Luca Pasolli

Gaussian Process Regression for Estimating Chlorophyll Concentration in Subsurface Waters From Remote Sensing Data

In this letter, we explore the effectiveness of a novel regression method in the context of the estimation of biophysical parameters from remotely sensed imagery as an alternative to state-of-the-art regression methods like those based on artificial neural networks and support vector machines. This method, called Gaussian process (GP) regression, formulates the learning of the regressor within a Bayesian framework, where the regression model is derived by assuming the model variables follow a Gaussian prior distribution encoding the prior knowledge about the output function. One of its interesting properties, which gives it a key advantage over state-of-the-art regression methods, is the possibility to tune the free parameters of the model in an automatic way. Experiments were focused on the problem of estimating chlorophyll concentration in subsurface waters. The achieved results suggest that the GP regression method is very promising from both viewpoints of estimation accuracy and free parameter tuning. Moreover, it handles particularly well the problem of limited availability of training samples, typically encountered in biophysical parameter estimation applications.

Polarimetric RADARSAT-2 imagery for soil moisture retrieval in alpine areas

In this work, the polarimetric capability of RADARSAT-2 images is exploited in the aim of soil moisture content retrieval in Alpine meadows and pastures. Three feature extraction methods are investigated: the simple polarimetric intensity and phase processing, the H/A/α polarimetric decomposition, and the Independent Component Analysis (ICA). The features extracted according to these strategies were assessed for their capability to improve the soil moisture estimation by considering both quantitative performance on a set of reference samples and qualitative analysis of the corresponding output soil moisture content maps. The method proposed for the soil moisture estimation was based on the Support Vector Regression technique combined with an innovative multi-objective model selection strategy. The results indicated that the use of polarimetric features such as HH and HV channels improved the estimation of soil moisture content in the investigated mountain area, especially because the HV channel was able to disentangle the vegetation effect on the radar signal. From the preliminary results presented in this paper, the use of the H/A/α polarimetric decomposition and the ICA technique seem to not determine a significant improvement in the soil moisture estimation.

Estimating Soil Moisture With the Support Vector Regression Technique

This letter presents an experimental analysis of the application of the ε-insensitive support vector regression (SVR) technique to soil moisture content estimation from remotely sensed data at field/basin scale. SVR has attractive properties, such as ease of use, good intrinsic generalization capability, and robustness to noise in the training data, which make it a valid candidate as an alternative to more traditional neural-network-based techniques usually adopted in soil moisture content estimation. Its effectiveness in this application is assessed by using field measurements and considering various combinations of the input features (i.e., different active and/or passive microwave measurements acquired using various sensor frequencies, polarizations, and acquisition geometries). The performance of the SVR method (in terms of estimation accuracy, generalization capability, computational complexity, and ease of use) is compared with that achieved using a multilayer perceptron neural network, which is considered as a benchmark in the addressed application. This analysis provides useful indications for building soil moisture estimation processors for upcoming satellites or near-real-time applications.

Estimation of Soil Moisture in an Alpine Catchment with RADARSAT2 Images

Soil moisture retrieval is one of the most challenging problems in the context of biophysical parameter estimation from remotely sensed data. Typically, microwave signals are used thanks to their sensitivity to variations in the water content of soil. However, especially in the Alps, the presence of vegetation and the heterogeneity of topography may significantly affect the microwave signal, thus increasing the complexity of the retrieval. In this paper, the effectiveness of RADARSAT2 SAR images for the estimation of soil moisture in an alpine catchment is investigated. We first carry out a sensitivity analysis of the SAR signal to the moisture content of soil and other target properties (e.g., topography and vegetation). Then we propose a technique for estimating soil moisture based on the Support Vector Regression algorithm and the integration of ancillary data. Preliminary results are discussed both in terms of accuracy over point measurements and effectiveness in handling spatially distributed data.

Comparison between SAR Soil Moisture Estimates and Hydrological Model Simulations over the Scrivia Test Site

In this paper, the results of a comparison between the soil moisture content (SMC) estimated from C-band SAR, the SMC simulated by a hydrological model, and the SMC measured on ground are presented. The study was carried out in an agricultural test site located in North-west Italy, in the Scrivia river basin. The hydrological model used for the simulations consists of a one-layer soil water balance model, which was found to be able to partially reproduce the soil moisture variability, retaining at the same time simplicity and effectiveness in describing the topsoil. SMC estimates were derived from the application of a retrieval algorithm, based on an Artificial Neural Network approach, to a time series of ENVISAT/ASAR images acquired over the Scrivia test site. The core of the algorithm was represented by a set of ANNs able to deal with the different SAR configurations in terms of polarizations and available ancillary data. In case of crop covered soils, the effect of vegetation was accounted for using NDVI information, or, if available, for the cross-polarized channel. The algorithm results showed some ability in retrieving SMC with RMSE generally <0.04 m 3 /m 3 and very low bias (i.e., <0.01 m 3 /m 3 ), except for the case of VV polarized SAR images: in this case, the obtained RMSE was somewhat higher than 0.04 m 3 /m 3 (≤0.058 m 3 /m 3 ). The algorithm was implemented within the framework of an ESA project concerning the development of an operative algorithm

