Rozprawy doktorskie na temat „Satellite Soil Moisture Retrievals”

La humedad del suelo es la variable que regula los intercambios de agua, energía, y carbono entre la tierra y la atmósfera. Mediciones precisas de humedad son necesarias para una gestión sostenible de los recursos hídricos, para mejorar las predicciones meteorológicas y climáticas, y para la detección y monitorización de sequías e inundaciones. Esta tesis se centra en la medición de la humedad superficial de la Tierra desde el espacio, a escalas global y regional. Estudios teóricos y experimentales han demostrado que la teledetección pasiva de microondas en banda L es optima para la medición de humedad del suelo, debido a que la atmósfera es transparente a estas frecuencias, y a la relación directa de la emisividad del suelo con su contenido de agua. Sin embargo, el uso de la teledetección pasiva en banda L ha sido cuestionado en las últimas décadas, pues para conseguir la resolución temporal y espacial requeridas, un radiómetro convencional necesitaría una gran antena rotatoria, difícil de implementar en un satélite. Actualmente, hay tres principales propuestas para abordar este problema: (i) el uso de un radiómetro de apertura sintética, que es la solución implementada en la misión Soil Moisture and Ocean Salinity (SMOS) de la ESA, en órbita desde noviembre del 2009; (ii) el uso de un radiómetro ligero de grandes dimensiones y un rádar operando en banda L, que es la solución que ha adoptado la misión Soil Moisture Active Passive (SMAP) de la NASA, con lanzamiento previsto en 2014; (iii) el desarrollo de técnicas de desagregación de píxel que permitan mejorar la resolución espacial de las observaciones. La primera parte de la tesis se centra en el estudio del algoritmo de recuperación de humedad del suelo a partir de datos SMOS, que es esencial para obtener estimaciones de humedad con alta precisión. Se analizan diferentes configuraciones con datos simulados, considerando (i) la opción de añadir información a priori de los parámetros que dominan la emisión del suelo en banda L —humedad, rugosidad, temperatura del suelo, albedo y opacidad de la vegetación— con diferentes incertidumbres asociadas, y (ii) el uso de la polarización vertical y horizontal por separado, o del primer parámetro de Stokes. Se propone una configuración de recuperación de humedad óptima para SMOS. La resolución espacial de los radiómetros de SMOS y SMAP (40-50 km) es adecuada para aplicaciones globales, pero limita la aplicación de los datos en estudios regionales, donde se requiere una resolución de 1-10 km. La segunda parte de esta tesis contiene tres novedosas propuestas de mejora de resolución espacial de estos datos: • Se ha desarrollado un algoritmo basado en la deconvolución de los datos SMOS que permite mejorar la resolución espacial de las medidas. Los resultados de su aplicación a datos simulados y a datos obtenidos con un radiómetro aerotransportado muestran que es posible mejorar el producto de resolución espacial y resolución radiométrica de los datos. • Se presenta un algoritmo para mejorar la resolución espacial de las estimaciones de humedad de SMOS utilizando datos MODIS en el visible/infrarrojo. Los resultados de su aplicación a algunas de las primeras imágenes de SMOS indican que la variabilidad espacial de la humedad del suelo se puede capturar a 32, 16 y 8 km. • Un algoritmo basado en detección de cambios para combinar los datos del radiómetro y el rádar de SMAP en un producto de humedad a 10 km ha sido desarrollado y validado utilizando datos simulados y datos experimentales aerotransportados. Este trabajo se ha desarrollado en el marco de las actividades preparatorias de SMOS y SMAP, los dos primeros satélites dedicados a la monitorización de la variación temporal y espacial de la humedad de la Tierra. Los resultados presentados contribuyen a la obtención de estimaciones de humedad del suelo con la precisión y la resolución espacial necesarias para un mejor conocimiento del ciclo del agua y una mejor gestión de los recursos hídricos.
Soil moisture is a key state variable of the Earth's system; it is the main variable that links the Earth's water, energy and carbon cycles. Accurate observations of the Earth's changing soil moisture are needed to achieve sustainable land and water management, and to enhance weather and climate forecasting skill, flood prediction and drought monitoring. This Thesis focuses on measuring the Earth's surface soil moisture from space at global and regional scales. Theoretical and experimental studies have proven that L-band passive remote sensing is optimal for soil moisture sensing due to its all-weather capabilities and the direct relationship between soil emissivity and soil water content under most vegetation covers. However, achieving a temporal and spatial resolution that could satisfy land applications has been a challenge to passive microwave remote sensing in the last decades, since real aperture radiometers would need a large rotating antenna, which is difficult to implement on a spacecraft. Currently, there are three main approaches to solving this problem: (i) the use of an L-band synthetic aperture radiometer, which is the solution implemented in the ESA Soil Moisture and Ocean Salinity (SMOS) mission, launched in November 2009; (ii) the use of a large lightweight radiometer and a radar operating at L-band, which is the solution adopted by the NASA Soil Moisture Active Passive (SMAP) mission, scheduled for launch in 2014; (iii) the development of pixel disaggregation techniques that could enhance the spatial resolution of the radiometric observations. The first part of this work focuses on the analysis of the SMOS soil moisture inversion algorithm, which is crucial to retrieve accurate soil moisture estimations from SMOS measurements. Different retrieval configurations have been examined using simulated SMOS data, considering (i) the option of adding a priori information from parameters dominating the land emission at L-band —soil moisture, roughness, and temperature, vegetation albedo and opacity— with different associated uncertainties and (ii) the use of vertical and horizontal polarizations separately, or the first Stokes parameter. An optimal retrieval configuration for SMOS is suggested. The spatial resolution of SMOS and SMAP radiometers (~ 40-50 km) is adequate for global applications, but is a limiting factor to its application in regional studies, where a resolution of 1-10 km is needed. The second part of this Thesis contains three novel downscaling approaches for SMOS and SMAP: • A deconvolution scheme for the improvement of the spatial resolution of SMOS observations has been developed, and results of its application to simulated SMOS data and airborne field experimental data show that it is feasible to improve the product of the spatial resolution and the radiometric sensitivity of the observations by 49% over land pixels and by 30% over sea pixels. • A downscaling algorithm for improving the spatial resolution of SMOS-derived soil moisture estimates using higher resolution MODIS visible/infrared data is presented. Results of its application to some of the first SMOS images show the spatial variability of SMOS-derived soil moisture observations is effectively captured at the spatial resolutions of 32, 16, and 8 km. • A change detection approach for combining SMAP radar and radiometer observations into a 10 km soil moisture product has been developed and validated using SMAP-like observations and airborne field experimental data. This work has been developed within the preparatory activities of SMOS and SMAP, the two first-ever satellites dedicated to monitoring the temporal and spatial variation on the Earth's soil moisture. The results presented contribute to get the most out of these vital observations, that will further our understanding of the Earth's water cycle, and will lead to a better water resources management.

Dans cette thèse, un algorithme de restitution à base de réseaux de neurones a été développé afin d’estimer l’humidité du sol à partir d’une combinaison d’observations satellitaires en micro-ondes, infrarouge et visible. Une estimation globale des valeurs mensuelles d’humidité du sol a été obtenue pour la période 1993-2000 et est fournie sur une grille à pixel de surface constante avec une résolution équatoriale de 0,25 ◦. Cette estimation de l’humidité du sol a été évaluée avec des données modélisées, des données de télédétec- tion et des observations in situ et a montré une bonne performance à différentes échelles spatiales et temporelles. Une analyse de contenu en information a montré que chacune des différentes observations satellites contribue à une information différente sur l’humidité du sol, avec les données micro-ondes actives plus sensibles à l’évolution temporelle et les données infrarouges thermiques reproduisant mieux les structures spatiales. En outre, une analyse de synergie a révélé que la combinaison de toutes les observations permet une réduction de l’incertitude de restitution de plus de 18 % et que la méthode des réseaux de neurones exploite de manière optimale la synergie des observations par comparaison avec autres approches. Une analyse a démontré la cohérence de l’humidité du sol resti- tuée avec d’autres produits satellitaires réprésentatifs d’autres paramètres hydrologiques (inondations, précipitations) à l’échelle du globe. Cela souligne le potentiel de nôtre jeu de données d’humidité du sol pour les études du cycle de l’eau terrestre. Enfin, il a été démon- tré que la méthode de réseaux de neurones proposée, constitue également un outil efficace pour évaluer les modèles de surface continentale ainsi que la modélisation des processus
In this thesis, a neural network based retrieval algorithm has been developed to compute surface soil moisture from a combination of microwave, infrared and visible satellite obser- vations. A global estimate of monthly mean soil moisture values has been computed for the period 1993-2000 and is provided on an equal-area grid with an equatorial resolution of 0. 25◦. This soil moisture estimate has been evaluated against modelled, remotely sensed and in situ observations and was found to perform well on different spatial and temporal scales. An information content showed that each of the various satellite observations con- tributes information about a different soil moisture variation, with the active microwave data being more sensitive to the temporal evolution and the thermal infrared data better capturing the spatial patterns. Furthermore, a synergy analysis revealed that the combina- tion of all observations permits a reduction of the retrieval uncertainty by more than 18% and that the neural network methodology optimally exploits the synergy of observations compared to other approaches. A joint analysis of various remotely sensed datasets of ter- restrial water cycle components demonstrated the coherence of the retrieved soil moisture with other retrieval products and with global hydrological processes. This underlined its potential to be used for observation-based studies of the terrestrial water cycle. Finally, it has been shown that the proposed neural network methodology also provides an effective tool to evaluate Earth System Models on both a variable and a process basis

Thesis (M.S.)--George Mason University, 2008.
Vita: p. 65. Thesis director: John Qu. Submitted in partial fulfillment of the requirements for the degree of Master of Science in Earth System Science. Title from PDF t.p. (viewed Jan. 11, 2009). Includes bibliographical references (p. 59-64). Also issued in print.

