Drought is a major environmental stressor limiting crop productivity, with dryland cotton systems particularly vulnerable due to their dependence on precipitation. This study presents an integrated framework combining remote sensing, machine learning, and phenological analysis to map dryland cotton and predict yield gain and loss across drought scenarios from 2000 to 2024. To address gaps in the USDA Cropland Data Layer (CDL), a cotton classification model was developed using Landsat imagery and a Random Forest algorithm, guided by the cotton growing calendar. The model achieved high performance, with an overall accuracy and F1-score of 0.91, an MCC of 0.82, and R2 values between 0.67 and 0.95 for annual area comparisons. Monthly evapotranspiration (ET) data from OpenET were used to differentiate dryland from irrigated cotton by comparing ET against crop water requirements and historical precipitation. Yield data from 184 dryland observations (2008-2018) were analysed using six drought severity indices. The Palmer Drought Severity Index (PDSI) was identified as the most sensitive indicator of cotton yield variability. A phenological framework was then applied using binary PDSI values (positive = wet, negative = drought) to model yield gain and loss. In addition, three ensemble machine learning models were applied across five scenarios. The highest predictive accuracy (86 %) and MCC (0.72) were achieved during the early phenological stages (0-30 and 30-60 days after green-up), underscoring that consecutive drought stress across these periods, particularly through flowering, greatly intensifies the risk of cotton yield reduction. To further validate these findings, raw observational datasets were analysed to identify vulnerable growth phases contributing to yield loss. The results confirmed that the early flowering phenological stage is particularly susceptible to drought stress, underscoring its critical role in determining cotton production outcomes. Therefore, the approach from this study offers innovative contributions for improving water resource management and guiding producer decisions in drought-prone cotton producing regions.