Multi-Objective Parameter Optimization in Support Vector Regression: General Formulation and Application to the Retrieval of Soil Moisture From Remote Sensing Data

This paper deals with the tuning of the free parameters of the Support Vector Regression technique used for the retrieval of geo/bio-physical variables from remotely sensed data. We propose to address this task in the framework of the multi-objective optimization. A multi-objective function is defined based on a set of two (or more) metrics (e.g., mean squared error MSE and determination coefficient R2 ) that quantify from different (and sometimes competing) perspectives the goodness of a given parameter configuration. Then the metrics are jointly optimized according to the concept of Pareto optimality. This allows preserving the meaning of each metric and deriving multiple optimal solutions to the tuning problem. Each solution leads to a different optimal trade-off among the considered metrics. The main advantages of the proposed multi-objective parameter optimization approach with respect to traditional mono-objective strategies are: (1) the intrinsic improved robustness and efficiency, since multiple metrics are jointly exploited in the tuning of the free parameters of the considered regression method; and (2) the possibility to select the parameter configuration that leads to the desired trade-off among different criteria and thus best meets both the application constraints and the requirements of the specific estimation problem. The experimental analysis was focused on the challenging application domain of soil moisture retrieval from microwave remotely sensed data. The results obtained on data sets associated with two different operative conditions are very promising and show the effectiveness of the proposed approach in comparison with more traditional tuning strategies based on a single metric and its usefulness in defining estimation systems for real application domains.

Estimation of Soil Moisture in Mountain Areas Using SVR Technique Applied to Multiscale Active Radar Images at C-Band

This paper presents an approach for retrieval of soil moisture content (SMC) from different satellite sensors with a focus on mountain areas. The novelties of the paper are: the extension of an already developed method to coarse resolution data (150 m) in mountain environment with high land heterogeneity, with only VV polarization and the proper selection of input features. During the result analysis, several algorithm characteristics were clearly identified: 1) the performances showed to be strongly related to input features such as topography and vegetation indices; 2) the algorithm needs a training phase; 3) the averaging window needs to be proper selected to take into account both the speckle noise and the characteristics of the area under investigation; and 4) the algorithm, being data driven, can be considered as site dependent. The experimental analysis is carried out on images acquired over the Südtirol/Alto Adige Province in Italy during 2010-2011 from the RADARSAT2 and Envisat ASAR in Wide Swath mode. SMC maps were compared with spatially distributed ground measurements, resulting in a root mean squared error (RMSE) value ranging from 0.045 to 0.07 m3/m3. Concerning the multiscale analysis, the results indicated that RADARSAT2 maps are able to detect the spatial heterogeneity and soil moisture dynamics at local scale, while ASAR WS SMC maps are able to identify mainly the two main classes of pasture and meadows. When these estimates are compared with SMC values from meteorological stations a RMSE value of 0.10 m3/m3 for both satellites indicated a reduced capability to follow the temporal dynamics.

Estimating Biophysical Parameters from Remotely Sensed Imagery with Gaussian Processes

Recently, a new machine learning approach that is based on the Gaussian process (GP) theory has been introduced in the literature. According to this approach, the learning of a machine (regressor or classifier) is formulated in terms of a Bayesian estimation problem, where the parameters of the machine are assumed to be random variables which follow jointly a Gaussian distribution. The purpose of this work is to investigate this approach in the context of the estimation of biophysical parameters. Experimental results obtained on synthetic and real data, which simulate the spectral behavior of the chlorophyll concentration in subsurface waters, are reported and compared with those yielded by the general regression neural network (GRNN) and the epsiv-insensitive support vector regression (SVR) methods.

Performance inter-comparison of soil moisture retrieval models for the MetOp-A ASCAT instrument

In this study we evaluate five different retrieval algorithms, applied on MetOp-A ASCAT backscatter data, in their ability to retrieve soil moisture on a global scale. Correlation and triple collocation analysis are performed using in situ and land surface model data as a reference. Results do not clearly identify one best algorithm. We therefore conclude that future work should focus on the exploitation of the strengths and weaknesses of different modelling approaches in a synergetic way rather than trying to find one model that suits every possible situation.

Multiobjective model selection for non-linear regression techniques

This paper proposes to model the critical issue of the choice of the free parameters of a supervised non-linear regression technique (the so called model selection issue) as a multiobjective optimization problem. In this framework, the multi-objective function is made up of a set of two or more quality metrics (e.g., MSE, R2, etc.) computed on the test (or validation) samples. A set of solutions is derived according to the concept of Pareto optimality. The advantages of the proposed approach with respect to the traditional ones (which typically optimize a single scalar metric) are mainly two: 1) the capability to derive solutions which jointly optimize the set of metrics considered and represent different possible optimal tradeoffs among them; and 2) the possibility for the user to effectively select the model that optimizes the requirements of the specific retrieval problem. Results achieved for the specific application of soil moisture estimation from microwave remotely sensed data with the Support Vector Regression (SVR) technique are reported. These results show the effectiveness of the proposed approach.