Soil moisture is an important element in hydrological land-surface processes as well as land atmosphere interactions and has proven useful in numerous agronomical, climatological and meteorological studies. Since hydrological soil moisture estimates are usually point-based measurements at a specific site and time, spatial and temporal dynamics of soil moisture are difficult to capture. Soil moisture retrieval techniques in remote sensing present possibilities to overcome the abovementioned limitations by continuously providing distributed soil moisture data atdifferent scales and varying temporal resolutions. The main purpose of this study is to derive soil moisture estimates for the Stockholm region by means of two different approaches from a hydrological and a remote sensing point of view and the comparison of both methods. Soil moisture is both modelled with the Topographic Wetness Index (TWI) based on digital elevation data and with the Temperature‐Vegetation Dryness Index (TVDI) as a representation of land surface temperature and Normalized Difference Vegetation Index (NDVI) ratio. Correlations of both index distributions are investigated. Possible index dependencies onvegetation cover and underlying soil types are explored. Field measurements of soil moistureare related to the derived indices. The results indicate that according to a very low Pearson correlation coefficient of 0.023, nolinear dependency between the two indices existed. Index classification in low, medium and high value categories did not result in higher correlations. Neither index distribution is found to berelated to soil types and only the TVDI correlates alongside changes in vegetation cover distribution. In situ measured values correlate better with TVDIs, although neither index is considered to give superior results in the area due to low correlation coefficients. The decision which index to apply is dependent on available data, intent of usage and scale. The TWI surface is considered to be a more suitable soil moisture representation for analyses on smaller scaleswhereas the TVDI should prove more valuable on a larger, regional scale. The lack of correlation between the indices is attributed to the fact that they differ greatly in their underlying theories. However, the synthesis of hydrologic modelling and remote sensing is a promising field of research. The establishment of combined effective models for soil moisture determination over large areas requires more extensive in situ measurements and methods to fully assess the models’ capabilities, limitations and value for hydrological predictions.

Soil moisture is an integral part of the Earth's hydrological cycle. Therefore, accurate estimation of the Earth's changing soil moisture is required to achieve sustainable land and water management to augment Numerical Weather Prediction (NWP) and forecasting skill, and to develop improved flood and drought monitoring capability. Unfortunately, most of the locations in the world do not have accurate soil moisture information on a relevant spatial and temporal scale. However, after latest advances in remote sensing and mesoscale models, it is now possible to estimate soil moisture using passive microwave satellite imaging such as Soil Moisture and Ocean Salinity (SMOS) and/or mesoscale model like Weather Research and Forecasting (WRF)-NOAH Land Surface Model (LSM). L-band passive remote sensing and WRF-NOAH LSM are potentially very useful for soil moisture sensing due to its all-weather capabilities and in-depth physics oriented relationship between soil emissivity and soil moisture, applicable for a diverse land use/land cover. In commensurate with new era in soil moisture remote sensing, this thesis explores the potential of SMOS satellite and WRFNOAH LSM for soil moisture retrieval over the temperate maritime climate. Also, soil moisture deficit (SM I) is found to be an integral component for irrigation scheduling, drought and flood prediction. Hence, the main focus of this thesis is the evaluation of soil moisture datasets as a method to effectively determine the SMD. All major areas of the improvement aided by the SMOS and WRF NOAH LSM arc addressed. Several novel approaches and investigations dealing with the SMOS soil moisture retrieval using Microwave Polarisation Difference Index (M PDI) and Radiative Transfer Equations arc examined. Input data (soil roughness, land surface temperature and vegetation opacity) sensitivity of different retrieval configurations are evaluated using the various algorithms. Thus, the thesis includes ( I) initial evaluation of SMOS satellite and ECMWF downscaled soil moisture using WRF-NOAH LSM with special reference to sensitivity of growing and non growing seasons; (2) assessment of land parameter retrieval model and tau-omega rationale; (3) a modified soil moisture retrieval algorithm from SMOS brightness temperature; and (4) sensitivity and uncertainty analysis of mesoscale model based product for SMD prediction. Further through this study, SMOS soil moisture downscaling schemes using artificial intelligence techniques with MODIS LST have also been proposed to improve the spatial resolution at a catchment scale and finally, data fusion techniques for improving soil l moisture deficit are presented with the SMOS and WRF-NOAII LSM. The overall finding indicates that the SMOS and WRF-NOAH LSM using ECMWF have been proven not only to improve data quality and soil moisture deficit estimation, but also have a great potential in fostering the soil moisture research and applications. The studies presented in this thesis will enhance our understanding of the Earth's water cycle, will help improve ECMWF forecast, SMOS algorithm, NWP and will lead to better water resource management practices.

Soil moisture is a key element in the hydrological cycle, regulating evapotranspiration, precipitation, infiltration and overland flow. However, its effective utilisation in hydrological modelling is still in a state of infancy. Generally, hydrological application of soil moisture data requires: 1) soil moisture data relevant to hydrology, and 2) appropriate hydrological model structure compatible with such data. This thesis focuses on tackling those two aspects by enhancing the compatibility between soil moisture observations and operational hydrological models' soil moisture state variable. In this study satellite soil moisture observations are mainly focused on, because of their increasing availability and large-scale global coverage

Soil moisture is an important variable in hydrological modelling used for real time flood forecasting and water resources management. However, it is a very challenging task to measure soil moisture over a hydrological catchment using conventional in-situ sensors. Remote sensing is gaining popularity due to its large coverage suitable for soil moisture measurement at a catchment scale albeit there are still many knowledge gaps to be filled in. This thesis focuses on investigating soil moisture estimation from remote sensing satellite and land surface model (LSM) coupled with a Numerical Weather Prediction (NWP) model. A hydrological-based approach has been conducted to assess/evaluate the estimated soil moisture using event-based water balance and Probability Distributed Model (PDM). An Advance Microwave Scanning Radiometer (AMSR) and a physically- based Land Parameters Retrieval Model (LPRM) have been used to retrieve surface soil moisture over the sturdy area. The LPRM vegetation and roughness parameters have been empirically calibrated by a new approach proposed in this thesis. The relevant parameters are calibrated on the hydrological model through achieving the best correlation between the observation-based catchment storage and the retrieved surface soil moisture. The development of the land surface model coupled with the NWP model is used to estimate soil moisture at different combinations of soil layers. The optimal combination of the top two layers is found to have the best performance when compared to the catchment water storage. Regression-based mathematical models have been derived to predict the catchment storage from the estimated soil moisture based on both satellite remote sensing and the LSM-NWP model. Three schemes are proposed to examine the behaviour of soil moisture products over different seasons in order to find the appropriate formulas in different scenarios. Finally, weighted coefficients and arithmetic average data fusion methods are explored to integrate two independent soil moisture products from the AMSR-E satellite and the LSM-NWP. It has been found that the merged output is a significant improvement over their individual estimates. The implementation of the fusion technique has provided a new opportunity for information integration from satellite and NWP model. Keywords: Soil moisture, Satellite remote sensing, satellite, land surface model, NWP model, rainfall-runoff model, water balance, PDM model

L'humidité du sol (SM) contrôle les bilans d’eau et d’énergie des surfaces continentales et joue ainsi un rôle clé dans les domaines de la météorologie, l'hydrologie et l'écologie. La communauté scientifique en télédétection micro-ondes a fait des efforts considérables pour établir des bases de données globales de l’humidité du sol en surface (SSM) découlant d'instruments micro-ondes actifs et passifs. Parmi ces instruments, SMOS (Soil Moisture and Ocean Salinity), lancé en 2009, est le premier satellite passif conçu spécifiquement pour mesurer SSM à partir d’observations en bande L (1.4 GHz) à l'échelle globale. La validation des données SMOS SSM sur différentes régions climatiques et pour des conditions environnementales variées est une étape indispensable avant qu’elles soient utilisées de manière opérationnelle. En effet, une meilleure connaissance de la précision des estimations de SSM et des incertitudes associées permettra non seulement d'améliorer les produits SMOS SSM, mais aussi d'optimiser les approches de fusion de données utilisées pour créer des produits multi-capteurs long terme. De tels produits sont développés dans le cadre du programme Climate Change Initiative (CCI) de l'Agence spatiale européenne (ESA) pour l’ensemble des variables climatiques essentielles (ECV), dont SSM. A la suite des chapitres d'introduction I à III, les résultats de cette thèse sont présentés en trois chapitres. Le chapitre IV présente une comparaison des produits SSM issus des capteurs passifs SMOS (bande L) et AMSR-E (bande C) en prenant pour référence les estimations SSM du système d'assimilation SM-DAS-2 du Centre Européen pour les Prévisions Météorologiques à Moyen Terme (CEPMMT). Cette évaluation est menée sur la période d’observation commune à SMOS et AMSR-E (2010- 2011), en utilisant des indicateurs classiques (corrélation, RMSD, Biais). En parallèle, le chapitre V présente une comparaison des produits SMOS SSM avec les produits SSM issus du capteur actif ASCAT en bande C en utilisant comme référence les simulations SSM d’un modèle des surfaces continentales (MERRA-Land), et en utilisant des indicateurs classiques, des méthodes statistiques avancées (triple collocation), et des diagrammes de Hovmöller sur la période 2010-2012. Ces deux évaluations ont montré que la densité de la végétation (paramétrée ici par l’indice foliaire LAI) est un facteur clé pour interpréter la cohérence entre le produit SMOS et les produits AMSR-E et ASCAT. Cet effet de la végétation a été quantifié pour la première fois à l’échelle globale pour les trois capteurs micro-ondes. Ces deux chapitres ont également montré que les trois capteurs SMOS, AMSR-E et ASCAT ont des performances complémentaires selon la densité de végétation et qu’il y a ainsi un potentiel intéressant en terme de fusion des jeux de données micro-ondes passifs et actifs. Dans le chapitre VI, avec l’objectif général d’étendre vers le passé les séries de données SSM de SMOSL3 et de développer un jeu de données SSM homogène sur 2003-2014, nous avons évalué l’utilisation d’une approche de régression linéaire multiple appliquée aux mesures de températures de brillance de AMSR-E (2003 - 2011). Les coefficients de régression ont été calibrés avec les produits SSM issus de SMOS sur 2010-2011. Le produit SSM résultant, qui fusionne les observations SMOS et AMSR-E, a été évalué par comparaison avec un produit SSM AMSR-E et les produits SSM MERRA-Land sur 2007-2009. Ces résultats préliminaires montrent que la méthode de régression linéaire est une approche simple et robuste pour construire un produit SSM réaliste en termes de variations temporelles et de valeurs absolues. En conclusion, cette thèse a montré que le potentiel de synergie entre les systèmes micro-ondes passifs (AMSR-E et SMOS) et actifs (ASCAT) est très prometteur pour le développement et l'amélioration de longues séries temporelles SSM à l'échelle mondiale, telles que celles produites dans le cadre du programme CCI de l'ESA
Soil moisture (SM) plays a key role in meteorology, hydrology, and ecology as it controls the evolution of various hydrological and energy balance processes. The community of scientists involved in the field of microwave remote sensing has made considerable efforts to build accurate estimates of surface SM (SSM), and global SSM datasets derived from active and passive microwave instruments have recently become available. Among them, SMOS (Soil Moisture and Ocean Salinity), launched in 2009, was the first ever passive satellite specifically designed to measure the SSM, at L-band (1.4 GHz), at the global scale. Validation of the SMOS SSM datasets over different climatic regions and environmental conditions is extremely important and a necessary step before they can be used. A better knowledge of the skill and uncertainties of the SSM retrievals will help not only to improve the individual products, but also to optimize the fusion schemes required to create long-term multi-sensor products, like the essential climate variable (ECV) SSM product generated within the European Space Agency’s (ESA's) Climate Change Initiative (CCI) program. After the introductory Chapters I to III, this dissertation consists of three main parts. Chap. IV of the dissertation evaluates the passive SMOS level 3 (SMOSL3) SSM products at L-band against the passive AMSR-E SSM at C-band by comparing them with a Land Data Assimilation System estimates (SM-DAS-2) produced by the European Centre for Medium Range Weather Forecasts (ECMWF). This was achieved over the common period 2010-2011 between SMOS and AMSR-E, using classical metrics (Correlation, RMSD, and Bias). In parallel, Chap. V of the dissertation evaluates the passive SMOSL3 products against the active ASCAT SSM at C-band by comparing them with land surface model simulations (MERRA-Land) using classical metrics, advanced statistical methods (triple collocation), and the Hovmöller diagram over the period 2010-2012. These two evaluations indicated that vegetation density (parameterized here by the leaf area index LAI) is a key factor to interpret the consistency between SMOS and the other remotely sensed products. This effect of the vegetation has been quantified for the first time at the global scale for the three microwave sensors. These two chapters also showed that both SMOS and ASCAT (AMSR-E) had complementary performances and, thus, have a potential for datasets fusion into long-term SSM records. In Chap. VI of the dissertation, with the general purpose to extend back the SMOSL3 SSM time series and to produce an homogeneous SM product over 2003-2014 based on SMOS and AMSR-E, we investigated the use of a multiple linear regression model based on bi-polarization (horizontal and vertical) brightness temperatures (TB) observations obtained from AMSR-E (2003 - 2011). The regression coefficients were calibrated using SMOSL3 SSM as a reference over the 2010-2011 period. The resulting merged SSM dataset was evaluated against an AMSR-E SSM retrievals and modelled SSM products (MERRA-Land) over 2007-2009. These first results show that the multi-linear regression method is a robust and simple approach to produce a realistic SSM product in terms of temporal variation and absolute values. In conclusion, this PhD showed that the potential synergy between the passive (AMSR-E and SMOS) and active (ASCAT) microwave systems at global scale is very promising for the development of improved, long-term SSM time series at global scale, such as those pursued by the ESA’s CCI program. It also provides new ideas on the way to merge the different SSM datasets with the aim of producing the CCI (phase 2) long-term series (a coherent "SMOS-AMSR-E" SSM time series for the period 2003 -2014), that will be evaluated further in the framework of on-going ESA projects

