SPATIO-TEMPORAL ESTIMATION OF SOIL MOISTURE IN A TROPICAL REGION USING A REMOTE SENSING ALGORITHM Liliana Marrufo (1) , Fernando González (1) , Alejandro Monsiváis-Huertero (2) and Judith Ramos (1) (1) Instituto de Ingeniería, Universidad Nacional Autónoma de México Av. Universidad 3000, 04510 Coyoacán, México, D.F. Email: LMarrufoV@iingen.unam.mx (2) ESIME Ticomán, Instituto Politécnico Nacional, Av. San José Ticomán No. 600, Gustavo A. Madero, C.P. 07340, México D.F. ABSTRACT To achieve a soil moisture estimation (m s ) in an accurate way is crucial to understand the water cycle response and avoid traditional runoff estimations where the m s is assumed as a constant value. The aim of this study is the implementation of optical-radar images into a model to estimate m s in the Zapotes Lagoon System in Tabasco, Mexico. The satellite images used were Landsat TM and ETM+ sensors, Envisat and also in situ measurements and MDT were available. The field measurements at the soil profile showed a clear pattern of the water movement into the basin corresponding to the lower parts of the system (lagoons). The land use analysis obtained with the optical images indicated a strong change in the floodplain due to the construction of protect barriers around the Villahermosa city losing its hydrological capacity. This allows the identification of three main covers (soil, vegetation and air) that were monitored and feed to the model (MIMICS) in order to estimate m s using the Envisat image. Results provided an empirical equation that relates the m s with the backscattering coefficient. Index Terms— soil moisture, MIMICS, ENVISAT, microwave remote sensing, supervised classification 1. INTRODUCTION The soil moisture (m s ) is an important factor in the system soil-plant-atmosphere due to the interchange of energy and mass fluxes at regional and local scales. The m s is influenced by climatological and surface conditions that modify its spatio-temporal characteristics [1, 3, 5]. The soil moisture content can be measured by different methods such as gravimetry, lisimitry, tensiometry, and reflectometry. These methods are highly accurate but are limited to punctual discrete measurements that cannot be extrapolated to a regional scale. Thus, to characterize extensive zones such as wetlands, it is necessary the design of expensive measurement campaigns. To circumvent this inconvenient, remote sensing (RS) techniques have been developed as an alternative to analyze spatially and temporally the m s behaviour in large areas in a synoptic way. Both active (radar) and passive (radiometer) microwave sensors, at frequencies lower than 10 GHz, have shown to be sensitive to variations in near-surface soil moisture [4]. The passive sensors provide data with low spatial resolution. In contrast, the active sensors collect data with high spatial resolution, allowing the synoptic estimation of m s [6]. The active microwave sensors have a major sensitivity to the dialectical properties at the surface. In particular, this sensitivity is associated to the permittivity which is related to the soil moisture. The active sensors estimate the backscattering coefficient (ı°) and have a major coverage at the surface due to its high spatial resolution. Also, as the frequency is lower there is a reduction to the attenuation effects associated to the atmosphere and vegetation. The ı° is composed of a direct contribution from soil and vegetation and the interactions between soil and vegetation. Thus, the ı° from soil is a function of the soil texture and the soil dielectric constant that depends upon the m s . Most of the algorithms to estimate m s [2, 8] have been implemented to semi-arid climates, where soils have a low water retention and low vegetation cover. However, the validity of such algorithms to study the effects of high temporally space variability in m s over tropical regions still remains. In tropical regions, and particularly wetlands, the soil has a dense vegetation cover and the layer containing the m s goes from 0-1.5m depth (vadose zone). This study proposes the implementation of an algorithm based on the integration of optical and radar images to estimation the m s in tropical zones analysing different soil water saturation conditions in a wetland located in Tabasco, South East of Mexico. 2. SITE OF STUDY The Zapotes Lagoon System is a floodplain of a riparian zone characterised by a low internal drainage and a water table within a depth of 50-150 cm. The Zapotes system is complex integrated by rivers and streams that respond to some hydraulic structures (i.e. barrires) and processes (i.e runoff) taking place within and without the system. These processes can be classified as inter-riparian (river) and trans-riparian (climate, geology, altitude, inundation stay time, slope, humidity gradients, etc) and the occurrence of periodical hurricanes. The study area is located at the southeast of the Subregion Centro in the Tabasco state, Mexico, in the coordinates 18º 00’N and 92º52’O (Fig.1). 3089 978-1-4577-1005-6/11/$26.00 ©2011 IEEE IGARSS 2011