71,72,73}. The experimental conditions of the camera varied for each region, for VIS images: f9, ISO 200 and exposure time of (frac{1}{8}) s; for UV photographs: f10, ISO 200 and exposure time of 2.5 s; and for IR imaging: f9, ISO 200 and exposure time of 25 s^{69,70,71,72,73}.

Two 150 W halogen lamps were used for VIS and IR photography (situated 30.0 cm from the paintings). For the ultraviolet photography the lamp (situated 25.0 cm from the paintings) was made up of five Tecno Lite® bulbs with an average power of 20 W each. The spectral information of the UV radiation of this lamp is produced at a wavelength of 350 nm. For the IR photographs we used a Fotga 2.047 adjustable IR filter that goes from 530 nm to 750 nm. Nevertheless, it is important to mention that the filter was used at 750 nm wavelength to increase the signal-to-noise ratio of the main peaks. A white balance was carried out for all the photographs⁷⁴ and a pigment reference named Cultural Heritage Science Open Source Pigments Checker (CHSOS) pigment reference was used. Computational tools such as Lightroom Classic® and Photoshop® were employed for lens adjustments and to create IRFC pictures, by aligning VIS and IR photographs in order to exchange the blue channel for the green channel in VIS, the red channel for the green channel in VIS and the IR red channel replaced the red channel in VIS^68,69. Panoramic images of each spectral region were generated with Lightroom Classic®.

RegionOfInterest

RegionOfInterest (available for free in the GitHub repository https://github.com/FaustoPacheco/RegionsOfInterest.git) was used to identify the composition of the artwork in terms of colours used by the artist^19,75,76.

RegionOfInterest operates in two main stages to quantify pixel intensity values. First, the user defines a region of interest on the image, which can be of any shape or size. In the second stage, the software isolates all pixels within a specified intensity range based on the initial user-defined region. This process uses OpenCV to convert colour images to grayscale, classifying pixels on a scale from 0 (black or no intensity) to 255 (white or maximum intensity). The software then creates histograms that plot ‘Count vs. Intensity’ to visualise the distribution of intensity values. Additionally, it produces a colour image highlighting pixels that fall within the chosen intensity range, showing their locations on the original image. Intensity measurements are performed using OpenCV library functions and have been compared to BT.601 and BT.709 colour-to-grey transformation methods, with uncertainties of ±0.2% and ±2.1%, respectively. The processing time for the software is approximately 1 second, though it can vary depending on image format, size, and computer performance. The quality of the analysis also depends on the image quality and lighting conditions. In this way, based on the pixel intensity values, it is possible to map within the artwork the locations of the different pigments and therefore, select regions of interest for spectroscopic in situ sampling¹⁹.

Developing GraphiteComparison

GraphiteComparison makes use of IR photographs of each of the artworks. The software, written in C, processes the IR photos and converts them into photos with black and white lines. These images show that everything black is within the intensity range with which the artist traced the sketches and everything white shows parts of the paint that are outside that intensity range. It is certain that for each of these images, all of the sketch lines are in black; however, all the black is not exclusively sketch lines, since the intensity range includes other pixels with the same intensity as the graphite, i.e. infrared-absorbing pigments and other materials. It seems relevant to us to mention that GraphiteComparison only computationally considers the result of the IR photographs, therefore, it is likely that there are traces of graphite and other materials used by the artist that absorb IR.

For this software, the graphite intensity means the level of light that a graphite pixel has. It consists of a range defined by two thresholds where it is more likely to find graphite in the paintings. Being the minimum intensity threshold U_min = 120 and the maximum intensity threshold U_max = 145 for the graphite intensity. A graphite painted pixel means a pixel that is also a sketch pixel, since the sketches were made in graphite. GraphiteComparison (available for free in the GitHub repository https://github.com/FaustoPacheco/GraphiteComparison.git) was used to obtain a very precise approximate percentage of the number of pixels with the same intensity of the graphite and, therefore, also the pixels with the intensity of graphite presented in the paintings.

Fiber optics reflectance spectroscopy (FORS)

Areas of interest for sampling were determined using MSI and computational softwares described above. In the case of Paisaje con casa campesina, 22 areas of interest were identified, for El puente 25 and for Paisaje con árbol 26 (see Supplementary Fig. S2). Spectrums of CHSOS pigment references were carried out to compare the ones of the paintings and determine pigments used by the artist. Spectrums were obtained in situ, using the ASD FieldSpec® instrument with a spectral range of 350–2500 nm, the sampling rate was 0.2 s per spectrum. Spectrums were collected and saved with the RS³ software, an optimisation and setting a white reference was needed. The spectra were averaged using ViewSpec^TMPro software. The collection of spectra was processed statistically using the principal component analysis (PCA) procedure of ChemFlow.