Drought is a natural hazard characterized by a high degree of complexity thus its investigation could be best performed only through a complex analysis involving a set of environmental variables and their intricate relationships. Therefore, the development of an effective early drought detection and monitoring system requires integration of observations on vegetation condition, climate-based drought index data, and several biophysical and social variables of the environment. One of the fundamental drought variables is of course soil moisture, a key parameter determining crop yield potential in drought-affected parts of the world like in the developing nations of the Horn of Africa. Hence soil moisture deficit can be regarded as an important component of, if not synonymous with especially agricultural drought. However, to date model components taken into consideration in the existing drought detection and monitoring systems are only data on precipitation and vegetation condition. Cognizant of this fact, the current work targets integrating historical and real-time monitoring of soil moisture conditions in existing prediction and warning systems to make efforts of early detection of drought events more efficient and reliable. The dissertation addressed the gap through investigation of spatial and temporal soil moisture dynamics in the Horn of Africa using data from microwave observations, and the subsequent definition of monitoring procedures and/or triggers suitable to drought early warning activities. For these purposes, satellite based soil moisture long-term time-series data, obtained from the Water Cycle Multi-Mission Observation Strategy (WACMOS), has been processed, investigated and analyzed using proper statistical methodologies. The specific objectives were investigating the historical time-series soil moisture spatio-temporal dynamics and assess interactions with land cover/vegetation types; to identify historical soil moisture anomalies and establish the relationships with historical agricultural drought events in the Horn of Africa and to explore the relationships between soil moisture and vegetation conditions, a commonly used drought variable. Results of the study has clearly revealed that the WACMOS soil moisture data set and methodology implemented proved useful to identify the behaviors of different vegetation types; the soil moisture index developed (SMCI), is an effective tool in identifying historical drought events; the highly significant correlation between the vegetation and soil moisture data observed at the time lag -1, justifies the potential use of the soil moisture data for drought detection purposes, in order to complement NDVI analysis for an effective drought early warning. Further quantitative validation of the results is possible and helpful if sufficient geospatial data about drought distribution in time and space are available. Eventually, potential of the satellite based WAMOS soil moisture data for detection and monitoring of drought can still be improved if the following points are taken into consideration: Guaranteeing that the soil moisture data retrieved from the different microwave instruments are physically consistent and data available in near-real time, the level of missing data which account for low accuracy need refining and development and the low spatial resolution of the data set is a limitation compromising the level of details many investigations require.

SMOS (Soil Moisture and Ocean Salinity) is ESA's Earth Explorer series satellite carrying the novel MIRAS (Microwave Imaging Radiometer with Aperture Synthesis) interferometric synthetic aperture radar. Its objective is monitoring and studying the planet's water cycle by following the changes in soil moisture levels and ocean surface salt concentrations on a global scale. The success of the mission calls for nearly uninterrupted operation of the science payload. However, the instrument experiences sporadically problems with its hardware, which cause losses of scientific data and may require intervention from ground to resolve. The geographical areas in which most of these anomalies occur, polar regions and the South-Atlantic anomaly, give cause to assume these problems are caused by charged particles in the planet's ionosphere. In this thesis, methods of predicting occurrence of hardware anomalies from indicators of Earth radiation environment are investigated.

Spatial-temporal distribution of soil moisture in the vadose zone is an important aspect of the hydrological cycle that plays a fundamental role in water resources management, including modeling of water flow and mass transport. The vadose zone is a critical transfer and storage compartment, which controls the partitioning of energy and mass linked to surface runoff, evapotranspiration and infiltration. This dissertation focuses on integrating hydraulic characterization methods with remote sensing technologies to estimate the soil moisture distribution by modeling the spatial coverage of soil moisture in the horizontal and vertical dimensions with high temporal resolution. The methodology consists of using satellite images with an ultrafine 3-m resolution to estimate soil surface moisture content that is used as a top boundary condition in the hydrologic model, SWAP, to simulate transport of water in the vadose zone. To demonstrate the methodology, herein developed, a number of model simulations were performed to forecast a range of possible moisture distributions in the Everglades National Park (ENP) vadose zone. Intensive field and laboratory experiments were necessary to prepare an area of interest (AOI) and characterize the soils, and a framework was developed on ArcGIS platform for organizing and processing of data applying a simple sequential data approach, in conjunction with SWAP. An error difference of 3.6% was achieved when comparing radar backscatter coefficient (σ0) to surface Volumetric Water Content (VWC); this result was superior to the 6.1% obtained by Piles during a 2009 NASA SPAM campaign. A registration error (RMSE) of 4% was obtained between model and observations. These results confirmed the potential use of SWAP to simulate transport of water in the vadose zone of the ENP. Future work in the ENP must incorporate the use of preferential flow given the great impact of macropore on water and solute transport through the vadose zone. Among other recommendations, there is a need to develop procedures for measuring the ENP peat shrinkage characteristics due to changes in moisture content in support of the enhanced modeling of soil moisture distribution.

Le satellite altimétrique interférométrique SWOT, dont le lancement est prévu pour 2020, devrait pour la première fois permettre une couverture globale en quelques jours d'un radar proche-nadir en utilisant la bande de fréquence Ka. Or, l'utilisation d'une telle bande de fréquence est encore mal documentée pour l'hydrologie continentale. En particulier, le contraste du coefficient de rétrodiffusion sur les sols et sur les surfaces en eau pour des angles de visée nadir et proche nadir est une problématique majeure. C'est ce qui fait l'objet de ces travaux de thèse. Dans un premier temps, l'étude de la variation des coefficients de rétrodiffusion en provenance d'altimètres en bandes C et Ku et de diffusiomètres utilisant les mêmes bandes de fréquence sur l'Afrique de l'Ouest a montré qu'il existe un lien quantifiable entre l'humidité du sol et le coefficient de rétrodiffusion. En région semi-aride ce lien se manifeste via une hausse des coefficients de rétrodiffusions durant la saison humide par rapport à la saison sèche. L'analyse avec des données annexes d'humidité du sol et de précipitations a pu également montrer que le radar nadir détecte plus précisément les changements d'humidité du sol par rapport à la diffusiométrie radar à visée latérale. Dans le but de mieux comprendre la rétrodiffusion en bande Ka, très peu documentée, deux campagnes de mesures radar ont été réalisées, l'une sur des surfaces en eau à rugosité contrôlée, l'autre sur un terrain contrôlé en rugosité et humidité du sol. En parallèle, un programme de simulation de la rétrodiffusion altimétrique a été développé pour pouvoir analyser les effets d'un faible nombre de variables sur des sols réalistes, dans le but de simuler les variations du coefficient de rétrodiffusion. Ces mesures et ces simulations ont ensuite pu être comparées aux séries temporelles issues du satellite altimétrique AltiKa, fonctionnant en bande Ka et lancé en février 2013, sur différents sites représentatifs des régions bio-climatiques d'Afrique de l'Ouest. Il en ressort que la bande Ka présente une forte sensibilité aux changements d'humidité du sol. Il est également montré que les coefficients de rétrodiffusion en provenance d'AltiKa sur les sols et sur l'eau peuvent être similaires au nadir
The radar altimetry interferometry satellite SWOT, which is to be launched in 2020, should provide for the first time a global coverage of a close-to-nadir radar altimeter in a few days using the Ka-band. Yet, the use of such a frequency band for continental hydrology is still poorly documented. In particular, the contrast of the backscattering coefficient over soils and over water bodies for nadir and close-to-nadir angles is a major issue. This is the reason for this work. First, the study of the backscattering coefficients from C- and Ku- band altimeters and scatterometers over West Africa has shown that there is a link between the surface soil moistureand the backscattering coefficient. In semi-arid regions, this link is seen through a rise of the backscattering coefficients during the rainy season compared to the dry season. The analysis with ancillary data such as the surface soil moisture and the precipitation estimations has also shown that nadir-looking radars detect more precisely the changes in surface soil moisture compared to side-looking radars. Still with the purpose to better understand the Ka-band surface scattering, poorly documented, two measurement campaigns were led, on the one hand over water surface with controlled roughness, on the other hand over bare soils with monitored roughness and surface soil moisture. In parallel, an altimetry backscattering simulation program has been developed to analyze the effect of a low number of variables on realistic grounds, with the aim of simulating the backscattering coefficient variations. These measurements and simulations were then compared with time series from the satellite altimeter AltiKa, which has been launched in 2013 and works at Ka-band, over different sites representaing the bioclimatic areas of West Africa. It led to a high sensitivity of the Ka-band to changes in the surface soil moisture. It has also been shown that backscattering coefficients at nadir-looking angle from AltiKa over grounds and over water bodies can be similar

