Artículos

 

Modern sands of the Gulf of Mexico: Discriminating fluvial and coastal sand composition

 

Arenas actuales del Golfo de México: Discriminación entre la composición de arenas fluviales y costeras

 

J.J. Kasper-Zubillaga^1* and A. Carranza-Edwards¹

 

¹ Instituto de Ciencias del Mar y Limnología UNAM, Circuito Exterior s/n, Cd. Universitaria Coyoacán, 04510, México D.F., México. *E-mail: kasper@icmyl.unam.mx

 

Recibido en marzo de 2003;
aceptado en julio de 2003.

 

Abstract

A forward stepwise linear discriminant analysis, A-Nova test and R-mode factor analysis were used on modern beach, dune and river sands to discriminate the geographic distribution of detrital modes between sands sourced by volcanic rocks (Trans-Mexican Volcanic Belt (TMVB) sands) and sedimentary rocks (Veracruz Embayment (VE) sands). Thirty-three beach, dune and river sand samples were collected across beach-dune profiles, in the inner reaches of the channel and at the river mouth during the dry season. The source terranes for these sands are constituted by Miocene to Pliocene volcanites and by Tertiary to Quaternary siliciclastic rocks. To avoid the constrained character of compositional data, point-count percentages were expressed as compositional modal Log ratios (i.e., Log Qt/Ft = total quartz/total feldspars, Log Qt/Lt = total quartz/total lithics, Log Ft/Lt = total feldspars/total lithics, Log Qm/Fk = monocrystalline quartz/potash feldspars, Log Qm/Fp = monocrystalline quartz/ plagioclase, Log Fk/Fp = potash feldspars/plagioclase, and Log Lv/Ls = volcanic lithics/sedimentary lithics). The linear discriminant and factor analyses showed that the geographical dispersal of detrital modes between the TMVB sands and VE sands are statistically significant by means of the Log Lv/Ls, Log Qt/Ft, Log Qt/Lt and Log Ft/Lt ratios. The A-Nova test and post-hoc analyses showed that the mean differences of the Log Lv/Ls, Log Qt/Ft and Log Qt/Lt ratios are highly significant for the VE river sands and the TMVB dune sands. The composition of the TMVB dune sands is controlled by a longshore drift of quartz supply. The VE river sands are depleted in Ls due to the mechanical and/or chemical weathering of Ls and the influence of TMVB on the composition of the VE river sands.

Key words: modern and detrital modes, linear discriminant analysis, A-Nova test, factor analysis, SW Gulf of Mexico.

 

Resumen

Se realizó un análisis de discriminación lineal, una prueba A-Nova y un análisis factorial en arenas actuales de playa, duna y río para discriminar la distribución geográfica composicional entre arenas derivadas de fuentes volcánicas (arenas del Cinturón Volcánico Trans-Mexicano, CVTM) y sedimentarias (arenas de la Cuenca Veracruzana, CV). Se recolectaron 33 muestras de arena de playa, duna y río a lo largo de perfiles de playa-duna, en la desembocadura y el lecho distal de los ríos en la época de secas. La roca fuente está constituida por rocas volcánicas del Mioceno y Plioceno y rocas silicoclásticas del Terciario al Cuaternario. Para evitar el carácter "cerrado" de los datos composicionales, los porcentajes obtenidos del conteo de granos se expresaron como índices logarítmicos (Log Ct/Ft = cuarzo total/feldespato total, Log Ct/Lt = cuarzo total/líticos totales, Log Ft/Lt = feldespato total/líticos totales, Log Cm/Fk = cuarzo monocristalino/feldespato de potasio, Log Cm/Fp = cuarzo monocristalino/plagioclasas, Log Fk/Fp = feldespato de potasio/plagioclasas y Log Lv/Ls = líticos volcánicos/líticos sedimentarios). El análisis de discriminación lineal y el análisis factorial mostraron que las diferencias geográficas composicionales de la arena de playa, duna y río del CVTM y de la CV son estadísticamente significativas mediante los índices Log Lv/Ls, Log Ct/Ft, Log Ct/Lt y Log Ft/Lt. Más aún, las pruebas A-Nova y post-hoc mostraron diferencias significativas en las medias de los índices Log Lv/Ls, Log Ct/Ft y Log Ct/Lt para la arena de duna del CVTM y de la arena de río de la CV. La composición de la arena de duna del CVTM está influída por transporte de cuarzo a lo largo de la costa. La ausencia de Ls en la composición de la arena de río de la CV está determinada por el intemperismo mecanico y/o químico de los Ls y la influencia del CVTM en la composición de la arena de río de la CV.

