The objective of this Ph.D. thesis is to build the scientific foundation for reliable irrigation quantification. During the last decade, irrigation quantification has evolved from being mainly model-based to include a wide variety of satellite-based methods. Satellites enable us to observe how humans, through irrigation, influence the soil-vegetation system, which is challenging to model with traditional physically based models. Therefore, this thesis aims to investigate the uncertainty and robustness of satellite approaches, provide novel methodologies for irrigation quantification, and tools for global upscaling by metamodeling.

The uncertainty of the satellite approaches depends on the uncertainty of satellite data. Although we like to think of satellite products of evapotranspiration and soil moisture as observations, they are merely estimates inferred from remote sensing models and can be significantly different. Despite this, a limited focus has been on assessing the irrigation uncertainty associated with satellite uncertainty, which this study aims to address by employing an ensemble approach to quantify the precision of an evapotranspiration-based approach. Many satellite approaches focus either on evapotranspiration or soil moisture-based approaches to quantify irrigation or use them in a joint approach. Less attention has been given to comparing the approaches and the main frameworks, which is essential to support a paradigm shift in irrigation monitoring. This study provides the first inter-comparison of systematic residuals between satellite estimates and rainfed land surface model baselines, created from an evapotranspiration-based, a soil moisture-based, and a joint approach, and compare irrigation estimates from different framework structures to provide insights into the strengths and limitations needed to further advance irrigation quantification methodologies. Lastly, the study aims to support sustainable water management by providing tools for easy access to global scale irrigation estimates, explored through machine learning-based metamodels. Metamodels allow quick extrapolation of irrigation knowledge in space and time beyond the capability of the model and satellite approaches. The following main conclusions can be drawn:
• Satellite uncertainty is an issue within both evapotranspiration and soil moisture-based approaches that must be addressed. For evapotranspiration-based approaches, we found that although the rainfed evapotranspiration significantly differed between the satellite ensemble members, the irrigation-induced evapotranspiration was more similar. This suggested that the precision of evapotranspiration-based irrigation estimates is high if the framework allows for appropriate identification and subtraction of a potentially biased rainfed baseline.• The inter-comparison of different approaches and frameworks underlined the advantage of using both evapotranspiration and soil moisture in a joint approach to create more robust irrigation estimates. Further, this study showed that the framework structure was important to harvest the potential synergy from a joint approach through the calibration strategy.• This study has mainly been focussed on the advantage of combining LSMs and satellite data to quantify irrigation. The main advantage of using this methodology is the strength of the baseline model to compensate for evapotranspiration biases and how it allows for mitigating precipitation uncertainty through model calibration. This resulted in more robust irrigation estimates closely linked to climate patterns.• Metamodels demonstrated their ability to predict irrigation closely linked to hydroclimatic conditions and remote sensing data. The spatiotemporal model transferability was found to improve as the spatial extent of the training data increased, thus encompassing a more diverse range of environments.This Ph.D. thesis provides knowledge on how satellite uncertainty may propagate into irrigation estimates, insight into the strengths and limitations of main approaches and frameworks, and how to synthesize our knowledge of irrigation. All is needed to build robust irrigation quantification methodologies and create fundamental improvements within agricultural monitoring and groundwater management. These results could provide feedback into the hydrological modeling community and improve irrigation schemes. Also of great interest within the climate modeling community to better represent and parametrize irrigation in coupled Earth system models, which would improve weather forecasts and climate predictions• The inter-comparison of different approaches and frameworks underlined the advantage of using both evapotranspiration and soil moisture in a joint approach to create more robust irrigation estimates. Further, this study showed that the framework structure was important to harvest the potential synergy from a joint approach through the calibration strategy.• The inter-comparison of different approaches and frameworks underlined the advantage of using both evapotranspiration and soil moisture in a joint approach to create more robust irrigation estimates. Further, this study showed that the framework structure was important to harvest the potential synergy from a joint approach through the calibration strategy.• Satellite uncertainty is an issue within both evapotranspiration and soil moisture-based approaches that must be addressed. For evapotranspiration-based approaches, we found that although the rainfed evapotranspiration significantly differed between the satellite ensemble members, the irrigation-induced evapotranspiration was more similar. This suggested that the precision of evapotranspiration-based irrigation estimates is high if the framework allows for appropriate identification and subtraction of a potentially biased rainfed baseline.• Satellite uncertainty is an issue within both evapotranspiration and soil moisture-based approaches that must be addressed. For evapotranspiration-based approaches, we found that although the rainfed evapotranspiration significantly differed between the satellite ensemble members, the irrigation-induced evapotranspiration was more similar. This suggested that the precision of evapotranspiration-based irrigation estimates is high if the framework allows for appropriate identification and subtraction of a potentially biased rainfed baseline.• Satellite uncertainty is an issue within both evapotranspiration and soil moisture-based approaches that must be addressed. For evapotranspiration-based approaches, we found that although the rainfed evapotranspiration significantly differed between the satellite ensemble members, the irrigation-induced evapotranspiration was more similar. This suggested that the precision of evapotranspiration-based irrigation estimates is high if the framework allows for appropriate identification and subtraction of a potentially biased rainfed baseline.• Satellite uncertainty is an issue within both evapotranspiration and soil moisture-based approaches that must be addressed. For evapotranspiration-based approaches, we found that although the rainfed evapotranspiration significantly differed between the satellite ensemble members, the irrigation-induced evapotranspiration was more similar. This suggested that the precision of evapotranspiration-based irrigation estimates is high if the framework allows for appropriate identification and subtraction of a potentially biased rainfed baseline.