L’humidité du sol à été déclarée Essential Climate Variable (ECV) en 2010 par l’European Space Agency (ESA) en support du travail du Groupe d'experts Intergouvernemental sur l'Évolution du Climat (GIEC). Dans des zones vulnérables comme l’Afrique de l’Ouest (agriculture faiblement irriguée et de subsistance, températures extrêmes et forte variabilité des précipitations), la valeur ajoutée d’informations concernant l’humidité du sol est importante, surtout dans un contexte de changement climatique. La première partie de ces travaux de thèse concerne la représentation de l'humidité en profondeur à grande échelle en utilisant le triptyque mesures in-situ/télédétection/modélisation. Ces 3 méthodes présentent chacune des limites: (i) la faible densité des réseaux in-situ (3 sites de mesures sur toute l'Afrique de l'Ouest), (ii) les estimations de SMOS uniquement en surface (0-5 cm) et (iii) les incertitudes des forçages de précipitation temps-réel utilisés dans les modèles de surface. Afin de réduire ces limitations, une méthode d'assimilation (filtre particulaire) des données SMOS à été implémentée dans un modèle de surface empirique (API) et comparées aux mesures in-situ AMMA-CATCH. Les résultats montrent une amélioration des humidités modélisée après assimilation. La seconde partie concerne l'impact des variations d'humidité du sol sur les rendements de mil. Une relation statistique a tout d'abord été déterminée à partir de données de rendements mesurés sur 10 villages autour de Niamey. Les résultats montrent que les anomalies d'humidité du sol sur 20 jours début Juillet et fin Août - mi Septembre (période reproductive et période de remplissage du grain), à une profondeur d'environ 30 cm, expliquent les variations de rendement mesuré à R2=0.77 sur l'ensemble de 9 villages. Cette relation à ensuite été appliquée à l'échelle du Niger à partir de données de rendement issues de la FAO et de cartes d'humidité en profondeur développées dans la première partie de la thèse. Les résultats montrent une corrélation à R2=0.62 sur les années 1998-2014. Puis la méthode a été appliquée à 3 autres pays du Sahel, montrant une corrélation de 0.77. La dernière partie de ces travaux concerne l'exploitation des résidus du schéma d'assimilation afin de réduire les incertitudes sur les précipitations. Les produits de précipitations satellites CMORPH, TRMM et PERSIANN, dans leur version temps-réel ont été comparées à des pluviomètres avant et après assimilation. Le résultat de cette étude montre une nette amélioration des intensités estimées. La méthode a ensuite été appliquée à un produit de précipitation utilisé au centre régional AGRHYMET pour le suivi agricole, le produit TAMSAT.Ces travaux de thèse ont permis d'approfondir les recherches concernant le potentiel des données d'humidité par satellite pour des applications agronomiques. Les perspectives de ces travaux portent principalement sur : (i) l'utilisation d'autres capteurs (SMAP, ASCAT, AMSR) pour augmenter la fréquence des observations d'humidité dans l'assimilation, (ii) sur des méthodes de désagrégation des coefficients pour la correction des précipitations à plus haute résolution spatiale et (iii) sur l'utilisation de données multispectrales (indices de végétation, température du sol, ...) pour un meilleur suivi des rendements
Soil moisture was declared Essential Climate Variable (ECV) in 2010 by the European Space Agency (ESA) in support of the work of the Intergovernmental Panel on Climate Change (IPCC). In vulnerable areas such as West Africa (poorly irrigated and subsistence agriculture, extreme temperatures and high variability of rainfall), the added value of informations on soil moisture is important, especially in a changing climate. The first part of this thesis concerns the representation of root-zone soil moisture on a large scale using the triptych in-situ measurements / remote sensing / modeling. These 3 methods each have limitations: (i) the low density of in-situ networks (3 measurement sites throughout West Africa), (ii) SMOS estimates only at the surface (0-5 cm) and (iii) the uncertainties of the real-time precipitation forcing used in surface models. In order to reduce these limitations, an assimilation method (particle filter) of SMOS data has been implemented in an empirical surface model (API) and compared to AMMA-CATCH in-situ measurements. The results show an improvement of the humidities modeled after assimilation. The second part concerns the impact of soil moisture variations on millet yields. A statistical relationship was first determined from yield data measured in 10 villages around Niamey. The results show that the 20-day soil moisture anomalies at the beginning of July and the end of August - mid September (reproductive period and grain filling period), at a depth of about 30 cm, explain the variations in yield measured at R2=0.77. This relationship was then applied to the Nigerien scale from FAO yield data and in-depth moisture maps developed in the first part of the thesis. The results show a correlation at R2=0.62 over the years 1998-2015. Then, the method was apply to 3 other sahelian countries, showing a agreement of 0.77. The last part of this work concerns the exploitation of the residuals of the assimilation scheme in order to reduce the uncertainties on the precipitations. The satellite precipitation products CMORPH, TRMM and PERSIANN, in their real-time version, were compared to rain gauges before and after assimilation. The result of this study shows a marked improvement in the estimated precipitations intensities. The method was then applied to a precipitation product used at the AGRHYMET regional center for agricultural monitoring, the TAMSAT product.This thesis work has led to further research into the potential of satellite moisture data for agronomic applications. The perspectives of this work mainly concern: (i) the use of other sensors (SMAP, ASCAT, AMSR) to increase the frequency of the observations of humidity in the assimilation, (ii) on methods of disaggregation of the coefficients for the correction of precipitation at higher spatial resolution and (iii) the use of multispectral data (vegetation indices, soil temperature, ...) for a better monitoring of yields

Dans les régions de haute biodiversité, caractérisées par des paysages dynamiques, la cartographie détaillée de l'utilisation des sols et du couvert végétal est communément obtenue par la classification d'images satellite. Cependant, les cadres conceptuels d'estimation d'erreurs sur les cartes sont éprouvés pour les zones tempérées et hautement industrialisées. Une nouvelle méthode est proposée pour l'évaluation de la fiabilité des cartes et une autre méthode pour l'estimation des erreurs de classification par ambiguïtés entre classes sur images satellites. La première méthode comprend un nouveau mode d'échantillonnage et une estimation par ensembles flous des incertitudes positionnelles et thématiques. Elle a été testée sur l'Inventaire Forestier Mexicain de l'an 2000. La deuxième méthode s'appuie sur la simulation d'images satellites avec le modèle de transfert radiatif DART et a été testée sur des images IKONOS de six types de forêts au Mexique, sur terrain plat et en forte pente
In highly bio-diverse regions, characterized by dynamic landscapes, detailed land use and land cover maps are commonly generated through the classification of remote sensing imagery. Yet, frameworks for the estimation of errors on maps have focused on mainly temperate zones, in highly industrialized countries. We propose a new method for the accuracy assessment of maps and a new method for the systematic estimation of ambiguities on satellite imagery. The first method comprises a novel sampling scheme and a fuzzy sets based framework whereby degrees of positional and thematic tolerance are user defined. The second method is based on structural and optical measurements and the simulation of satellite imagery with the Discrete Anisotropic Radiative Transfer (DART) model. The first method was successfully tested on the Mexican National Forest Inventory map of year 2000 and the second method was successfully tested on real IKONOS imagery of forest cover types on flat and rugged terrain

