Francesco Posa

Soil Moisture Retrieval From Remotely Sensed Data: Neural Network Approach Versus Bayesian Method

Neural network (NN) approaches and statistical methods, based on a Bayesian procedure, are applied and compared in soil moisture (SM) retrieval from remotely sensed data. The principles and the practical implementations of Bayesian procedures and NNs are briefly discussed in terms of the advantages and disadvantages of each. Experimental tests are carried out by using the same set of training and test data for each method. The methodologies have been applied to two sets of data to retrieve SM from bare soils and to verify their accuracy. One data set contains scatterometer and radiometer data acquired on a variety of agricultural fields in different polarizations, frequencies, and incidence angles. The other is made up of five experiments carried out with a C-band scatterometer on rough and smooth soils at different polarizations and incidence angles. There are significant similarities in the performance of each method; they both retrieve the same features and trends in the analyzed data sets. Algorithm performances change according to SM level and data configuration. The main difficulties are found in retrieving low SM values, and in this case, the error on estimates is reduced when the data with two polarizations or two incidence angles are inserted in the inversion procedure. One major difference between the methodologies is that the NN performance improves, with respect to the Bayesian method, when more inputs are presented as two polarizations or two incidence angles in the training phase.

Bayesian algorithm for the estimation of the dielectric constant from active and passive remotely sensed data

An inversion technique based on the merging of microwave remotely sensed data is applied to ground-based radiometer and scatterometer data acquired for the same area. The purpose of this technique is to retrieve the dielectric constant of bare soils. The algorithm is based on a Bayesian approach and combines prior information on the dielectric constant and surface roughness with observed data, in order to obtain a marginal posterior probability density function. The function describes how the probability is distributed within the range of the dielectric constant values, given the measured values of emissivity and backscattering coefficient. The algorithm allows for the incorporation of all the available sources of information, such as multipolarization and multifrequency data. Several criteria, which have been used to compare the predicted and the observed values, show that for dielectric constant values higher than 10 the best performance is achieved when data with one polarization and one or two frequencies are exploited. For dielectric constant values of less than 10, the configuration with two polarizations produces the best estimates.

Cassini Radar Data: Estimation of Titan's Lake Features by Means of a Bayesian Inversion Algorithm

The analysis derived from the Cassini SAR imagery reflects the complex Titans surface morphology with a wide range of backscattering coefficients and peculiar features such as periodic structures and lakelike features, which were observed on July 22, 2006, when polar areas were first imaged, and are considered good candidates to be filled with liquid hydrocarbons. In this paper, the modeling description of lakes is addressed by means of a double-layer model which considers an upper liquid-hydrocarbon layer and a lower layer compatible with the radar response of the neighboring areas. This model is introduced into a Bayesian framework for the purpose of inferring the likely ranges of some parameters and, in particular, of the optical thickness of the hypothesized liquid-hydrocarbon layer and of the wind speed. The main idea is to use the information contained in the parameter probability density function, which describes how probability is distributed among the different values of parameters according to the various scenarios considered. The analysis carried out on lakes and surrounding areas on flybys T16 and T19 determines optical thickness values from 0.2 to 6. For T25 flyby, the inferred values of optical thickness indicate that a limit value of optical thickness may be 9. Considering that, beyond these values, the signal from the bottom layer is completely attenuated, information on the wind speed on the upper layer can be inferred. The found mean values of wind speed are around 0.2-0.3 m/s according to different hypotheses on the upper layer dielectric constant.

Cassini Radio Detection and Ranging (RADAR): Earth and Venus observations

The Cassini Radio Detection and Ranging (RADAR) was operated in scatterometric and radiometric modes during the Venus 1 and Earth swingbys to verify its functionality. At Venus, only the thermal emission from the thick absorbing atmosphere was detected. At Earth both the radar echo and the microwave emission from the surface were detected and reveal ocean surface disturbances, the rough, high, and cold Andes mountains, and surface features including a small reservoir in Brazil. Instrument performance appears to be excellent.

Bayesian fusion of active and passive microwave data for estimating bare soil water content

Evaluates an approach to improve the estimation of soil moisture from remotely sensed data. We focus on two types of sensors: a radiometer and a scatterometer operating at a frequency of 4.6 GHz, which both observe the same portion of the Earth surface. Active and passive microwave systems are sensitive to changes in the dielectric properties of the soil and surface morphological properties. Indeed they show complementary capabilities useful for the quantification of these soil parameters. In this context, a retrieval algorithm based on a Bayesian approach has been developed. Our analysis indicates that an improvement in soil moisture estimation accuracy can be obtained when passive radiometric measurements and active radar data are fused with respect to the estimation from a single source. The evaluated soil moisture values show a reasonable agreement in comparison with in situ measurements.