The variable analysed in FORS is the reflectance intensity across different wavelengths. PCA processes these intensities to identify principal components that explain the most variance in the spectral data. The first principal component (PC1) corresponds to the violet to blue wavelengths, peaking at 438 nm, with major contributions from 530 nm to 750 nm, reaching a maximum at 620 nm. The second principal component (PC2) better differentiates violet (403 nm) and yellow-orange (570 to 620 nm) wavelengths, along with contributions from 765 nm to 1150 nm, peaking at 922 nm. The third principal component (PC3) covers a range from violet (387 nm to 437 nm) to green (with a peak at 531 nm) and from orange to red (615 nm to 725 nm), showing peaks at 800 nm and 1000 nm, with additional peaks at 893 nm and in the 1340 to 1820 nm region.

A baseline correction to adjust for signals from underlying materials was applied, ensuring that the reflectance data more accurately represent the surface features of the artworks. We used CHSOS pigment reference spectra from known materials to understand and account for any interference from underlying layers. By comparing PCA results from samples with known underlying materials versus those without, we assess the impact of these materials on the analysis. Variations in PCA component loadings and scores help determine their influence^77,78.

The analysis was conducted over the entire spectral range collected, allowing for a comprehensive understanding of the spectral features. Data preprocessing involved normalisation to account for variations in intensity and baseline correction to eliminate systematic variations not related to pigment characteristics, as mentioned above. The number of principal components retained was determined based on cumulative explained variance, with a typical threshold of 90%. Cross-validation techniques were used to assess the stability of our PCA results, ensuring robustness in our selected parameters.

X-ray fluorescence spectroscopy (XRF)

XRF was used to determine the elemental composition of the pigments. Areas of interest for sampling were determined in the same way as FORS. In the case of Paisaje con casa campesina, 22 areas of interest were identified, for El puente 15, for Paisaje con árbol 25, for Paisaje A 10 and for Paisaje B 13 (see Supplementary Fig. S9). Spectrums of CHSOS pigment reference were carried out to compare the ones of the paintings and determine pigments used by the artist. The irradiation was carried out in situ, with an X–ray tube, AMPTEK Mini X, with silver white (E = 22.1 keV), which is operated at 35 kV and 15 μA, with a beam of X-rays that impacted the sample at an angle of 67.5° with respect to its front surface. The characteristic rays, emitted from the sample, were received by a pure silicon (Si) detector, AMPTEK XR–100, located at an angle of 45° to the primary X-ray beam. The detector was linked to a digital pulse processor, AMPTEK PX4, which incorporated the multichannel analyser and power supply. X-ray spectra are obtained using the AMPTEK XRF-FP software, which displays the data on a computer. Before starting measurements a calibration was carried out with Molybdenum (Mo) and Titanium (Ti) plates. For every measurement the exposition time was 120 s. The analysis of the spectra obtained was carried out with the PyMca software.

Signature review

Signatures were processed using visible images of the corresponding section for each painting: Paisaje con árbol labelled as A, Paisaje con casa campesina labelled as B, and El puente labelled as C (see Supplementary Fig. S1). The latter were cropped and resized using GNU Image Manipulation Program (GIMP) maintaining the aspect ratio such that most of the strokes would superimpose. The obtained images were used in two instances. First, a feature review is done by placing a grid and enclosing graphical features where blue boxes indicate embellishment features, red boxes blurred lines, and magenta lines baseline alignment, the grid labels each rectangle as (x,y) according to the corresponding discrete values of the horizontal and vertical axes respectively. Secondly, images were reprocessed using GIMP, signatures were cropped and transformed to binary to highlight their morphological features. Also, each individual letter was cropped to compare against each other, and quantitative comparisons between paintings were made using the SSIM algorithm of scikit-image in Python.

Sampling along with microscopy and spectroscopy analysis

Two-millimetric samples (approximately 4.0 mm²) were collected, with the approval of the Manager of the museum’s Collections Department. These samples were collected from areas previously identified as damaged and examined with Optical Microscope, SEM-EDX, Raman spectroscopy and FTIR-ATR.

Scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDX)

The aforementioned samples were analysed in order to know their microstructure and elemental composition. They were first observed using an AMPScope stereoscope for macro photography. For microanalysis, the sample was placed on a metal base and adhered to it with double-sided carbon tape and it was not covered. Once ready, it was observed in a Hitachi SEM S-3700 scanning electron microscope at low vacuum, 15 KV and 10.5–11 mm WD, with backscattered electron images to analyse its structure. The SEM has an EDX detector from the company IXRF Systems with which elemental analyzes were taken.