nÃo hÃ
A observaÃÃo da superfÃcie terrestre por meio de satÃlites em Ãrbita de nosso planeta tornou-se corriqueira no mundo contemporÃneo. As inferÃncias de variÃveis ambientais diversas feitas a partir de imagens e dados fornecidos por satÃlites cada vez mais aumentam em qualidade e aplicabilidade de maneira que um nÃmero crescente de hidrologistas, meteorologistas, climatologistas e outros profissionais e leigos em geral fazem uso intensivo delas em estudos e pesquisas, em polÃticas governamentais ou na tomada de decisÃo. Uma dessas variÃveis à a umidade superficial do solo, que representa uma importante componente do ciclo hidrolÃgico terrestre, essencial em vÃrios processos naturais ambientais e cujo conhecimento à importante no gerenciamento dos recursos hÃdricos e terrestres, gerenciamento agrÃcola e na modelagem do meio ambiente e agrÃcola. As informaÃÃes derivadas de satÃlites, apesar de ainda apresentarem algumas limitaÃÃes tÃcnicas, podem facilitar bastante o monitoramento ambiental ao se tornarem, muitas vezes, mais Ãgeis e mais econÃmicas do que mediÃÃes locais in situ. Em paÃses em desenvolvimento e de limitados recursos financeiros, tais como o nosso, a informaÃÃo por satÃlites cresce em valor. No Estado do CearÃ, isso desponta ainda mais em virtude das suas dificuldades econÃmicas e sociais. Em vista disso, à proposta, neste trabalho, a aplicaÃÃo nesse estado do Nordeste de um mÃtodo de inferÃncia, por satÃlite, do estado geral da umidade superficial do solo expresso pelo Ãndice de Seca Temperatura-VegetaÃÃo (ISTV), que à indicativo do grau da umidade, estando a ela relacionado. Esse Ãndice à obtido a partir da combinaÃÃo de informaÃÃes da Temperatura da SuperfÃcie Continental (TSC) e do Ãndice de VegetaÃÃo por DiferenÃa Normalizada (IVDN), inferidos por meio de imagens no visÃvel e no infravermelho que podem ser fornecidas por satÃlites meteorolÃgicos operacionais, de Ãrbita polar, tais como os da sÃrie NOAA (National Oceanic and Atmospheric Administration, EUA). No mÃtodo, foi escolhido, da literatura cientÃfica, um algoritmo de cÃlculo da TSC que apresenta certa facilidade de uso, sendo diretamente dependente da FraÃÃo de Cobertura de VegetaÃÃo (FCV) e que pode fornecer boas inferÃncias dessa temperatura. Nesse algoritmo, foram testadas, de forma inÃdita, algumas diferentes formulaÃÃes da FCV encontradas na literatura especializada, representando uma delas o estado da arte no assunto. Foram usadas imagens provenientes do satÃlite NOAA-17, recepcionadas na FUNCEME, e um software especÃfico, dessa FundaÃÃo, para se processar as imagens e implementar a metodologia abordada. Alguns testes foram feitos para duas regiÃes relativamente pequenas do semiÃrido cearense, com destaque para uma delas englobando a Bacia Experimental de Aiuaba (BEA), comparando-se as informaÃÃes do satÃlite NOAA-17 com dados in situ (provenientes de sondas no solo) e com dados advindos dos satÃlites ambientais Terra (dados de TSC, disponÃveis na Internet) e Aqua (dados de umidade superficial do solo). Procurou-se mostrar as diferenÃas qualitativas entre os mapeamentos obtidos, de umidade superficial do solo, e entre estes e os oferecidos pela modelagem em geral. Os resultados encontrados mostraram-se promissores para a utilizaÃÃo no territÃrio cearense do ISTV (no modelo trapezoidal) por meio de satÃlites NOAA, com o algoritmo de Kerr para o cÃlculo da TSC e com a FCV dada pelo Scaled Difference Vegetation Index (SDVI), com o fim de se estimar o estado geral da umidade superficial do solo sobre grandes Ãreas. Entretanto, recomenda-se mais validaÃÃo local posterior do mÃtodo usado, para detecÃÃo de possÃveis erros ou limitaÃÃes nÃo vislumbradas nestes primeiros testes, visando sua definitiva aplicaÃÃo operacional no Cearà e mesmo no semiÃrido do Nordeste.

La télédétection par satellite est aujourd'hui une composante à part entière de l'océanographie. Elle permet d'effectuer des mesures de vents, de température de surface (SST), de couleur de l'eau, de topographie, ... avec des couvertures spatiales et temporelles bien supérieures à celles obtenues par des méthodes in situ. Cependant, il n'existe pas à l'heure actuelle de mesure satellitaire de salinité de surface des océans (SSS), et celle-ci reste sous échantillonnée à la fois spatialement et temporellement. La salinité étant un paramètre important pour la circulation des masses d'eau océaniques, son observation globale et régulière constituerait un apport conséquent à l'océanographie physique. C'est pourquoi de nombreuses équipes scientifiques à travers le monde relèvent actuellement le défi technologique de la télédétection de la SSS par satellite, et particulièrement en Europe grâce à la mission de l'Agence Spatiale Européenne « Soil Moisture and Ocean Salinity » (SMOS). Au cours de ma thèse, j'ai étudié la faisabilité de la mesure de la SSS à l'aide d'un radiomètre hyperfréquence en bande L (i.e. fréquence = 1.4 GHz <=> longueur d'onde = 21 cm), en estimant les sources d'incertitude sur la SSS qui sera restituée dans le cadre de la mission SMOS. Pour cela, j'ai codé un modèle direct, qui simule les processus physiques intervenant depuis la surface océanique jusqu'à l'antenne du radiomètre. Ce modèle est constitué d'un modèle d'émissivité de la mer à « deux échelles » (i.e. on distingue les vagues selon qu'elles soient « grandes » ou « petites » par rapport à la longueur d'onde du radiomètre), et d'un modèle de transfert radiatif à travers l'atmosphère. Le modèle d'émissivité m'a permis d'estimer la sensibilité de la température de brillance (Tb) de l'océan aux paramètres géophysique océanique (i.e. SSS, SST, et rugosité de surface induite par le vent ou la houle), ainsi que l'incertitude sur cette sensibilité en comparant les résultats obtenus à partir de paramétrisations différentes. J'ai conclu de ces études que la sensibilité de la Tb à la SSS est relativement bien connue (de l'ordre de quelques dixièmes de Kelvin par psu) mais que l'effet de la rugosité est très incertain à cause de l'imprécision des modèles de spectre des vagues, alors que cet effet ne semble pas être négligeable (la sensibilité de la Tb au vent étant comprise entre 0.12 à 0.25 K/(m/s) selon le modèle de spectre). Le modèle de transfert radiatif m'a permis d'estimer les différentes contributions de l'atmosphère (atténuation des rayonnements la traversant et émission propre), ainsi que la sensibilité de ces contributions aux paramètres atmosphériques (i.e. profils de température, pression et humidité relative). En bande L, l'atmosphère est quasiment transparente (épaisseur optique ~ 0.01 néper) et sa température de brillance est de l'ordre de 2 K. Ces effets sont peu sensibles aux paramètres atmosphériques, particulièrement à la vapeur d'eau. Je présente aussi dans la thèse des comparaisons du modèle avec des mesures radiométriques en bande L récentes (campagnes WISE 2000, WISE 2001 et EuroSTARRS) ainsi que les conclusions sur la validité des différents modèles de spectre de mer étudiés.

Dans un contexte de développement durable, la gestion des sols et des ressources en eau est un enjeu primordial non seulement d’un point de vue environnemental mais aussi socio-économique. L’humidité, la rugosité, la composition et la structure du sol sont des variables clés pour la compréhension et la modélisation des catastrophes naturelles telles que l’érosion, la sécheresse ou les inondations. Pour des sols nus agricoles (très propices au ruissellement), de nombreuses études ont déjà montré le potentiel des données RADAR acquises en bande C pour la cartographie de l'humidité et la rugosité du sol. Cependant l’application de ces méthodes dans un cadre opérationnel était limitée.Dans ce contexte, les travaux de cette thèse présentent un premier volet sur l’analyse de la sensibilité aux états de surface (EDS) du sol du signal en bande X du capteur TerraSAR-X à très haute résolution spatiale et temporelle. Différentes configurations TerraSAR-X ont été analysées et les résultats ont permis de définir les configurations instrumentales optimales pour caractériser chaque paramètre d’EDS du sol. La comparaison de la sensibilité du capteur TerraSAR-X à celle des capteurs en bande C montre que le capteur TerraSAR-X est sans conteste le plus adapté pour estimer et cartographier l’humidité du sol à des échelles fines (50 m²).Le second volet était de développer une méthode permettant d’estimer et de cartographier l’humidité des sols nus agricoles. Dans ce but, les méthodes d'inversion généralement utilisées en bande C ont été testées sur les données en bande X. La précision sur les estimations d’humidité issues de l'algorithme d’inversion du signal TerraSAR-X a été déterminée et l’applicabilité de la méthode sur de nombreux sites d'étude a été testée avec succès. Une chaine de traitements cartographiques allant de la détection des sols nus à l’estimation de l’humidité et ne nécessitant qu’une seule image TerraSAR-X a été développée. Cette chaine innovante de traitements cartographiques « automatique et autonome » devrait permettre d’utiliser les données TerraSAR-X pour cartographier l’humidité du sol en mode opérationnel
In the context of sustainable development, soil and water resources management is a key issue from not only the environmental point of view, but also from a socioeconomic perspective. Soil moisture, roughness, composition, and slaking crusts are some key variables used to understand and model natural hazards, such as erosion, drought and floods. For agricultural bare soils (most subject to runoff), numerous studies have already shown the potential of C-band RADAR data for the mapping of soil moisture and roughness. However, the application of these methods in operational settings remained limited.In this context, the first objective of this thesis was to analyse the sensitivity of X-band TerraSAR-X sensors to soil surface characteristics (SSC) at high spatial and temporal resolutions. Different TerraSAR-X configurations were evaluated and results were used to define the optimal instrumental configuration for the characterization of each SSC parameter. The comparison of TerraSAR-X sensor sensitivity with equivalent levels recorded with the C-band sensor showed that the TerraSAR-X sensor is undoubtedly the most suitable of the two when estimating and mapping soil moisture at a fine scale (50 m²).The second objective of this work was to develop a method to estimate and map soil moisture levels of agricultural bare soil. To achieve this goal, methods that are commonly used to retrieve soil moisture from C-band, have been tested on X-band data. The accuracy of soil moisture estimations using an empirical algorithm was determined, and validated successfully over numerous study sites. A mapping process based uniquely on TerraSAR-X data, both for bare soil detection and for the estimation of soil moisture content, was developed. This innovative chain of « automatic and autonomous» mapping processing steps should enable the utilization of TerraSAR-X data for the mapping of soil moisture levels in operational conditions

Un regain d'intérêt pour la télédétection de la salinité de surface de l'océan (SSS) par radiométrie en bande-L (21cm) est apparu dans les années 1990 et a conduit au lancement des missions spatiales SMOS (nov. 2009) et Aquarius (juin 2011). Cependant, en raison du faible rapport signal sur bruit, l'inversion de la SSS à partir des mesures radiométriques en bande-L est très difficile. Ce travail porte sur l'étude de la signature radiométrique en bande-L des propriétés de la surface de l'océan (en particulier SSS et rugosité) à partir des mesures du radiomètre aéroporté en bande-L CAROLS, acquises dans le golfe de Gascogne en 2009 et 2010. Une première étude a montré que la SSS déduite des mesures du radiomètre CAROLS était précise à mieux que 0.3 pss dans une zone de forte variabilité spatio-temporelle avec une meilleure précision que les modèles océanographiques côtiers. La seconde étude qui combine les mesures passives (CAROLS) et active (diffusiomètre en bande-C STORM) a mis en évidence l'amélioration des nouveaux modèles de rugosité par rapport aux modèles pré-lancement satellitaires. Par ailleurs, l'étude a montré l'importance de la prise en compte des moyennes et grandes échelles de rugosité (> 20 cm) pour l'interprétation des mesures radiomé- triques loin du nadir.