Palabras clave: arenas actuales y procedencia, análisis de discriminación lineal, prueba A-Nova, análisis factorial, Golfo de México sur occidental.

 

Introduction

Sand detrital modes are controlled by the source-rock composition, climate, relief and drainage system (Basu, 1976; Ibbeken and Schleyer, 1991; Le Pera and Critelli, 1997). Beach and dune sands reflect provenance from continental sources, and are controlled by longshore transport and fluvial discharges (Komar, 1976; Critelli et al., 1997). The spatial distribution of beach, dune and river sand composition, however, has not yet been tested using multivariate analyses. Moreover, multivariate methods such as A-Nova tests and post-hoc group comparisons have not been included together with the linear discriminant analysis (LDA) and factor analysis (FA) to test for significant variables and group differences. The purpose of this study is to detect the spatial distribution of the sand modes during the dry season in order to assess the feasibility of such an approach for the characterization of sand provenance and discrimination among sand composition environments of the area by means of LDA, A-Nova and FA.

 

Study area

The area is located in the southwestern coastal Gulf of Mexico, at 19°25'-19°48' N latitude and 96°20'-96°27' W longitude (fig. 1). In the Tampico Embayment (TE) and Veracruz Embayment (VE), the relief corresponds to plains, while in the Trans-Mexican Volcanic Belt (TMVB), the mountainous relief reaches the coastline (Lugo-Hubp et al., 1990). In figure 2 we observe how the altitude (200 and 1000 m) contour lines delineate these geomorphologic embayments. The terrain gradient is maximum in the TMVB sand zone (cf. slope 0.09500), and gradually decreases towards the north (cf. slope 0.00704) and south (cf. slope 0.00487). Across the VE sand zone, the average gradient is 0.01900. Between the mountainous relief and the plains there is an area of hills. In the TMVB sand zone, torrential streams dominate, whereas the VE sands are directly affected by rivers that originate in lands higher than 1000 m in the mountainous relief areas with high precipitation (sometimes higher than 4000 mm) (fig. 3). In the study area, the rainy season is from June to October and the dry season from November to May (CETENAL, 1970), with a dominant humid semi-tropical climate. The area is a wave-dominated zone (waves 0.9 m in height), influenced by surface water currents with a southward component in spring and average velocities of 6 cm s^-1 (Fernández-Eguiarte et al., 1992a, b). The study area is drained by torrential streams close to TMVB and by two rivers with a perennial regime in VE (figs. 1, 2). The area is the extension of the TMVB (Negendank et al., 1985), characterized by Miocene-Pliocene andesites of calcalkaline composition (Cantagrel and Robin, 1979; Negendank et al., 1985) and alkaline basalts (Negendank et al., 1985) (fig. 1) inferred to reflect the Cocos Plate subduction. The TE and VE are sedimentary deposits composed of Tertiary and Quaternary alluvial strata of siliciclastic and minor carbonate composition (fig. 1).

