The endorheic Basin of Mexico (BM), also called the Valley of Mexico Basin (19° 29′52″N, 99° 7′37″W), is located in the central part of the Trans-Mexican Volcanic Belt Province belonging to the Mexican Transition Zone (Morrone et al., 2017; Morrone et al., 2022), with a mean altitude of 2,276 m above sea level. According to Arce et al. (2019) the area of the basin is of 9,620 km². The basin is delimited by the volcanic ranges of Sierra de las Cruces to the west, Sierra Nevada to the east, Sierra Chichinautzin to the south, and the Apan-Tezontepec volcanic range to the north (Arce et al., 2019; Martínez-Abarca et al., 2021) (Figure 1).

The predominant soil types in the area of the former Great Lake of Mexico are Phaeozem, Vertisol, and Solonchac; the latter is characterized by a high content of soluble salts (Sedeño-Díaz et al., 2019; CONABIO, 2023) (Supplementary Material; Figure 1A).

Today, the predominant climate in the Basin of Mexico (BM hereafter) is temperate [C (w0), C (w1), and C (w2)] according to Köppen’s classification (García, 2004). Temperature ranges from 13°C to 25°C, with a mean annual temperature of 16°C; the mean annual precipitation varies between 700 mm and 900 mm, with the rainy season in the summer and up to 5% of rainfall in the winter (Alcocer and Williams, 1996; López-López et al., 2016) (Supplementary Material, Figure 1B).

During the Upper Cretaceous and Lower Tertiary, the BM was an open basin, with a surface drainage to the south through two main rivers, which produced large deposits of alluvial material at the bottom of the basin. Subsequently, during the Pleistocene, the basin closed off (Palma et al., 2022). In this sense, the lithological layers in the BM are dominated by alluvial deposits (with sedimentary rocks) and andesite-basalt (Supplementary Material, Figure 2).

The former Great Lake of Mexico was located within the BM (Figure 1) and was split into five lakes during the dry season, namely,: Texcoco, Zumpango, Xaltocan, Chalco, and Xochimilco (Berres, 2000; Montero-Rosado et al., 2022). Nowadays, only four lakes remain: Chalco, Texcoco (Nabor Carrillo reservoir, built in 1983, covering a small area located where the great Texcoco Lake once was), Xochimilco, and Zumpango, all with major modifications (Figure 1). Texcoco and Xochimilco are currently protected natural areas considered wetlands and designated as Ramsar sites (Numbers 2469 and 1,363, respectively); however, both lakes receive the effluents of different wastewater treatment facilities (López-López et al., 2016; Morales-García et al., 2020). For its part, Lake Zumpango, with a storage capacity of 100 million m³, functions like a regulating vessel. During approximately 2 months in the rainy season, it receives part of the water from the Cuautitlán River and the Emisor Poniente through the Santo Tomás canal. The water thus stored is used for agriculture, and the rest is diverted to the “Gran Canal del Desagüe”, which drains into the Pánuco basin. Lake Chalco is mainly filled with rainwater; however, municipal and industrial wastewater treated and untreated is also discharged into it (Ortiz-Zamora and Ortega-Guerrero, 2007). The original connectivity between lakes was lost due to urban settlements in the former lacustrine area.

When Alexander von Humboldt visited Mexico, in addition to analyzing the water of the BM lakes, he drafted a map of the BM, including the lakes, which had already lost a considerable area by that time (the early 1800s). In the present work, we studied the same lakes Humboldt analyzed, except for Lakes Xaltocan and San Cristobal since both became extinct (Figure 2).

We conducted a morphometric analysis of the BM to determine its perimeter, maximum and mean areas, and the compactness coefficient (defined as the ratio between the basin perimeter and the circumference of a circle with an area equal to the basin area, as proposed by Gravelius, 1914). The model designed in this work from the surface of the Great Lake of Mexico was developed using the Digital Elevation Model (DEM) by Hydrosheds.org (2023), with a 3s resolution, and public shapefiles from Mexico’s Institute of Statistics, Geography and Computer Sciences (Instituto Nacional de Estadística, Geografía e Informática; INEGI, 2023). The map entitled Carte de la Vallée de Mexico, drafted by Humboldt and published in 1811, was obtained from the website at David Rumsey Map Collection (2023). This map was georeferenced by the nearest neighbor method, using the landmarks marked in Humboldt’s map to elaborate a new map superimposed on the DEM of Mexico Basin mentioned above. All spatial analyses were performed using the QGIS geographic information system (open-source system).