碩士
國立屏東科技大學
森林系所
105
Soil moisture is an important feature of hydrological circle that affect the tree growth and vegetation distribution. For this reason, estimation of soil moisture dynamics and distribution patterns of forest land with a long period can provide important information for forest management as well as a reference basis. However, the data which from meteorological station in-situ or field surveys can’t represent dynamic of soil moisture and spatial distribution. Taiwan region was selected as the study area. The satellite images of MODIS were collected from 2008, 2010, 2012, 2014, and 2016, each year were divided into Spring (March to May), Summer (June to August), Autumn (September to October), Winter (November to January), so we had 20 periods images in total. Land surface temperature (LST) and Normalized difference vegetation index (NDVI) were used to estimate Temperature vegetation dryness index (TVDI), which represented the soil moisture index of dryness in this study. After the verification of precipitation and terrain on temporal and space, it is feasible that TVDI could represent soil moisture. Soil moisture of ecological zones in Taiwan was estimated by TVDI. The results showed that the spatial patterns of TVDI in the ecological zones of Taiwan were included the tropical dry forest that is the driest area in order to the low vegetation cover and the moistest area is temperature mountain system. The others ecological zone, the orders of soil moister degree were subtropical mountain systems, tropical moist deciduous forest, tropical rain forest and the tropical moisture deciduous forest that located in Eastern Hengchun Peninsula and the lowland of Kaohsiung.

Soil moisture influences the water and energy cycles of terrestrial environments, and thus plays an important climatic role. However, the behavior of soil moisture at large scales, including its impact on atmospheric processes such as precipitation, is not well characterized. Satellite remote sensing allows for indirect observation of large-scale soil moisture, but validation of these data is complicated by the difference in scales between remote sensing footprints and direct ground-based measurements. To address this problem, a method, based on information theory (specifically, mutual information), was developed to determine the useful information content of satellite soil moisture records using precipitation observations. This method was applied to three soil moisture datasets derived from Advanced Microwave Scanning Radiometer for EOS (AMSR-E) measurements over the contiguous U.S., allowing for spatial identification of the algorithm with the least inferred error. Ancillary measures of biomass and topography revealed a strong dependence between algorithm performance and confounding surface properties. Next, statistical causal identification methods (i.e. Granger causality) were used to examine the link between AMSR-E soil moisture and the occurrence of next day precipitation, accounting for long term variability and autocorrelation in precipitation. The probability of precipitation occurrence was modeled using a probit regression framework, and soil moisture was added to the model in order to test for statistical significance and sign. A contrasting pattern of positive feedback in the western U.S. and negative feedback in the east was found, implying a possible amplification of drought and flood conditions in the west and damping in the east. Finally, observations and simulations were used to demonstrate the pitfalls of determining causality between soil moisture and precipitation. It is shown that ignoring long term variability and precipitation autocorrelation can result in artificial positive correlation between soil moisture and precipitation, unless explicitly accounted for in the analysis. In total, this dissertation evaluates large-scale soil moisture measurements, outlines important factors that can cloud the determination of land surface-atmosphere hydrologic feedback, and examines the causal linkage between soil moisture and precipitation at large scales.

碩士
國立中央大學
土木工程研究所
100
Soil moisture is an important factor for the exchange of water between the land surface and plant transpiration. It has tremendous effects on agriculture, the environment and climate. It is hard to evaluate long term land surface dryness by field investigation or ground survey. Using remote sensing technology can get soil moisture information extensively. The Advanced Microwave Scanning Radiometer for EOS (AMSR-E) provide global soil moisture product, the spatial resolution is 25km. The spatial resolution is not good enough to satisfy the demand for agricultural planning or drought monitoring. In the literary, using Moderate Resolution Imaging Spectroradiometer (MODIS) satellite image to observe land surface water content is feasible. A land surface drought index called Normalized Multi-Band Drought Index (NMDI) based on two short wave infrared (SWIR) channel in MODIS as the soil moisture sensitive band, is used for estimating land surface soil moisture, and the spatial resolution is up to 1km. The main objective of this study is to estimate soil moisture conditions of the Central American region using MODIS and AMSR-E data in 2010 and 2011 dry season.