 

Materials and methods

A total of 33 sediment samples were collected. Sand sampling was carried out only during the dry season where the VE river sands are still draining towards the sea (fig. 1). Petro-graphic analysis was carried out on samples characterized by sandy mode fraction following the Indiana traditional method of point counting of 250 grains. A forward stepwise LDA was carried out using petrographic ratios and the mean grain-size (Mz) based on the petrographic and textural data from Kasper-Zubillaga and Dickinson (2001) (table 1). Also, an A-Nova test, post-hoc Tukey comparison tests and R-mode FA were carried out to assess the relations and differences among the compositional modal variables of the sands. The ratios were expressed as Log Qt/Ft (total quartz = mono and polycrystal-line quartz/total feldspars = potash feldspars + plagioclase), LogQt/Lt (total quartz/total lithics = volcanic + sedimentary), Log Ft/Lt (total feldspars/total lithics), Log Qm/Fk (monocrys-talline quartz/potash feldspars), Log Qm/Fp (monocrystalline quartz/plagioclase), Log Fk/Fp (potash feldspars/plagioclase) and Log Lv/Ls (volcanic lithics/sedimentary lithics) (table 2). This was done to avoid the constrained character of compositional data (Weltje, 2001). For the LDA, the variables were entered in a N = 33 x n = 8 matrix, in which Nis the number of samples and n is the number of variables. For the A-Nova tests, each column was labeled with each variable to test separately for N = 33. For the R-mode FA, the samples and variables were grouped in a N = 19 x n = 8 matrix for the beach, dune and river sands close to TMVB, and in a N = 14 x n = 8 matrix for the southern beach, dune and river sands close to VE (figs. 1, 2). Hence, for the FA, samples a, b, c, d, e, f, g, m, n, o, p, q, r, s, 1, 2, 3, 4 and 5 were grouped as TMVB sands in the FA correlation matrix, and samples h, i, j, k, l, t, u, v, w, x, 6, 7, 8 and 9 were grouped as VE sands in the FA correlation matrix.

 

Results

The Qt-Ft-Lt ternary plots of beach, dune and river sands are shown in figure 4. Overall, the TMVB and VE sands are feldspatholithic (Qt30Ft24Lt46). Furthermore, significant differences between the overall composition of the TMVB and VE sands, in percent values (table 1), were determined. This was carried out with a one-tailed t-test for Qt, Ft, Lt, Lv and Ls, in which the null-hypothesis of equal means is rejected between the TMVB and VE sands for quartz abundances (t_observed = 2.09, df = 31, p = 0.05, n = 33, t_critical = 1.69). Additionally, the LDA results show, from the partial lambda values, that the best variables that discriminate between the TMVB and VE sands are the Log Qt/Ft, Log Qt/Lt, Log Ft/Lt and Log Lv/Ls ratios (table 3). Moreover, the best canonical score plots were between canonical variate 1 vs canonical variate 2 (root 1 vs root 2) and canonical variate 2 vs canonical variate 3 for the dune and river TMVB and VE sands (figs. 5, 6). The best separated groups were the TMVB river sands distinct from the VE river sands, followed by the TMVB beach sands distinct from the VE beach sands, and the TMVB dune sands distinct from the VE dune sands (fig. 5). Discrimination between the TMVB and VE beach, dune and river sands is controlled by the Log Qt/Ft, Log Qt/Lt, Log Ft/Lt and Log Lv/Ls ratios. However, a comparison of means test may elucidate the contribution of the variable in terms of separate groups. Therefore, an A-Nova test and post-hoc Tukey comparison of means test were carried out, and we found that the TMVB and VE dune and river sands are significantly different when applying these tests (table 4). Moreover, factor 1 grouped the Log Qt/Ft, Log Qt/Lt, Log Qm/Fk, Log Qm/Fp and Log Lv/Ls ratios, and Mz as high variable loads for the TMVB sands (table 5). Additionally, factor 1 grouped the Log Qt/Ft, Log Qt/Lt, Log Ft/Lt, Log Qm/Fp and Log Lv/Ls ratios for the VE sands. This suggests that the composition of the TMVB beach, dune and river sands is controlled by the Log Qt/Ft, Log Qt/Lt, Log Qm/Fk, Log Qm/Fp and Log Lv/Ls ratios, as well as Mz, and that the VE sand composition is controlled by the Log Qt/Ft, Log Qt/Lt, Log Ft/Lt, Log Qm/Fp and Log Lv/Ls ratios (table 5). Nevertheless, only the Log Qt/Ft, Log Qt/Lt and Log Lv/Ls ratios are the three significant variables entered in the LDA that contribute to separate among TMVB and VE dune and river sands. These results are based on the overall statistical significance of the LDA, A-Nova, post-hoc and FA tests for the three most common significant variables entered in the analyses (i.e., Log Qt/Ft, Log Qt/Lt and Log Lv/Ls ratios) (tables 3, 4, 5).

