Monitoring of water dynamics in rice ecosystems based on remote sensing time series

• Autores: Lucia Tornos Castillo
• Directores de la Tesis: Alicia Palacios-Orueta (dir. tes.), José Antonio Domínguez Gómez (codir. tes.)
• Lectura: En la Universidad Politécnica de Madrid ( España ) en 2017
• Idioma: español
• Tribunal Calificador de la Tesis: Francisco Javier Salas Rey (presid.), Concepción González García (secret.), María Inmaculada Aguado Suárez (voc.), Carlos J. Novillo Camacho (voc.), Silvia Merino de Miguel (voc.)
• Programa de doctorado: Programa de Doctorado en Investigación Forestal Avanzada por la Universidad Politécnica de Madrid
• Materias:
  • Matemáticas
    • Estadística
      • Series temporales
  • Ciencias de la tierra y del espacio
    • Geología
      • Teledetección (geología)
    • Hidrología
      • Aguas superficiales
• Enlaces
  • Tesis en acceso abierto en: Archivo Digital UPM
• Resumen

  • Water resources conservation is becoming an increasing concern worldwide. Rice ecosystems, as the consumer of 30% of freshwater resources used for crops worldwide (Karaba et al., 2007) are a key element to ensure water availability. Besides, rice fields hold important aquatic and terrestrial biodiversity, and produce crucial staple food at global level, being the most important element in the diet of half of the world’s population (Van Nguyen & Ferrero, 2006). In a context of increasing rice demand and water scarcity (Bouman et al., 2007), with serious implications in water resources availability and environmental conservation, obtaining information about water use in rice ecosystems is becoming a crucial need. However, the provision of reliable and updated information on rice paddy distribution and changes (Dong e al., 2015), as well the provision of accurate flooding and phenology detection methods is still difficult at global level. Furthermore, improving the estimation of evapotransporation (ET) and its temporal and spatial variations is essential to improve water management planning (Lee et al., 2004; Chemin and Honda, 2006).

    Remote sensing has become a key tool to monitor, map and observe rice ecosystems at different time and spatial scales (Kuenzer & Knauer, 2013), providing reliable information in an instantaneous and cost-effective manner. Time-series analysis of remote sensing data provides important information to assess ecosystem dynamics (Zhang e al., 2003. Additionally, the use of spectral indices time series is also a solid approach to identify intra-annual and interannual dynamics, and has been applied in the fields of land-use identification, phenology detection and natural phenomena monitoring, among others (Gumma et al, 2014; Shihua et al., 2014; Simonneaux et al., 2008).

    In the present thesis, a methodology is proposed to monitoring water and vegetation dynamics in rice ecosystems based on spectral indices time series, and the use of remote sensing time series in combination with ET models. The options explored are intended to be applied with low field-data requirements and inputs, aimed to be feasible in regions lacking general agricultural information, and especially in developing countries. We used MODIS time series to achieve results with enough time frequency and adequate spectral bands to detect both water and vegetation phenomena. The specific objectives were (1) to explore the potentiality of Spectral Indices annual time series to map rice ecosystems; (2) to assess the potential of different spectral indices for monitoring rice phenology and flooding dynamics by combining phenometric and statistical time series approaches and (3) to evaluate the Priestley Taylor- Jet Propulsion laboratory (PT-JPL) daily model performance in obtaining E (ET, expressed in terms of energy) estimates in rice ecosystems.

    The results of this thesis confirmed the potential of vegetation indices to provide reliable rice maps and highlight the importance of exploring angular indices to improve the identification of land cover dynamics. In particular, the use of the Shortwave Angle Slope Index (SASI) has demonstrated its potential to provide remarkable results in discriminating rice from other crops or land uses, in combination with spectral matching techniques. The existence of a significant response to flooding events showed in the Normalized Difference Water Index (NDWI(1), NDWI(2)) and SASI illustrates the importance of SWIR spectral wavelengths to detect flooded and wet soils. Specifically, SASI exhibited a strong capacity to identify changes in soil water content, which may encourage its use in wetland monitoring studies. Based on the results obtained, we have proposed a specific combination of indices for assessing rice flooding events and phenological stages in relation to different management practices. The PT-JPL-daily model reproduced adequately the main dynamics of  in rice ecosystems during the rice growing season, but it was less reliable in reproducing the repeated ups and downs present out of this period.