dc.description.abstract
Microwave observations of the Earths surface are sensitive to various environmental variables, including the water content in the soil and vegetation. Since vegetation and soil moisture influence the global carbon-, energy-, and hydrological cycle, their monitoring and mapping is pivotal to provide accurate input in global circulation and climate models. The advantage of microwave remote sensing, compared to well known visible near-infrared (VNIR) remote sensing, is that it is not impeded by cloud cover or dependent on solar illumination. For vegetation monitoring vegetation, optical depth (t) is often used, which is an attenuation parameter in the microwave domain and is related to the water content of the vegetation. So far t has mainly been retrieved from passive microwave observations using radiative transfer models. However, long-term active microwave observations are available from a series of scatterometers, which were originally developed for monitoring ocean winds, making them valuable instruments for monitoring land surface parameters. Although active microwave observations are more sensitive to surface roughness and vegetation geometry, their advantage over passive microwave observations is their better spatial resolution, radiometric accuracy and independence of surface temperature. Consequently, the aim of this thesis is to retrieve t from backscatter observations in order to improve our understanding of vegetation dynamics. Chapter I starts with the motivation for this research. This is followed by a paragraph describing the research questions and objectives and the thesis outline. In Chapter II an introduction to microwave theory is presented. This chapter first deals with the radar equation and scattering properties of natural media. After that follows a literature review focused on previous studies which have assessed the sensitivity of backscatter observations to vegetation dynamics and other land surface parameters. Chapter III presents the retrieval of vegetation optical depth from Metop Advanced Scatterometer (ASCAT) backscatter observations using model parameters of the vegetation correction term within the TU Wien soil moisture retrieval algorithm. A first comparison between vegetation optical depth derived from passive microwave observations (tp) and vegetation optical depth from ASCAT (ta) is performed. Global spatial patterns of ta and tp are qualitatively compared to each other. A temporal comparison is performed by calculating the Spearman Rank Coefficient between climatologies of ta and tp. The strong spatial and temporal correspondence between ta and tp suggest that ta is sensitive to vegetation dynamics in most regions. However, in boreal forests low mean values for ta are found compared to tp. A low temporal correlation is found in deserts and tropical forests, which is attributed to the low natural variability of vegetation in these regions. Furthermore, the retrieval of ta enables the investigation of the effect of the vegetation parameterization in the TU Wien soil moisture retrieval algorithm. Overall, the vegetation parameterization as implemented in the TU Wien algorithm improves the soil moisture retrievals. However, in regions with high inter-annual variability in vegetation dynamics the soil moisture retrieval is degraded, most likely due to the fixed climatology of the correction term. A comprehensive inter-comparison of vegetation products is performed and described in Chapter IV. The inter-comparison is done between ta from Metop ASCAT observations, a cross-ratio (CR) from VH and VV observations from SAC-D Aquarius, tp from AMSR2 observations and Leaf Area Index (LAI) from SPOT VEGETATION. Spatial patterns of the different products are compared and all products follow the expected patterns according to land cover and climate class. Low values for ta are found in high latitude boreal forests and these are attributed to low backscatter values during frozen conditions. It is suggested that these low values in ta are likely to cause a bias in the TU Wien soil moisture product. A temporal comparison between the products shows that the seasonal trajectories of ta are able to follow vegetation dynamics as found in LAI and tp. In deciduous broadleaf forests a disparity is found between the products derived from scatterometers and the other products. This brings to light a different response of scatterometers compared to radiometers, which is possibly caused by leaf fall and the resulting double-bounce scattering. Lastly, phenological parameters, i.e. start of season (SOS) and peak of season (POS), are calculated for all products and compared with the aim to identify differences in timing. Spatial patterns of SOS and POS are tightly coupled between all products, but lags are found between the microwave and VNIR products which vary with land cover and climate. The study confirms the potential of ta to monitor vegetation and phenological parameters. More importantly, it presents a first global comparison between ta and cross-polarized data. The strong coupling between ta and CR suggests that CR may be used in soil moisture retrieval algorithms to improve vegetation parameterization. One of the disadvantages of the ta retrieval was that it is only available as a seasonal product, i.e. 366 values. However, the estimation of the model parameters within the TUWien soil moisture retrieval algorithm has been improved in a way that ta can now be calculated for every day. Chapter V investigates if the time series of ta, and subsequently the TUWien vegetation correction term, are also sensitive to vegetation dynamics by comparing them to LAI. Furthermore, the ability of ta to reproduce inter-annual variability in vegetation dynamics is assessed. Time series of ta are retrieved over Mainland Australia for the period 2007 - 2014. This period contains years with distinct climatic conditions, including the Millenium Drought (2000 - 2009), and two years with large amounts of rainfall as a result of a change in climate modes, due mainly to the El Niño Southern Oscillation. It is found that ta and LAI are tightly coupled, especially over sparsely vegetated regions and grasslands. Over croplands they start to deviate, which is a result of a lag between the two products. In deciduous broadleaf forests negative correlations between the two products are found, as is also found in the previous two chapters, Chapter III and IV. Significant differences between the mean values of drier years and the anomalously wet years are found. Patterns of increased ta correspond to those of LAI and surface soil moisture. Especially in central Australia large changes in ta and LAI are found, where the flush of grasses in a normally barren region effects both products. This thesis developed and validated the retrieval of ta from spaceborne active microwave observations and assesses its ability to monitor vegetation dynamics. It also identifies and analyses differences that arise between VNIR, passive and active microwave remote sensing, especially in boreal and deciduous forests. Overall, the ta satisfactorily follows vegetation patterns and dynamics as observed in VNIR and passive microwave vegetation products. Therefore, missions like Metops European Polar System - Second Generation and Sentinel-1 could retrieve or use ta for vegetation mapping and algorithmic improvements.
en