Micro-Raman spectroscopy

One sample was analysed in order to get information about pigments with micro-Raman spectroscopy. We used a WiTec alpha 300R micro-Raman Spectrometer and a diode laser emitting at 532 nm, with an objective 50x. All the spectra were measured in the spectral range 500–4500 cm⁻¹. Acquisition time was optimised for each area of the sample. All spectra were compared with reference pigments from CHSOS.

Fourier transform infrared-attenuated total reflectance spectroscopy (FTIR-ATR)

Two samples were analysed using a Nicolet 6700 Thermo Scientific spectrophotometer with a diamond ATR crystal. The spectra were collected in the range 4000–500 cm⁻¹. FTIR-ATR was used to identify the chemical composition of the base material of watercolours: paper, and determine if the artist made any preparation to it.

Quantifying the probability of attribution

In our study, we adopted a multi-faceted analytical approach to evaluate the probability of attribution for each artwork to Fausto Pacheco. Our methodology is grounded in both qualitative assessments and quantitative measurements across several key criteria. This approach allows for an objective and technical evaluation of the works, addressing limitations posed by the lack of precise dating and the irregular quality of some of Pacheco’s pieces. The analysis of pigments and support materials aids in associating the works with a specific period and the resources available during Pacheco’s time. Likewise, the identification of stylistic patterns, such as brushstrokes and recurring themes, provides a consistent visual framework that strengthens the attribution of authorship. Additionally, aging studies help distinguish between original works and potential forgeries. Together, these indicators offer a multidimensional approach that supports the authentication process, providing a solid foundation for more precise and reliable attribution of the works.

The percentages assigned to each aspect of the artwork were defined through a collaborative process involving art conservation professionals, stylistic experts, and material scientists. These experts helped determine the relative importance of features like pigments, brushstroke style, support materials, recurring motifs, and aging in relation to Pacheco’s known techniques and historical context:

Type of pigments used (30%): identifying the specific pigments employed in Pacheco’s watercolours can be a strong indicator of authenticity, as certain pigments would have been available during his time or would be characteristic of his palette. We conducted a thorough examination of the pigments present in each artwork using FORS and XRF analysis. Each pigment identified is assigned a percentage based on its presence in the artwork and its known usage by the artist.

Composition and brushstroke style (25%): the technique of applying paint, such as the direction and type of brushstrokes, can offer valuable clues about authorship. Pacheco had a distinctive style that can be recognised through detailed analysis of his artistic execution. Brushstroke techniques used in each artwork were evaluated against a catalogue of brushstroke known by art experts, with higher scores assigned to those exhibiting techniques closely associated with Pacheco’s known methods.

Support materials (paper, canvas, etc.) (15%): the materials used as support, such as the type of paper or cardboard, can be linked to a specific period or to Pacheco’s preferred materials. The physical properties of the paper and other materials used in each artwork were analysed, focusing on texture, weight (grammage), and surface finish. A comparative analysis with a known work by Pacheco was conducted to assess material consistency.

Recurring motifs and themes (20%): frequent themes in Pacheco’s works, such as rural landscapes and Costa Rican scenes, can serve as stylistic indicators of authenticity. Thematic analysis involved reviewing the subject matter and themes present in each artwork, scored based on their relevance and representation in the context of Pacheco’s oeuvre.

Aging and conservation analysis (10%): the state of preservation and aging analysis of the works can provide clues about their authenticity. We assessed signs of aging and any restoration efforts undertaken on the artworks. Each artwork receives a percentage based on the extent of aging and restoration, with higher percentages assigned to those that show minimal intervention consistent with Pacheco’s style.

Once each characteristic has been evaluated and assigned a percentage, these values are integrated to calculate a cumulative authenticity probability for each artwork. The qualitative assessments were conducted by art experts who possess extensive knowledge and have studied Pacheco’s works in depth. To guide their evaluations, these experts were provided with a structured evaluative table (see Supplementary Table S2) that outlines specific criteria and corresponding percentage ranges for assessing the alignment of the artworks with Pacheco’s known characteristics and practices. The cumulative probability reflects the overall alignment of the piece with Pacheco’s techniques and style. The final score is expressed as a percentage, where higher values indicate a stronger likelihood of attribution to Fausto Pacheco.

