Accurate crop yield prediction is vital towards optimizing agricultural productivity. Machine Learning (ML) has shown promise in this field; however, its application to legume crops, especially to lupin, remains limited, while many models lack interpretability, hindering real-world adoption. To bridge this literature gap, an interpretable ML framework was developed for predicting lupin yield using Sentinel-2 remote sensing data integrated with georeferenced yield measurements. Data preprocessing involved computing vegetation indices, removing outliers, addressing multicollinearity, normalizing feature scales, and applying data augmentation techniques to correct target imbalance. Subsequently, six ML models were evaluated representing different algorithmic strategies. Among them, XGBoost showed the best performance ($R^2$ = 0.8756) and low error values across $MAE$, $MSE$, and $RMSE$ metrics. To enhance model transparency, SHapley Additive exPlanations (SHAP) values were applied to interpret the feature contributions of the XGBoost model. The Enhanced Vegetation Index ($EVI$) and Normalized Difference Vegetation Index ($NDVI$) were found to be key predictors of crop yield, both showing a positive correlation with higher values reflecting greater vegetation vigor and corresponding to increased yield. These were followed by $B03$ (green) and $B12$ (short-wave infrared), which captured key reflectance properties associated with chlorophyll activity and water content, respectively. Both of them substantially influence photosynthetic efficiency and plant health, ultimately affecting yield potential.

Cultivation-stage carbon accounting is essential for supporting low-emission decision-making. In this context, transparent links between field operations, inputs, and reporting indicators are required. This study quantified gross cultivation-stage greenhouse gas (GHG) emissions from 25 commercial lentil fields in Central Italy. The assessment included the same 14 low-input and 11 intensive fields monitored over two cultivation seasons within a field-to-storage boundary. The observational design compared management bundles, and indicators were calculated at field level and summarized using arithmetic means across fields. A data workflow supported by an Application Programming Interface (API) was used to organize operational records, calculation assumptions, metadata, and emissions outputs. Mean field-level area-normalized emissions were 555 and 632 kg CO₂eq ha⁻¹ for the low-input and intensive systems, respectively. However, because of its higher yield, the intensive system had approximately 22% lower mean product-normalized GHG intensity than the low-input system, namely 0.50 vs. 0.64 kg CO₂eq kg⁻¹ of harvested lentils. Hotspot analysis at the operation level showed that mechanized operations, particularly tillage, harvesting, and seeding, dominated the low-input profile. In contrast, fertilization was the main hotspot in the intensive system. Using producer-provided crude-protein contents, estimated protein yield was higher in the intensive system, while protein-normalized emissions partly overlapped between systems. Overall, interpretation depended on the selected functional unit.