The water bodies studied were monitored in two contrasting seasons: dry (March 2022) and rainy (October 2022). In each sampling period, the following environmental variables were recorded with the use of a Quanta^® multiparameter probe (Hydrolab DS5): water and air temperature (°C), dissolved oxygen (DO mg L^-1), pH, salinity (UPS), turbidity (NTU), and conductivity (µS cm^-1). Additionally, 500 mL water samples were collected in polyethylene bottles, in duplicate. Samples were transferred in the dark and refrigerated (4 °C) for subsequent testing in the laboratory. The following water quality parameters were determined with HACHDR 3900, Hach^® spectrophotometer techniques: nitrites (NO₂, mg L^-1), nitrates (NO₃, mg L^-1), ammonium (NH₄, mg L^-1), total nitrogen (TN, mg L^-1), orthophosphates (PO₄, mg L^-1), total phosphorus (mg L^-1), alkalinity (mg L^-1), biochemical oxygen demand (BOD₅, mg L^-1), color (U-PtCo), and total suspended solids (TSS mg L^-1). Total hardness was quantified by titration with EDTA (CaCO₃, mg L^-1). Alkalinity (CaCO₃, mg L^-1) was quantified by titration according to APHA (2005) techniques. In addition, at each study site, 100 mL of water was collected in Whirl-Pak bags for microbiological analyses. These water samples were tested for total and fecal coliforms according to the APHA MPN technique (2005).

Separately, chlorophyll a (Chl a) was quantified as per the standardized APHA technique (APHA, 2005). To this end, 500 mL of water collected from each study site was filtered through Whatman filters (0.45 µm), followed by extraction in acetone (90%) for 24 h in the dark under refrigeration (4 °C) prior to the Chl a analysis.

Another 50 mL water sample was acidified and digested to determine Cl⁻, SO₄ ^2-, Na⁺, K⁺, Ca²⁺, and Mg²⁺ using Inductively Coupled Plasma Optical Emission Spectrometry (ICP OES) techniques (Perkin Elmer Optima 8,000) at the CICATA Legaria laboratory of the Instituto Politécnico Nacional.

To obtain a single parameter indicating the water quality characteristics of the current lakes, we calculated the water quality index (WQI) proposed by Dinius (1987) for each study lake and period. This index includes 13 environmental variables: dissolved oxygen saturation (%), atmospheric and water temperature, pH, biochemical oxygen demand, alkalinity, nitrates, conductivity, chlorides, hardness, true color, and fecal and total coliforms. The output of this WQI ranges from 0 to 100, which is suitable for a better understanding of water quality and decision-making on water use, and as support of management programs. Mean WQI values were assessed using two different data sets: individual lakes (WQI values for all study periods), and study periods (WQI values for all study lakes for each study period).

All data were assessed for normality and homoscedasticity. Then, ANOVA followed by Tukey’s multiple comparison test or Kruskal–Wallis followed by Duncan’s multiple range test were performed for parametric or non-parametric data, respectively, to test the data for significant differences between lakes and between study periods. Box-plot graphs were constructed to depict the physicochemical and biological variables evaluated, by study lake and study period, using the XLSTAT software (Addinsoft, 2023). Heatmaps were constructed with the mean values ​​and the standard error of the main physicochemical characteristics evaluated. A principal component analysis (PCA) was performed, after a factorial analysis of all the physicochemical factors evaluated, to identify the variables with the greatest contribution to the trends between study sites. The data included the four lakes studied, the factors recorded in the field, and the results of the physicochemical testing carried out in the laboratory; all the variables were transformed to log(x+1). The PCA was performed with Pearson’s correlation using the XLSTAT software (Addinsoft, 2023).

In the case of Na⁺, Ca²⁺, K⁺, Mg²⁺, Cl⁻, and HCO₃ ⁻, a Piper diagram was constructed using the free software Easy Quim V.5 (Vazquez-Suñe and Serrano-Juan, 2012) to classify the lakes according to their ionic composition.

