Innovative Use of Earth Observation and Land Surface Modeling for Tracking the Effects of Irrigation on the Terrestrial Water Cycle

In recent years, human water needs have been steadily increasing and they are currently dominated by agricultural activities for food production worldwide. Furthermore, the increase in population and climatic change are expected to raise the current demand of water highlighting the necessity for more efficient irrigation systems. In this context, the combined effect of human pressure (i.e. irrigation) and the increase of extreme natural phenomena, such as drought events, has a strong impact on the global water budget with a local depletion of water resources, especially groundwater. However, our understanding of the impact of human activities on the water cycle is challenged by the lack of data (such as irrigation benchmark data) and by the difficulty of land surface models (LSM) to represent human processes like irrigation. These knowledge gaps affect water and food security, because they undermine both the ability to accurately monitor and forecast drought events, and the capacity to safely manage water resources.

This thesis aims at shedding light on the utility of new earth observation (EO) data to characterize agricultural drought conditions and to detect water cycle modifications induced by anthropogenic activities like irrigation in order to help models to improve their ability to more realistically represent the terrestrial water cycle. Based on that, this research work tries to address two main important research questions:

What is the added value of new EO data (i.e., satellite-based soil moisture) in drought monitoring and to which extent are these data able to provide potential information on crop production with respect to LSMs?

Can the new generation of high resolution EO data help LSMs to better represent the impact of human activities like irrigation on the terrestrial water cycle?

To answer question (1) the activity focuses on the use of an innovative long-term record of satellite-based soil moisture for the development of a standardized agricultural drought index for a regional scale analysis. The novelty of this research is to establish the relation between drought indices and crop yields, through a comparison with a benchmark crop dataset, as well as to analyze the additional information contained in satellite observations about agricultural productivity and water uses as compared to ground-based rainfall and modeled soil moisture. The main findings highlight the crucial role of soil moisture in limiting the crop productivity during drought periods and consequently its key contribution for agricultural drought analysis. Another important aspect highlighted in the results is that satellite estimates of soil moisture contain added information about both water scarcity conditions and anthropogenic impacts on water resources (i.e., irrigation), compared to soil moisture estimates based on model simulations (which do not account for human-related processes). Those outcomes strongly link the agricultural drought analysis with the second part of this research.

To answer question (2), the activity explicitly focuses on the quantification of the water used for agricultural purposes. Hydrological studies are converging on the synergistic use of models and satellite data to detect and quantify irrigation. The parameterization of irrigation in large-scale LSMs is improving but not sufficient by themselves to provide correct irrigation estimates, because they are still hampered by the lack of information about dynamic crop rotations, and by out-of-date maps of irrigated areas as well as unknown timing and amount of irrigation. On the other hand, satellite observations are directly affected by, and hence potentially able to detect, irrigation. Therefore, combining LSMs and satellite information through data assimilation (DA) can offer the optimal solution to quantify water used for irrigation at the desired spatio-temporal scale. This research aims at building an innovative and reliable DA system able to investigate the potential of high-resolution microwave EO data from the Sentinel-1 mission to improve irrigation quantification. The main assumption is that the joint update of soil moisture and vegetation model states, through the ingestion of 1-km radar backscatter, can improve irrigation estimation. This is a topic which was not investigated in previous literature. In this context, the optimization of a coupled system comprising a LSM and a backscatter observation operator, is firstly investigated. Results highlight the importance of equipping the LSM with an irrigation scheme to avoid strong biases between satellite observations and backscatter model predictions over intensively irrigated areas, where anthropogenic activities cannot be neglected. Secondly, backscatter observations are assimilated in the calibrated DA system, taking into account the effects of different backscatter polarizations. DA introduces both improvements and degradations in soil moisture, vegetation and irrigation estimates. The spatial and temporal scale of the results have a large impact on the analysis and more contradicting results are found for an analysis at the plot scale, which highlights the need for very high-spatial resolution EO data and model parameterizations. Above all, this study sheds light on the limitations resulting from poorly-parameterized irrigation schemes included in LSMs which prevents large improvements in the irrigation simulation due to DA and points out on future implementations and input developments needed to improve LSM estimates of hydrological variables.

The research activities focus on regions characterized by a significant human pressure on the terrestrial water cycle. The first part includes the Karnataka and Maharashtra states, located in central India, whereas the second part includes irrigated areas with a different climate in Europe, i.e. Po Valley, in northern Italy, and Niedersachsen in Germany.