Analysis of biological factors

The effect of pigments on isolated microorganisms was established following a modified version of the agar diffusion assay as described by Gabrahie et al.⁷⁹. In the case of bacteria, two B. cereus strains were used; one of the isolates (M1P) was obtained from El puente⁴³ and the other strain was an isolate from food (M211) that was used for comparison purposes. Both strains were maintained as glycerol stocks at −20 °C as part of the bacterial collection of the Research Center for Tropical Diseases (CIET-UCR). Before the experiments, each strain was grown on Blood Agar plates that were incubated at 35 °C for 24 h. One or two colonies were used to prepare a bacterial suspension in 0.1% Sterile Peptone Water equal to the 0.5 MacFarland Standard. Each suspension was used to prepare uniform bacterial lawns on Mueller–Hinton plates that were allowed to dry for 30 min before cutting the agar with a sample disc collector to make five holes per plate.

Two watercolours were analysed: Ultramarine blue and Yellow Ochre (Winsor & Newton, London, England). These colours were selected as they are similar to the ones identified with the chemical analysis of the pigments of the three paintings analysed. Each of them was half diluted using sterile water to obtain different concentrations (1:4, 1:8, 1:16, 1:32) and a volume of 25 μL of each dilution was dispensed into one of the holes on the agar plate; the hole in the middle of the agar plate was filled with distilled sterile water as a control. The agar plates were incubated at 35 °C for 24 h and the inhibition halo surrounding each hole was measured (mm).

In the case of fungi, three isolates also obtained from El puente⁴³, one from Paisaje A and three from Paisaje B were included in this study. On June 22nd 2023, the latter two paintings were sampled for microorganisms. Twenty-two sampling sites were selected for each painting after analysing their multispectral images. The sites of interest were swabbed and the swabs were then transferred to the Section of Medical Mycology of the School of Microbiology of the University of Costa Rica in peptone water at 4 °C. At the laboratory, the Blood agar and Agar Sabouraud Dextrose plates were inoculated and incubated for 24 h at 37 °C and 3 weeks at room temperature (RT) (ca. [20–25] °C), respectively. Yeasts were identified using the VITEK® 2 System while filamentous fungi employing the MBT (Maldi Biotyper, Bruker Daltonics, Bremen, Germany). Finally, the spectra were analysed using the Bruker Library and the MSI (Mass Spectra Identification) platform (msi.happy-dev.fr) developed by the Assistance Publique-Hôpitaux de Paris, the Sorbonne University of Paris, and the Belgian Coordinated Collections of Microorganisms (BCCM).

The isolates were preserved at the Fungal Collection of the School of Microbiology of the University of Costa Rica. The MIC of the same watercolours (Ultramarine blue and Yellow Ochre) on the fungus was determined by using the M38–A guidelines from the Clinical Laboratory and Standards Institute (CLSI)⁸⁰. The mycelial isolates were grown for 7 days at RT in test tubes containing Potato Dextrose Agar. A suspension of conidia for each isolate was standardised to (1–5) × 10⁶ conidia/mL in 0.85% sterile saline. Then, 200 μL of these suspensions were added to test tubes with 9.8 mL of RPMI 1640 (Roswell Park Memorial Institute) (Thermo Fisher Scientific, USA). The yeasts were grown in Agar Sabouraud dextrose for 24 h at RT. A suspension of the fungi was standardised to 0.5 McFarland ([1–5] × 10⁶ blastospores/mL) and then 9.8 μL was added to 9.8 mL of RPMI 1640. The watercolours were diluted in the following concentrations with distilled sterile water: 1:2, 1:4, 1:8, 1:16, 1:32, 1:64, 1:128 and 1: 512. Then, 100 μL of each dilution was added to a microdilution plate of 96 wells. One hundred microliters of the fungal suspension were also added to the wells; hence the final concentrations of the paintings went from 1:4 to 1:1024. The microdilution plates were incubated for 72 h at RT. After the incubation time, the content of each well was plated and distributed on Agar Sabouraud Dextrose plates and incubated for 72 h at RT. The MIC was determined as the minimal concentration that inhibited fungal growth.

Finally, an aerial volumetric air sampling was performed of the vault were the paintings Paisaje A and Paisaje B were kept at the MAC it was conducted using a Burkard Personal Volumetric Air Sampler (Burkard Manufacturing Co. Ltd., United Kingdom).

Introduction

Historical context

Results and discussion

Paisaje con árbol

Paisaje con casa campesina

Paisaje A and Paisaje B

El puente

Signature review

Analysis of biological factors

Conclusions

Materials and methods

Multispectral imaging

RegionOfInterest

Developing GraphiteComparison

Fiber optics reflectance spectroscopy (FORS)

X-ray fluorescence spectroscopy (XRF)

Signature review

Sampling along with microscopy and spectroscopy analysis

Scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDX)

Micro-Raman spectroscopy

Fourier transform infrared-attenuated total reflectance spectroscopy (FTIR-ATR)

Quantifying the probability of attribution

Analysis of biological factors

Related Articles

Responses