Dottorato di Ricerca in Ingegneria Civile e Industriale, Ciclo XXXI
Il contenuto d’acqua nel suolo svolge un ruolo fondamentale all’interno di numerosi processi che avvengono sulla superficie terrestre, ed in particolare di quelli che fanno parte del ciclo idrologico. In tal senso il contenuto d’acqua nel suolo rappresenta una variabile chiave anche nell’ambito della generazione dei deflussi nei bacini idrografici per effetto degli eventi pluviometrici, e la corretta caratterizzazione della sua evoluzione temporale risulta estremamente funzionale ad una efficace previsione degli eventi di piena. Dato il ruolo di estremo interesse occupato nell’evoluzione dei processi non solo idrologici ma anche ad esempio climatici e agricoli, crescente attenzione è stata dedicata alla modellazione del contenuto d’acqua nel suolo ai diversi fini applicativi, nonché al monitoraggio strumentale della grandezza, che avviene sia in situ, a scala sostanzialmente puntuale con sensori caratterizzati da elevata accuratezza e risoluzione temporale, che da remoto. Con riferimento al secondo caso, il monitoraggio da satellite ha avuto notevoli sviluppi negli ultimi anni, arrivando a fornire informazioni su scala globale che si distinguono per risoluzioni spaziali e temporali sempre più spinte, anche se riferite ai soli primi centimetri di suolo. Queste tre opzioni per la descrizione dell’andamento del contenuto d’acqua nel suolo devono essere viste come complementari, in virtù delle loro diverse peculiarità, nonché delle limitazioni e degli errori che le caratterizzano. In tal senso, un’interessante opportunità è costituita dalle tecniche di data assimilation sviluppate per integrare in maniera ottimale, sulla base delle relative incertezze, le osservazioni con le previsioni da modello. Una potenziale applicazione è l’assimilazione delle osservazioni da satellite all’interno dei modelli afflussi-deflussi, al fine di migliorare le stime delle variabili di stato che rappresentano il contenuto d’acqua nel suolo, e da queste la simulazione delle portate fluviali. Numerosi studi sono stati svolti sul tema, con risultati spesso contrastanti, evidenziando un grande potenziale per questo genere di applicazione, ma anche la necessità di approfondire le numerose scelte procedurali tipicamente richieste in un lavoro di data assimilation. Le tecniche di data assimilation comunemente usate forniscono soluzioni ottime per problemi con precise ipotesi di base (ad esempio l’assenza di errori sistematici), attraverso il confronto fra osservazioni e stime da modello (che devono essere eventualmente ‘mappate’ qualora rappresentino grandezze diverse, ad esempio contenuto d’acqua riferito a diversi volumi/spessori di suolo) basato sulle relative varianze d’errore. Numerose soluzioni sono state proposte per affrontare i vari steps richiesti dal data assimilation, che si sono dimostrati avere un ruolo decisivo sui risultati finali. Le soluzioni proposte riguardano tanto i modelli, ad esempio attraverso una migliorata rappresentazione delle incertezze di stima o con modifiche alla struttura che siano funzionali all’assimilazione delle osservazioni satellitari, che le osservazioni. Le operazioni condotte sulle osservazioni ai fini della successiva integrazione in modelli previsionali hanno costituito il tema principale di questo lavoro. Generalmente, nelle fasi che precedono l’assimilazione delle misure da satellite di contenuto d’acqua nel suolo sono analizzate le seguenti questioni: la verifica della qualità delle osservazioni satellitari, la differenza fra gli spessori di terreno indagato dal sensore e riprodotto nel modello, la correzione delle differenze sistematiche fra i dataset di osservazioni e simulazioni da modello, la caratterizzazione delle varianze degli errori random. Procedure di quality check sono messe a punto per scartare osservazioni ritenute troppo poco attendibili; in tal senso sono fondamentali gli indicatori inclusi nei dataset satellitari, che, fornendo ad esempio informazioni sulle condizioni ambientali durante la misura o feedback dall’algoritmo di stima, consentono una caratterizzazione della qualità del dato. Il setup delle procedure di quality check è funzione ovviamente dell’applicazione finale, tenendo conto degli effetti derivanti tanto dall’utilizzo di un dato poco accurato che dalla sua eliminazione. Un altro aspetto di cui tenere conto riguarda la profondità di suolo in cui è rilevato il dato di contenuto d’acqua da satellite, limitata a pochi centimetri, laddove i volumi di controllo dei modelli sono generalmente maggiori. A tal fine, la struttura di alcuni modelli è stata modificata inserendo uno strato superficiale di spessore ridotto. Una soluzione di uso comune (talvolta anche nel caso di modelli multilayer) è la propagazione dell’informazione superficiale allo spessore di interesse attraverso un filtro esponenziale, che restituisce un indice indicato come SWI (soil water index). La semplicità di questo approccio, basato su un unico parametro, ne ha determinato un’ampia diffusione in vari ambiti applicativi, e dataset globali di contenuto d’acqua da satellite ottenuti con questo metodo sono attualmente in distribuzione. L’eventuale presenza di differenze sistematiche fra il dato da satellite in corso di processamento e la stima da modello deve essere poi corretta, andando ad inficiare in caso contrario le prestazioni del generico sistema di data assimilation, finalizzato alla sola riduzione degli errori random. Diversi approcci sono al riguardo disponibili; quelli di uso predominante risultano indirizzati al matching delle caratteristiche complessive dei due dataset (ad esempio in termini di varianza). Tuttavia, quando la correzione risulta preliminare al data assimilation, pare più appropriato l’uso di tecniche che cerchino di tenere conto della struttura di errore dei due dataset, in modo da effettuare il matching della sola parte informativa (anche nota come segnale), separando quindi i contributi legati all’errore. Sull’osservazione così preprocessata si effettua, quindi, una stima della varianza degli errori random, che contribuirà a determinare il suo peso quando sarà combinata con la previsione ‘a priori’ del modello. Una inadeguata caratterizzazione in questa fase impedisce di giungere al valore di ‘analisi’ ottimale, caratterizzato cioè da varianza di errore minima, e può portare anche al peggioramento delle performance iniziali del modello. Anche per questo step sono stati suggeriti diversi approcci, fra cui quello di uso consolidato è denominato Triple Collocation (TC), e si basa sull’utilizzo di tre dataset indipendenti per i quali si assume la stazionarietà della varianza di errore. Un metodo alternativo, in grado di fornire una stima tempovariabile della grandezza qui indagata, è la propagazione analitica degli errori (EP, error propagation) associati agli input e ai parametri attraverso le equazioni del modello da cui deriva l’osservazione (la misura da satellite del contenuto d’acqua non è in alcun caso diretta ma prevede il processamento delle grandezze effettivamente misurate dai sensori di bordo). Questo secondo approccio tuttavia non garantisce stime in magnitudo plausibili come la TC, non tenendo conto del contributo degli errori dovuti alla struttura del modello. L’analisi delle operazioni di preprocessing e caratterizzazione degli errori delle osservazioni da satellite di contenuto d’acqua nel suolo è stata principalmente svolta attraverso lo sviluppo di due applicazioni. Nella prima applicazione sono trattati i temi del quality check delle osservazioni satellitari e, soprattutto, del trasferimento dell’osservazione superficiale di contenuto d’acqua da satellite a spessori di suolo di maggiore interesse applicativo, usando l’approccio del filtro esponenziale di largo uso in letteratura, in un contesto di verifica della capacità della stima derivata da dati satellitari di riprodurre l’andamento osservato in situ del contenuto d’acqua su strati di spessore maggiore. L’aspetto innovativo introdotto nel lavoro di tesi è costituito dalla messa a punto di uno schema di propagazione degli errori originale, finalizzato alla caratterizzazione per via analitica dell’andamento temporale delle varianze degli errori random del SWI. Le equazioni di propagazione degli errori sono state ricavate e poste in una pratica forma ricorsiva, consentendo di tenere in conto fattori che notoriamente introducono inaccuratezze negli output del filtro esponenziale. Con l’approccio proposto diventa, infatti, possibile propagare le varianze d’errore tempovariabili disponibili in alcuni dataset satellitari di contenuto d’acqua superficiale, nonché valutare gli effetti sul SWI in termini di varianza di errore legati alla disponibilità temporale di misure in input e all’incertezza nel parametro del filtro. Una valutazione preliminare della procedura di propagazione degli errori proposta è stata effettuata verificando l’effettiva corrispondenza fra varianza d’errore del SWI stimata ed effettivi scostamenti rispetto a misure in situ di riferimento; contestualmente sono state anche testate diverse configurazioni della procedura di quality check usando gli indicatori disponibili per il prodotto satellitare usato. Nella seconda applicazione sono trattati i temi della correzione delle differenze sistematiche fra i dataset di osservazioni e simulazioni da modello, e della caratterizzazione delle varianze degli errori random nelle osservazioni, ai fini della valutazione degli effetti dell’assimilazione di misure satellitari di contenuto d’acqua del suolo sulle performance di modelli afflussi-deflussi. Lo studio, svolto durante un soggiorno di ricerca presso il gruppo di Idrologia del CNR-IRPI di Perugia ed in particolare con i ricercatori Luca Brocca e Christian Massari, presenta diversi aspetti innovativi, il primo dei quali è costituito dall’elevato numero (diverse centinaia) di bacini di studio, distribuiti nel continente europeo e complessivamente rappresentativi di diverse condizioni climatiche e fisiografiche, laddove i lavori precedenti su queste tematiche coinvolgevano generalmente aree geografiche ridotte e/o un numero contenuto di bacini. Il dataset di partenza, inclusivo di valori di portata, precipitazione, temperatura e osservazioni satellitari di contenuto d’acqua nel suolo per quasi 900 bacini, è stato costruito dal gruppo di Idrologia del CNR-IRPI. Un secondo aspetto d’interesse riguarda l’aver considerato, oltre ad osservazioni da sensori di tipo sia attivo che passivo provenienti da diverse missioni spaziali, diverse scelte procedurali per le fasi di rimozione delle differenze sistematiche e di caratterizzazione degli errori delle osservazioni. Nel complesso, sebbene le metodologie utilizzate costituiscano delle pratiche riconosciute e usate in questi ambiti, l’utilizzo di procedure comuni per un così largo numero di bacini rappresenta un’applicazione raramente riscontrata in letteratura che ha come principale pregio quello di consentire di superare le soggettività introdotte con la scelta di soluzioni sito-specifiche sovente fatte in precedenti studi su scala più ridotta e talvolta orientate all’ottimizzazione dei risultati finali della procedura di data assimilation. Un terzo tema analizzato, oggetto di attenzione recente nella letteratura del settore, è legato alla presenza di bias di tipo ‘moltiplicativo’ nelle serie temporali di contenuto d’acqua nel suolo da modello e derivate da satellite, ancora presenti in seguito alla fase di rimozione delle differenze sistematiche, e al suo effetto sugli output di portata ottenuti assimilando l’osservazione. Con riferimento all’obiettivo generale del miglioramento della previsione idrologica, in questa applicazione i benefici dell’assimilazione dei dati da satellite sono apparsi variabili, confermando in qualche modo i risultati contrastanti presenti in letteratura. Quale contributo a questo dibattito, lo studio fornisce indicazioni sulla bontà dell’assimilazione di diversi prodotti satellitari in diverse aree geografiche e sotto diverse condizioni preliminari (ad esempio differenti regimi climatici ma anche differenti accuratezze degli input pluviometrici disponibili), e sugli effetti dei diversi approcci metodologici usati per le operazioni preliminari all’assimilazione nel modello. La tesi è strutturata come segue. Il capitolo 1 è costituito da una breve introduzione alle tematiche del lavoro, mentre il capitolo 2 ha per oggetto il contenuto d’acqua del suolo (definizioni, fattori e processi che ne determinano le dinamiche spaziali e temporali, cenni al ruolo nelle varie applicazioni incluse quelle idrologiche) e le caratteristiche dei vari approcci con cui ne viene descritta l’evoluzione (modellazione, misure in situ e da remoto). Il capitolo 3 è incentrato sul data assimilation, fornendo una panoramica dei diversi approcci, una sintesi di risultati ed evidenze relativi all’assimilazione delle misure di contenuto d’acqua nel suolo, e la formulazione matematica dei metodi più comunemente utilizzati per tale scopo. Nel capitolo 4 è fornito un inquadramento teorico su problematiche e metodologie relative alle operazioni di preprocessing e di caratterizzazione degli errori delle osservazioni. Nei capitoli 5 e 6 sono mostrate nel dettaglio le due applicazioni sopra descritte che costituiscono l’aspetto peculiare di questa tesi.
Università della Calabria

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies
Soil moisture is the total amount of water present in the upper 10 cm of soil and it represents the water in land surface which resides in the pores of the soil which is not in river, lakes or groundwater and which depends of the weather conditions, soil type and associated vegetation, among others. Soil moisture assessments are important to understand the hydrological cycles and biophysical processes caused by global climate changes (Finn et al., 2011). Usually, soil moisture has been mapped with airborne microwave radiometers (Klemas et al., 2014) to measure the water retained in the spaces between soil particles. Its importance is due to the microorganism metabolic activity, regulation of the soil temperature and carriage of nutrients, among others. Soil moisture typically takes the form of small ice crystals, vapour, or small parts of liquid water in cold desert soils (Campbell & Claridge, 1982). Antarctic soils are composed by basically no organic and very low moisture content (Campbell and Claridge, 1987). Antarctica is a sensitive area to balance the global climate and its changes and its soil ecosystems are strongly regulated by variables of the abiotic environment and due to this, a research measures the incidence and spatial occurrence of the layer freezing to know how regional climate change could affect the energy exchange of this layer and its invertebrate communities (Wlostowski et al., 2017). Also, knowing how the dynamic of the surface varies in polar regions is transcendent to predict the impact of climate change in global sea-level rise in the future (Quincey & Luckman, 2009).