As for the information generated by Alexander von Humboldt, we reviewed his diary (Humboldt, 1802–1804). From it, we obtained his records about the test results of the water samples collected from the center of the lakes (Zumpango, San Cristobal, Texcoco, Xochimilco, and Chalco). These tests and observations included color, odor, taste, density, and the reactions from testing water samples with turmeric Curcuma paper (alkalinity) and the reactants lead acetate (hydrogen sulfide), silver nitrate (chloride), lead nitrate (sulfate), lime water (carbonates), and nitric acid (bicarbonates) (Richter and Engshuber, 2014).

With Humboldt’s results, a table was elaborated to summarize all the test results. Given that the results obtained by Humboldt are semi-quantitative, they were coded to provide quantitative data for a PCA. The coding involved using a 0-to-5 scale, where 0 means no reaction to the test and 5 indicates the maximum reaction reported by Humboldt. Variables such as colour and taste, were considered as categorical. The PCA aims to identify the similarities and differences between lakes, as well as any environmental gradients in the water bodies based on the lake parameters recorded by Alexander von Humboldt. The PCA was performed with Pearson coefficient and using the XLSTAT software (Addinsoft, 2023). The same procedure used for the PCA of our data was used for Humboldt´s data.

The analysis of basin morphometry indicates that the BM comprises an area of 9,219.3 km², with an approximate total length of 142.5 km in its main axis, oriented in a southwestern–northeastern direction, a maximum width of 112.7 km, a perimeter of 529.8 km, and a compactness coefficient score of 1.55. According to Faye and Ndiaye (2021), the compactness coefficient (IK) can be used to classify lakes into four different shapes: circular (1), squat (1–1.15), intermediate (1.15–1.5), and elongated (1.5 and above); therefore, the BM has an elongated shape, with its main axis oriented as mentioned above (Table 1). Some of these values differ from those obtained by Arce et al. (2019), namely, 9,620 km² in area, 100 km in length, and 80 km in width; separately, Alcocer and Williams (1996), based on the work of Alvarez and Navarro (1957), reported a total length of 125 km, 90 km in width, and a mean area of 9,600 km², which are similar to the figures reported by Arce et al. (2019).

In this study, using GIS tools, the Digital Elevation Model (DEM) from Hydrosheds.org (2023), and considering 2,250 masl as the high watermark of the lakes as proposed by Torres-Alves and Morales-Nápoles (2020), we derived a new shape for the Great Lake of Mexico, estimating an area of 1,280 km² and a perimeter of 644.2 km (Figure 1). Based on this new shape, it is possible to observe that the hills named La Estrella, Yuhualixqui, El Marquez, and El Elefante, and the volcanoes Xico, Guadalupe, La Caldera, and Xaltepec were previously islands located within the Great Lake of Mexico, as well as the area that currently harbors the Tultepec and Melchor Ocampo towns, located in the northern area of the former lake, since this area exceeds 2,250 m asl. This result is consistent with the alluvial deposits in the BM (Supplementary Material, Figure 2). Several authors have proposed models that estimate the surface area of the Great Lake of Mexico. Cruickshank García (1998) estimated that the lake region measured almost 2000 km²; Torres-Alves and Morales-Nápoles (2020) indicated an area of 1,000 km²; and Alcocer and Williams (1996) calculated an area of 7,868 km².

Therefore, considering the original lacustrine area of 1,280 km² as the original (baseline) area (100%), the remaining lacustrine area in Humboldt’s time (1803) was 428.26 km², accounting for 33.46% of the original area. This implies a loss of two-thirds of the original lake area by the early nineteenth century. Today, the remaining lacustrine area is a mere 2.83% of the original area of the Great Lake of Mexico (Table 2). Given its current role as a regulating vessel, Lake Zumpango has increased in area by 7.66% compared to the area shown in Humboldt’s map (Figure 2). The lake with the greatest loss of area from Humboldt’s time to date is Lake Texcoco (95.57%). For its part, Lake Xochimilco represents 11.32% of the area of all lakes in Humboldt’s time; this scenario may be due to the increasing number of chinampas being built in this area. Tussupova et al. (2020) state that many saline lakes around the world are drying up rapidly due to anthropogenic activities, causing local adverse effects on health (lung diseases) air quality (excess dust), and ecological impacts (e.g., biodiversity), among others. In the case of the BM lakes, the decision to drain the lakes and urban growth since the Spanish colonial period has brought about major consequences.