 

Discussion

It can be observed that dune sands tend to concentrate towards the Qt pole, whereas the beach and river sands tend to concentrate towards the Lt pole (fig. 4). This suggests a quartz-rich trend for the dune sands. Moreover, the t-test results of null-hypothesis rejection suggest that the overall quartz abundances of the sand samples from TMVB and VE are significantly different. Additionally, the LDA, A-Nova and FA results showed that the Miocene-Pliocene andesites and basalts control the VE river sand composition. This is supported by the highest concentration of Lv in the VE river sands (a more positive Log Lv/Ls ratio than the Log Lv/Ls ratio from the TMVB river sands) (tables 1, 2), which suggests: (a) that abrasion of Ls is occurring throughout the long VE rivers due to the longer chemical/mechanical weathering acting upon the Ls in the VE river sands (Arribas et al., 2000); and/or (b) that VE rivers are highly influenced by the TMVB supply at their heads, giving the VE river sands minimal maturation despite the long detritus transport (Le Pera et al., 2001) (fig. 1; tables 1, 2). The Log Qt/Lt ratio also controls the compositional differences between the TMVB dune sands and VE dune sands. A quartz supply derived from TE by a longshore drift southwards may be contributing to the composition of the TMVB dune sands by effect of marine/aeolian selective transport that enriches the dune sands with more quartz in the TMVB dune sands compared to the VE dune sands (Kasper-Zubillaga and Dickinson, 2001; Frihy et al., 1999) (table 2). This quartz enrichment effect in the dune may also be related to depletion of labile grains in humid climatic conditions (Mack, 1984). Hence, the major compositional control of the TMVB dune and VE river sands is based on the quartz content of the dune and the volcanic lithic content in the river.

Multivariate procedures showed that the Log Qt/Ft and Log Qt/Lt ratios and the Log Lv/Ls ratio are the three significant variables entered in the LDA that contribute to separate among TMVB and VE dune and river sands. Moreover, Log Qt/Ft and Log Lv/Ls are the two most important variables with geological significance to discriminate between TMVB and VE dune sands and TMVB and EV river sands. TMVB dune sand and VE river sand composition is controlled by a longshore drift of quartz supply, depletion of Ls by mechanical and/or chemical weathering and the influence of TMVB on the VE river sand composition. Multivariate procedures are a useful tool to elucidate the relationship between source rocks and sediments.

 

Acknowledgements

Financial support was provided by CONACYT (Mexico's Council of Science and Technology, project 3477T). Reviews by E. Le Pera and S. Critelli are highly appreciated. We thank Eduardo Morales de la Garza for his collaboration with the drawings.

