Evaluation of the crop sequence boundary (CSB) dataset for field boundary mapping and spatial overlap analysis supporting pesticide risk assessment

• CSB datasets validated using field data is California. • CSB shows moderate to high classification accuracy for major crops in California. • CSB often splits fields into multiple polygons, especially for non-perennial crops. • CSB is better option for overlap analysis in pesticide risk assessment. Accurate delineation of field boundaries is essential for translating pixel-level remote sensing data into actionable field-scale agricultural decision-making. This study evaluates the accuracy of the United States Department of Agriculture (USDA) Crop Sequence Boundaries (CSB) dataset, which derives crop field polygons from eight years of Cropland Data Layer (CDL) data by analyzing crop rotation history and spatial contiguity. Focusing on California, with a detailed case study in Kern County, we assess accuracy of the CSB dataset against high-quality, independent references from the Kern County Department of Agriculture and Measurement Standards (KAFB) and the California Department of Water Resources (DWR). Evaluation metrics include total and crop-specific acreage, number of fields, alignment of field boundaries, and F1 scores for crop type classification. Results indicate that the CSB dataset underestimates total cultivated area by 6.3 % relative to KAFB and 5.4 % relative to DWR. Crop-specific comparisons using KAFB as a reference show that the CSB dataset performs reliably for orchard crops—accurately capturing pistachios (0 % difference), slightly overestimating almonds (+5.2 %), and modestly underestimating citrus (−7.5 %)—while exhibiting larger overestimates for grapes (+17 %), cotton (+10.7 %), and alfalfa (+31.8 %), and significantly underrepresenting “other” crops (−31.7 %). The CSB also overestimates the number of fields, reporting more than twice as many as KAFB (28,022 vs. 11,914), due to field subdivisions driven by within-field variability in the gridded satellite data. F1 scores demonstrate medium to high crop classification accuracy, with values ranging from 73 % for alfalfa to 88 % for citrus and 86 % for almonds, indicating strong performance for perennial crops. Despite segmentation limitations, the CSB dataset offers substantial value through its provision of temporally consistent, field-scale crop type data with medium to high classification accuracy in a user-friendly format. To illustrate its practical use, we applied the CSB dataset in an overlap analysis with selected listed species ranges, an essential step in in pesticide risk assessments under the Endangered Species Act (ESA). This overlap analysis, which evaluates the potential likelihood of pesticide exposure, demonstrated that CSB significantly improves accuracy of overlaps between agricultural lands and endangered species habitats by providing more accurate, year-specific data compared to current regulatory approaches that use the Use Data Layer (UDL) as inputs. While both UDL and CSB are derived from the same underlying CDL dataset, —the annual resolution of CSB offers greater benefits for pesticide risk assessment. Overall, the CSB dataset demonstrates strong potential as the best available data source for practical applications such as overlap analysis for pesticide risk assessment, although continued refinement is needed to enhance its accuracy in field boundary detection.