L-band active-passive and L-C-X-bands passive data for soil moisture retrieval, two different approaches in comparison

In the context of the project HYDRO-POL, a study was carried out to test the efficiency of two different approaches: the use of L band active and passive data or the use of L-C-X bands passive data to retrieve soil moisture of bare soils. Simulated data are generated implementing classical superficial scattering models: IEM model for active L-band and L-C-band passive data, GO model for X-band passive data. Data are simulated considering different roughness conditions and moisture content. As the inversion problem is very complex, artificial feedforward backpropagation neural networks (NN) were employed. The best performing NNs are chosen to simulate a retrieval with a dataset artificially added with noise. In each case, the best retrieved parameter is the real part of the dielectric constant, while roughness parameters, especially autocorrelation length, is not very well retrieved. In many cases, retrieved values are out of range, so that the simulated values and targets appear unrelated. Applying a very generic filter that eliminates values very far from the proper range, correlation coefficients grow up. This filter cleans up the resulting data removing a small part of them. After this filtration, correlation coefficients relative to the real part of the dielectric constant surpass 0.82. In spite of the filtering process, roughness parameters retrieval is of inferior quality. On smooth soil, the three considered configurations work in an equivalent way, excellently retrieving the real part of the dielectric constant, without a need for filtration. On medium and rough soil, inversion results generally more difficult, so that performance gets worse. Active-passive approach results more efficient than the L-C-X one

Use of a C-Band Ground-Based Scatterometer to Monitor Surface Roughness and Soil Moisture Changes

Eight experiments on remote sensing of soil moisture and surface roughness were carried out over bare fields with a microwave C-band scatterometer from 1998 until 2001. This device is able to provide backscattering coefficients in the range of +10 dB and −40 dB for incidence angles between 10° and 60°. The objective is to assess the conditions (of roughness and incidence angle) in which it is possible to separate the effects of roughness and soil moisture, thus allowing a reliable estimation of soil moisture from backscattering coefficients. In particular, starting with measurements carried out at different incidence angles, we experimented with an approach to normalize the backscattering coefficients to a reference angle with the principal aim of comparing data sets acquired in different conditions and putting in evidence the radar response to soil moisture variations. A sensitivity analysis, performed over the whole acquired data set, confirmed that the dependence of backscattering coefficients on the incidence angle is influenced more by surface roughness than by soil moisture, as indicated also by theoretical models and other similar data sets. A significant result is that a simple model for comparing data acquired with different incidence angles works better for rough surfaces: only in this case an acceptable correlation between backscattering coefficients and soil moisture is retained. In order to better understand this behavior, the experiments carried out in 2001 were designed to acquire radar measurements on a test site where soil moisture was controlled with artificial irrigation and constantly monitored during the dry-down phase. This allowed a direct estimation of the relationship between radar responses and soil moisture, a quantitative evaluation of the sensitivity of our device and a test for the model developed using the previous acquisitions.

Extraction of soil parameters: two case studies using Bayesian fusion of multiple sources data

This work addresses the possibility of estimating soil moisture values starting from remotely sensed data in the microwave domain. The inversion approach is developed in a Bayesian framework, in order to merge point measurement derived from different sensors. The results indicate that the best combination to obtain reliable estimates of soil moisture is the case when backscattering coefficients and brightness temperature are considered with different polarisations. Furthermore, the introduction of prior information helps the inversion procedure to resolve the unavoidable ambiguities present in such problems.

Retrieval of soil moisture profile by combined C-band scatterometer data and a surface hydrological model

The objective of this work is to develop a method to use radar scatterometer data and a hydrological model in order to retrieve soil behaviour at a level greater than C-band microwave penetration depth. For microwave measurements a C-band FM-CW scatterometer has been employed in two campaigns; the device is able to provide backscattering coefficients in the range of+10 dB and -40 dB for incidence angles between 10° and 60°. Subsequently, microwave scatterometer data have been analysed to estimate their sensitivity to the soil moisture patterns of topsoil comparing them with ground truth measurements. For the validation of these radar data, a coupled heat and moisture balance model has been run to predict the hydrological behaviour of the same topsoil starting from point ground truth measurements. In a second run, soil moisture values derived from scatterometer data should have been used for the initialisation of the model. First attempts have been carried out to propagate the surface physical parameters to unreachable soil layers, such as vertical soil moisture profiles.

Cross-comparison and validation of MODIS AQUA cloud mask by using CLOUDSAT and CALIPSO datasets

This paper presents a cross-comparison of the data acquired by the MODIS, CLOUDSAT and CALIPSO sensors in order to understand the limit of the developed cloud-mask algorithm and to provide a quantitative validation assessment by using exclusively remotely sensed data. The comparison has been carried out by considering both the cloud mask and the intermediate levels such as the brightness temperatures and the reflectance values for different channels from which the cloud mask is derived. The preliminary analysis indicates a general good agreement among the different sources. A main underestimation of cloud cover is present on the sea and especially for high thin clouds. First results indicate that in order to increase the cloud cover accuracy the threshold for the intermediate levels (brightness temperature and reflectance values) may be changed by taking into account also the cloud vertical profiles.