 

References

Arribas, J., Critelli, S., Le Pera, E. and Tortosa, A. (2000). Composition of modern stream sand derived from a mixture of sedimentary and metamorphic source rocks (Henares River, Central Spain). Sedimen. Geol., 133: 27-48.         [ Links ]

Basu, A. (1976) Petrology of Holocene fluvial sand derived fromplutonic source rocks: Implications for paleoclimatic interpretation. J. Sediment. Petrol., 46: 694-709.         [ Links ]

Cantagrel, J.M. and Robin, C. (1979). K-Ar dating in eastern volcanic rocks - relations between the andesite and the alkaline provinces. J. Volc. Geoth. Res., 5: 99-114.         [ Links ]

CETENAL (1970). Carta de Climas (Veracruz 14Q-VI). Scale 1:500,000. Secretaría de la Presidencia and Instituto de Geografía, México.         [ Links ]

Critelli, S., Le Pera, E. and Ingersoll, R.V. (1997). The effects of source lithology, transport, deposition and sampling scale on the composition of southern California sand. Sedimentology, 44: 653-671.         [ Links ]

Fernández-Eguiarte, A., Gallegos-García, A. y Zavala-Hidalgo, J., (1992a). Oceanografía Física 1: Masas de Agua y Mareas de los Mares Mexicanos. Atlas Nacional de México (Sheet IV.9.1, summer). Instituto de Geografía, UNAM, México.         [ Links ]

Fernández-Eguiarte, A., Gallegos-García, A. y Zavala-Hidalgo, J., (1992b). Oceanografía Física 1: Masas de Agua y Mareas de los Mares Mexicanos. Atlas Nacional de México (Sheet IV.9.2, winter). Instituto de Geografía, UNAM, México.         [ Links ]

Frihy, O.E., El-Askary, M.A., Deghidy, E.M. and Moufaddal, W.M. (1999). Differentiating fluvio-marine depositional environments in the Nile Delta using textural and compositional components. J. Afr. Earth Sci., 28: 599-618.         [ Links ]

Ibbeken, H. and Schleyer, R. (1991) Source and Sediment. Springer Verlag, Berlin, 286 pp.         [ Links ]

Kasper-Zubillaga, J.J. and Dickinson, W.W. (2001). Discriminating depositional environments from modern source terranes using modal analysis. Sediment. Geol., 143: 149-167.         [ Links ]

Komar, P.D. (1976). Beach Processes and Sedimentation. Prentice Hall, New Jersey, 429 pp.         [ Links ]

Le Pera, E. and Critelli, S. (1997) Sourceland controls on the composition of beach and fluvial sand of the northern Tyrrenian coast of Calabria, Italy: Implications for actualistic petrofacies. Sediment. Geol., 110: 81-97.         [ Links ]

Le Pera, E., Arribas, J., Critelli, S. and Tortosa, A. (2001) The effects of source rocks and chemical weathering on the petrogenesis of siliciclastic sand from the Neto River (Calabria, Italy): Implications for provenance studies. Sedimentology, 48: 357-378.         [ Links ]

Lugo-Hubp, J., Aceves-Quezada, F., Córdoba Fernández de Arteaga, C., Espinasa-Pereña, R., García-Arizaga, M.T. y Melo-Gallegos, C. (1990). El Relieve como Atractivo Natural. Estado de los Componentes Naturales del Medio Ambiente, Hoja V.2.1, scale 1:4000000. Atlas Nacional de México, Instituto de Geografía, Univ. Nal. Autón. México.         [ Links ]

Mack, G.E. (1984). Exceptions to the relationship between plate tectonics and sandstone composition. J. Sedimen. Petrol., 54: 212-220.         [ Links ]

Negendank, J.F., Emmermann, R., Krawczyk, R., Mooser, F., Tobschall, H. and Werle, D. (1985) Geological and geochemical investigation on the eastern Trans-Mexican Volcanic Belt. Geofís. Int., 24: 477-575.         [ Links ]

Vidal-Zepeda, R. (1990). Precipitación: Clima, Hoja IV.4.6, scale 1:4000000. Atlas Nacional de México. Instituto de Geografía, Univ. Nal. Autón. México.         [ Links ]

Weltje, G.J. (2001) Quantitative analysis of detrital modes: Statistically rigorous confidence regions in ternary diagrams and their use in sedimentary petrology. Earth Sci. Rev., 57: 211-253.         [ Links ]