These results highlight the extreme reduction of the lake area in the BM, which has brought with it a shift in the cultural perception of the BM landscape from a lakescape in the pre-Hispanic era to an urban landscape today (Biar, 2020).

The values of the water variables tested in the five lakes in the early 1800s and obtained from Humboldt´s diary (Humboldt, 1802–1804) are summarized in Table 3.

Based on Table 3, Lakes Texcoco and Chalco were denser and more saline than the rest of the BM lakes; however, Lake Chalco had colorless clear water (i.e., no turbidity). On the other hand, Lake Xochimilco water was odorless and showed the lowest density. Currently extinct, Lake San Cristobal had turbid water with a density intermediate between Lakes Texcoco and Xochimilco. Contrary to some reports (Berres, 2000; Torres-Alves and Morales-Nápoles, 2020), Lake Chalco was relatively saline, not entirely freshwater. Humboldt recorded this water characteristic in the early 1800s, and Caballero and Ortega-Guerrero (1998) mentioned that Lake Chalco had been saline in some periods of the recent past.

The principal component analysis of the lakes based on the variables evaluated by Humboldt shows an explained variance of 83.23% in its first two components (Figures 3A, B). In the convex hulls diagram, Lakes Texcoco, Chalco, and Xochimilco are separate units, while Lakes Zumpango and San Cristobal are grouped in the same cluster (Figure 3A). The dispersion of the lakes along the environmental gradients formed by the variables tested by Humboldt is notorious. On the far right (Figure 3B), we can see the vectors of the variables that are associated, including density, and several tests such as the turmeric Curcuma paper, Pb(NO₃)₂, AgNO₃, lime water, Pb(Ac)₂, HNO₃, sodium carbonate taste, hydrogen sulfide odor, and yellowish color, all of which attain the highest values ​​at the far right of the diagram; Lake Texcoco Lake is located at this end of the gradient. In contrast, on the far left of the diagram, the vectors corresponding to transparency, alkaline taste, colorless, greyish color, and musty taste are observed. Lake Xochimilco is located on the far left of the diagram, characterized by its non-turbid, colorless water with an alkaline taste, characteristics that contrast with those of Lake Texcoco. The environmental conditions in Lakes Xochimilco and Texcoco also form a gradient ranging from less dense and less saline water to denser and more saline conditions. Along this gradient, Lakes Zumpango and San Cristobal are positioned on the lower left quadrant, characterized by grayish water with greater turbidity and a musty taste; regarding the tests with reactants, these lakes were less reactive than Texcoco but more than Xochimilco. Finally, Lake Chalco, positioned in the upper right quadrant, was characterized by reacting to turmeric curcuma paper and silver nitrate, and by a higher density than that of Lakes Xochimilco, Zumpango, and San Cristobal, but lower than that of Lake Texcoco. Lake Chalco was positioned opposite to Lakes Zumpango and San Cristobal since its water was more transparent, colorless and with an alkaline taste, and reacted to silver nitrate and turmeric paper (Figure 3B).

According to Humboldt’s results, the high Cl⁻, HCO₃ ⁻, and CO₃ ^2- contents are worth noting; together with alkalinity, these results reveal the alkaline soda-lake nature of Lake Texcoco. According to Jones et al. (1998), soda lakes are characterized by the predominance of NaCl, NaHCO₃, and Na₂CO₃, which are consistent with the findings of Humboldt from his qualitative chemical tests. In contrast, Lake Xochimilco turned out to be a freshwater lake with conditions opposite to those of Lake Texcoco, Lake Chalco stood out for its high alkalinity, with carbonate-rich waters, and ranked second in terms of its high density. Lakes Zumpango and San Cristobal showed intermediate characteristics between Lake Texococo (soda lake) and Lake Xochimilco (freshwater lake). Based on the above, the BM lakes showed an environmental gradient ranging from freshwater to saline soda lakes, with a strongly alkaline lake that does not reach a soda-lake state (Lake Chalco). Likewise, the spatial view of the BM is consistent with Humboldt’s perception in the early 1800s that everything was interconnected (Holl, 2018). Humboldt claimed that the lack of water might turn the valley sterile and unhealthy, increasing the salinity and aridity. He observed that the aquatic plants covering the lakes released hydrogen sulfide, which could be perceived when the wind blew across Lake Texcoco (Humboldt, 2003, 256).