The European Union's Earth Observation program, Copernicus, aims to create a vast amounts of global data from satellites and ground-based, airborne and seaborne measurement systems with the goal of providing information to help service providers, public authorities, and other international organizations improve European citizens' quality of life. With the aim of reaching this goal, a new family of missions called Sentinels has been developed by the European Spatial Agency, ESA, specifically for the operational needs of the program. These missions carry a range of cutting-edge technologies, such as radars and multi-spectral imaging instruments, for land, ocean and atmospheric monitoring. Multiple kinds of high-resolution data are now available to the scientific community, which is working to adapt and develop the existing models and algorithms to the new information. With this objective, two of the most important variables that contribute to the water cycle, rainfall and river discharge, were selected in this thesis to be estimated by the use of the information obtained from Sentinel-1 and Sentinel-2 sensors. The monitoring of these two variables is fundamental in many hydrological applications, like flood and landslide forecasting and water resources management, and their impact is clearly visible from space. In-situ measurements are the traditional data source of them, but the worldwide declining number of stations, their low spatial density and the data access problem limit their use. Satellite sensors have been therefore adopted to support and, in some cases, substitute the existing gauge network in estimating river discharge and rainfall, thanks to the strong growth in technologies and applications. Two valuable examples of this are SM2RAIN algorithm, which allows to estimate rainfall from Soil Moisture (SM) observations by exploiting the inversion of the soil water balance equation, and the CM approach, a non-linear regression model capable of linking the ground measurements of river discharge to the near-infrared (NIR) reflectance ratio between a dry and a wet pixel chosen around the border of a river. Notwithstanding their usefulness, until now several limitations affected these two methodologies. The main issue with satellite derived rainfall data was their low spatial resolution which could not overcome 10 km, a quantity insufficient to obtain accurate information over many areas and posing important constraint on their use for many applications and fields, which require more and more detailed information. Similarly, the resolution of the available NIR data was not suitable to provide information for narrow rivers (< 250 m wide), nor to study river features and patterns that were here averaged within a single pixel. The recently available high-resolution data from the Sentinel Missions of the Copernicus program offer an opportunity to overcome these issues. The data from Sentinel-1 mission can be used to obtain a high spatial resolution SM product, named S1-RT1, that is adopted in this thesis to derive 1 km spatial resolution (500 m spacing) rainfall data over the Po River basin from it, through the algorithm SM2RAIN. The rainfall derived from the 25 km ASCAT SM product (12.5 km spacing), resampled to the same grid of S1-RT1, is compared to the latter to evaluate the potential benefits of such product. SM2RAIN algorithm needs to be calibrated against a benchmark, which poses important limitations on the applicability of the analysis in data scarce regions. In order to overcome this issue, a parameterized version of SM2RAIN algorithm is previously developed relying on globally distributed data, to be used along with the standard approach in the high-resolution rainfall estimation. The performances of each obtained product are then compared, to assess both the parameterized SM2RAIN capabilities in estimating rainfall and the benefits deriving from Sentinel-1 high spatial resolution. For the river discharge estimation, the use of Sentinel-2 NIR reflectances within the CM approach is investigated to support the hypothesis that a higher satellite product’s spatial resolution, i.e., 10 m (vs. a medium-resolution, i.e., 250 m), is able to better identify the periodically flooded pixels, more related to the river dynamics, with obvious advantages for river discharge estimation. Moreover, the improved resolution allows both a finer distinction between vegetation, soil and water and the characterization of water turbidity in the river area, which is important to correctly estimate the river discharge using this approach. A new formulation enriched by the sediment component is proposed along with a procedure to localize the periodically flooded pixels without the intake of calibration data, which is a first step towards a completely uncalibrated procedure for the river discharge estimation, fundamental for ungauged rivers. The obtained results show that the high-resolution information from Copernicus actually increase the accuracy of the satellite derived products. Good estimates of rainfall are obtainable from Sentinel-1 when considering aggregation time steps greater than 1 day, since to the low temporal resolution of this sensor (from 1.5 to 4 days over Europe) prevents its application to infer daily rainfall. In particular, the rainfall estimates obtained from Sentinel-1 sensors outperform those from ASCAT in specific areas, like in valleys inside mountain regions and most of the plains, confirming the added value of the high spatial resolution information in obtaining spatially detailed rainfall. The use of a parameterized version of SM2RAIN produces performances similar to those obtained with SM2RAIN calibration, attesting the reliability of the parameterized algorithm for rainfall estimation in this area and fostering the possibility to apply SM2RAIN worldwide even without the availability of a rainfall benchmark product. Similarly, the river discharge estimation from Sentinel-2 reflectances from selected stations along two Italian rivers, the Po and the Tiber, confirms that reliable performance can be obtained from high-resolution imagery. Specifically, over both the stations the new formulation improves the river discharge accuracy and over the Po River the best performances are obtained by the uncalibrated procedure. Google Earth Engine (GEE) platform has been employed for the data analysis, allowing to avoid the download of big amounts of data, fostering the reproducibility of the analysis in different locations. Il programma per l’Osservazione della Terra dell’Unione Europea, Copernicus, mira a creare una grande quantità di dati globali da satelliti e sistemi di misurazione terrestri, aerei e marittimi con l'obiettivo di fornire informazioni per assistere i fornitori di servizi, le autorità pubbliche e altre organizzazioni internazionali a migliorare la qualità della vita dei cittadini Europei. Per raggiungere questo obiettivo, l’Agenzia Spaziale Europea (ESA) ha sviluppato una nuova famiglia di missioni satellitari, denominate Sentinel, progettata specificatamente per le esigenze operative del programma. Queste missioni trasportano una gamma di tecnologie all'avanguardia per il monitoraggio terrestre, oceanico e atmosferico, come radar e strumenti di scansione multispettrale. Diversi tipi di dati ad alta risoluzione sono ora disponibili per la comunità scientifica, che sta lavorando per adattare e sviluppare i modelli e gli algoritmi esistenti alle nuove informazioni. Per questa ragione, due delle variabili più importanti del ciclo dell'acqua, la precipitazione e la portata fluviale, sono state selezionate in questa tesi per essere stimate attraverso le informazioni ottenute dai sensori Sentinel-1 e Sentinel-2. Queste variabili sono state scelte perché il loro monitoraggio è fondamentale in molte applicazioni idrologiche, come la previsione di alluvioni e frane e la gestione delle risorse idriche, e inoltre il loro impatto è chiaramente visibile dallo spazio. La fonte tradizionale di questi dati sono le stazioni di misura in situ, ma il decrescente numero dei sensori operativi, la loro ridotta rappresentatività spaziale e i problemi relativi all’accesso e alla condivisione dei dati ne ostacolano l’uso a livello globale. I sensori satellitari sono stati quindi adottati come supporto o, in alcuni casi, alternativa alla rete di misura esistente per la stima delle precipitazioni e delle portate fluviali, grazie alla costante evoluzione delle tecnologie impiegate e della ricerca sulle loro applicazioni. Due validi esempi in tal senso sono l'algoritmo SM2RAIN e l'approccio CM: il primo permette di stimare le precipitazioni in un’area a partire da una serie temporale di umidità del suolo della stessa (ottenibile da satellite), grazie all'inversione dell'equazione di bilancio idrico del suolo; il secondo è un modello di regressione non lineare capace di ottenere una stima della portata fluviale attraverso l’uso di sensori satellitari sensibili alla radiazione del vicino infrarosso (NIR), grazie alle differenze di riflettanza tra l’acqua e il terreno in questa regione dello spettro elettromagnetico. Nonostante la loro utilità, fino ad ora queste due metodologie hanno mostrato diversi limiti. Il problema principale con i dati di precipitazione derivati da satellite è la loro bassa risoluzione spaziale che non eccede i 10 km, una quantità insufficiente per ottenere informazioni accurate in molte regioni del mondo e che pone un importante vincolo al loro utilizzo in diversi campi e tipi di applicazione, richiedenti invece un grado di dettaglio sempre maggiore. Allo stesso modo, la risoluzione dei dati satellitari NIR disponibili non è sufficiente a fornire informazioni per fiumi stretti (< 250 m di larghezza), né a studiare le caratteristiche e i dettagli dei fiumi che vengono invece mediati all'interno di un singolo pixel. La recente disponibilità di dati ad alta risoluzione delle missioni Sentinel del programma Copernicus, però, offre l'opportunità di superare questi problemi. Le informazioni ottenute dalla missione Sentinel-1 possono essere infatti usate per ottenere un prodotto di umidità del suolo ad alta risoluzione spaziale, chiamato S1-RT1, che è stato selezionato in questa tesi per derivarne dati di pioggia a 1 km di risoluzione spaziale (500 m di spaziatura) per il bacino del fiume Po, attraverso l’utilizzo dell’algoritmo SM2RAIN. I benefici potenziali di tale prodotto sono stati valutati attraverso il confronto della pioggia ottenuta con quella derivata dal prodotto satellitare di umidità del suolo ASCAT, caratterizzato da 25 km di risoluzione spaziale (12.5 km di spaziatura) e opportunatamente ricampionato sulla griglia di S1-RT1. L’algoritmo SM2RAIN ha bisogno di essere calibrato attraverso l’uso di un prodotto di riferimento, cosa che pone importanti limiti al suo utilizzo in regioni con scarsa disponibilità di dati osservati. Per superare questo problema, è stata quindi sviluppata una versione parametrizzata dell’algoritmo SM2RAIN indipendente dai dati osservati, da affiancare alla versione standard per la stima della precipitazione ad alta risoluzione. Le performance di ciascun prodotto ottenuto sono state quindi confrontate, in modo da valutare sia le capacità del prodotto parametrizzato di stimare la pioggia senza il supporto di dati di calibrazione, sia i benefici derivanti dall’uso dei dati ad alta risoluzione spaziale di Sentinel-1. Per quanto riguarda la stima della portata fluviale, invece, l’approccio CM è stato applicato alle immagini NIR ottenute dal sensore Sentinel-2 per verificare l’ipotesi che una migliore risoluzione spaziale del prodotto satellitare adottato (10 m di Sentinel-2 contro una risoluzione media di 250 m dei suoi predecessori) sia capace di meglio identificare i pixel periodicamente allagati e quindi sensibili alle variazioni della portata fluviale, con conseguenti benefici alla stima di quest’ultima. La maggiore risoluzione rende inoltre possibile una più accurata distinzione tra vegetazione, suolo e acqua, nonché la caratterizzazione della torbidità dell’acqua nel tratto di fiume selezionato, fattori importanti per stimare correttamente la portata fluviale usando quest’approccio. È stato dunque possibile introdurre una nuova formulazione dell’approccio CM arricchita della componente dei sedimenti, nonché una procedura per la localizzazione dei pixel periodicamente allagati senza l’utilizzo di dati di calibrazione, che rappresenta un primo passo verso una procedura completamente non calibrata per la stima della portata fluviale, fondamentale in fiumi non strumentati. I risultati ottenuti mostrano che le informazioni ad alta risoluzione provenienti dal programma Copernicus possono effettivamente migliorare l’accuratezza dei prodotti di pioggia e portata fluviale ottenuti da satellite. È possibile ottenere stime attendibili di pioggia da Sentinel-1 quando tempi di aggregazione maggiori di un giorno sono presi in considerazione, dato che la ridotta risoluzione temporale del sensore (da 1.5 a 4 giorni per l’Europa) ne previene l’applicazione per l’ottenimento della pioggia a risoluzione giornaliera. In particolare, le stime della pioggia ottenute dai sensori Sentinel-1 hanno prestazioni migliori di quelle di ASCAT in aree specifiche, come le valli all’interno delle regioni montuose e la maggior parte delle pianure, confermando il valore aggiunto dalla elevata risoluzione spaziale nell’ottenimento di un prodotto di pioggia spazialmente dettagliato. L’uso della versione parametrizzata di SM2RAIN mostra prestazioni molto simili a quelle ottenute dalla calibrazione dello stesso, dimostrando l’affidabilità dell’algoritmo parametrizzato per la stima della pioggia nell’area considerata e la possibilità di applicare SM2RAIN in tutto il mondo anche senza la disponibilità di un prodotto di pioggia osservata di riferimento. Similmente, la stima della portata fluviale a partire dalla riflettanza misurata da Sentinel-2 per le stazioni selezionate lungo due fiumi italiani, il Po e il Tevere, conferma che buone prestazioni possono essere ottenute dall’utilizzo di immagini satellitari ad alta risoluzione. Specificatamente, la nuova formulazione permette di migliorare l’accuratezza della portata fluviale stimata in entrambe le stazioni, e, per il fiume Po, le migliori prestazioni sono ottenute dall’uso della procedura non calibrata, provandone la validità. Va infine sottolineato l’impiego della piattaforma Google Earth Engine per l’analisi dei dati di Sentinel-2, che ha permesso di evitare lo scaricamento di ingenti quantità di dati, favorendo la riproducibilità delle analisi anche in diverse località.

Thesis (M.S.)--University of Wisconsin--Madison, 1994.
Typescript. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 42-43).