In this study, short-term soil moisture values were estimated in the CONUS region using Cyclone Global Navigation Satellite System (CYGNSS) observations between August 2018 and October 2022 along with ancillary data including precipitation, temperature, normalized difference vegetation index (NDVI), land cover classification (LCC), elevation, and soil type. The study employs Long Short-Term Memory (LSTM) neural networks to demonstrate the effectiveness of this method in accurately estimating daily soil moisture at International Soil Moisture Network (ISMN) stations. Furthermore, the study analyzes SHapley Additive exPlanations (SHAP) values to understand how each feature contributes to the model’s predictions.

Crop yield estimation at the national scale is on the rise with a clear increase in freely available satellite images; providing field-level crop masks and their corresponding Earth Observation (EO) data-based predictors. In line with increasing EO data, the yield estimation problem is moved from a univariate to a multivariate time series analysis problem. Indeed, most of the state of the art methods applied to yield estimation are enabled to increase the accuracy of the yield estimation model using these multisource EO data. However, one of the major drawbacks of these methods is the lack of explainability of the yield. In this study, we will try to understand what EO-based time series data says about the agricultural practices differences among the geographically distributed fields and which kinds of dissimilarities exist among these multi-source EO data, and what regional/global factors cause these dissimilarities. In order to understand this, the study will go through the shape based and feature based time series similarity metrics that often highlight different characteristics of the time series data. While the temperature’s parameters (i.e., the incident solar radiation and 2 m dewpoint temperature), not the temperature itself, highlight geographical variation in yield data and behave as a geotag, these climate variables are not the only cause, but are the contributing ones, driving the yield variation distribution patterns of nationwide data.

Earth Observation (EO)-based crop yield estimation, which focuses on leveraging crop conditions at any time t has recently played a critical role in the development of nationwide crop monitoring. Following the developments in open data policy in remote sensing, the high-resolution freely available EO data provides annual crop masks and enables the understanding of more complex patterns of the agricultural practices at field level. Despite the detailed information provided by EO-imaging data at the field level, the resolution of the target variable, commune-level yield, restricts the effective usage of EO imaging data and, in turn, imposes limitations on leveraging the large amount of EO data, which could provide accurate yield estimation by a physics aware data-driven estimation models. In this paper, we explore the challenges of the uncertainties associated with data-driven yield estimation using the Turkey cotton dataset, which comprises EO-based time series from Sentinel-1 and Sentinel-2 imaging satellites, together with climate variables and soil properties. These uncertainties can arise from discrepancies in the EO-based crop mask data, an unreliable statistical dataset, and the variance in the spatio-temporal characteristics of descriptive features and low-resolution yield statistics that are used in the estimation.

In this study, we examined the effectiveness of integrating satellite-based crop biophysical parameters, meteorological conditions, and soil properties for the end and mid-season cotton yield prediction in the continental United States (CONUS) region. We employed six machine learning algorithms: decision tree (DT), random forest (RF), adaptive boosting (Ad-aBoost), gradient boosting (GB), light gradient boosting machine (LightGBM), and extreme gradient boosting machine (XGBoost). By employing this rigorous approach to hyperparameter tuning based on Bayesian optimization, the XGBoost method was found as the best method for both mid-season and end-season cotton yield prediction. Furthermore, we investigated the global importance of temporal and static features using the Shapley Additive Global importancE (SAGE) method to understand the driving factors of cotton yield prediction. As a result of global feature importance analysis, precipitation (P), enhanced vegetation index (EVI), and leaf area index (LAI) were found as the most important temporal features, while silt and pH were found as the most important soil properties.

Cotton, a vital crop in the global textile industry, faces challenges from climate and ecosystem changes. Accurate cotton yield prediction is crucial for the economy and environmental sustainability, and it requires a deep understanding of the complex relationship between its parameters and yield. To achieve this, a comprehensive approach integrating climatic factors, soil parameters, and biophysical parameters observed through high-resolution remote sensing satellites was employed. This study utilized a multisource dataset to develop a predictive model for cotton yield over Turkiye, allowing accurate yield estimation and understanding the impact of the Earth Observation (EO)-based yield predictors on the model. Specifically, we utilized the Explainable Boosting Machine (EBM) algorithm to model and predict cotton yield while offering insights into selecting EO predictors. Additionally, we conducted a performance evaluation of our proposed approach in comparison to popular boosting-based algorithms like eXtreme gradient boosting (XGBoost), adaptive boosting (AdaBoost), and Light gradient boosting (Light-GBM).

This study aimed to assess the influence of Earth observation (EO) time series data, specifically soil properties, climate variables, and Enhanced Vegetation Index, on predicting cotton yield using an explainable artificial intelligence model. By utilizing statistical yield data acquired at the commune level in Turkey between 2019-2021, we developed a model for predicting cotton yield. The model employed the Long Short-Term Memory (LSTM) architecture and incorporated the SHapley Additive exPlanations (SHAP) method as a post-hoc method to explain how EO features impact the cotton yield and to interpret the relationship between these features and the variations in yield data.

Cotton is under the threat of climate and ecosystem change, and has an essential role in the global textile industry. This makes its yield prediction essential for both economics and sustainability. The potential cotton yield can be predicted by integrating climatic factors, soil parameters, and biophysical parameters observed by high temporal & spatial resolution remote sensing satellites. This study used a multisource dataset to create an explainable and accurate predictive model for cotton yield prediction over the continental US (CONUS). A recently proposed glass-box method called Explainable Boosting Machine (EBM), which provides transparency, reliability, and ease of interpretation, was implemented. Accuracy performance was compared with common machine learning (ML) methods for predicting cotton yields. The EBM showed higher accuracy against other glass-box methods and competitive results with black-box models. With the help of the EBM, the importance of individual features and their pairwise interactions was revealed without applying any post-hoc methods. The study findings showed that the precipitation (P), enhanced vegetation index (EVI), and leaf area index (LAI) are the three most important dynamic features. The dynamic features are the driver of the created model with 78% of the overall feature importance, followed by pairwise interactions of the features with 16% contribution. Lastly, static features contribute 6% to the overall feature importance. The study highlights the importance of using multi-source data and interactions of the input features and providing an interpretable model to understand the inner dynamics of cotton yield predictions.

Predicting biophysical parameters with high accuracy and fast speed based on remote sensing-based modeling is an attractive topic. In this context, the revisit time, coverage, and illumination condition in-dependency make the Polarimetric Synthetic Aperture Radar (PoISAR) data is an attractive tool. In this paper, one of the most studied biophysical parameters, Leaf Area Index (LAI), is chosen to assess Polynomial Chaos Expansion (PCE) regression, commonly used metamodeling due to its precise and rapid approximation performance. Experimental analysis based on AgriSAR 2009 campaign, including oat and canola, is given to validate the PCE in the regression. According to the accuracy analysis, the Pearson correlation of 88% and 95% for oat and canola, respectively, were achieved.

Crop phenology monitoring using Synthetic Aperture Radar (SAR) data is gaining popularity within the remote sensing community due to SAR's all weather and large coverage imaging capability. This paper introduces a polynomial chaos expansion (PCE) based regression algorithm to retrieve BBCH scale of crops, which identifies the phenology of crops in a standardized system. The impact and applicability of the proposed methodology is successfully illustrated using the TerraSAR-X dual-Pol imagery that was acquired over the cultivation period of paddy-rice fields located in Turkey. To assess the applicability of the methodology, root mean square and correlation analysis were performed under different amount of training data and number of inputs.

Space-borne interferometric SAR has advanced significantly in the last decades, with many successful Earth monitoring applications. The key point of this success lies in the fact that interferometric SAR supplies unprecedented phase and amplitude information characterizing target's physical parameters. This paper presents the role of interferometric SAR, specifically considering the bistatic mission TanDEM-X, for shallow lake water level estimation by using sediment storage.