CATCH-GH: Effort-Offset Gradient Boosting with Water-Body-Conditional Conformal Prediction for Pre-Trip Catch Forecasting in Ghana’s Volta Basin Artisanal Fisheries
Description
Trip-level dataset of artisanal fishing in the Volta Region of Ghana, January 2019 to December 2024 (2024 partially observed). It contains 9,248 fishing trips by 160 fishers (anonymous codes) at 20 landing sites on four water bodies: Volta Lake, Lower Volta River, Keta Lagoon and Avu Lagoon. Each row is one trip, with 50 variables: date fields (year, month, day of year, season), location (landing site, water body, town, district), effort (crew size, trip length, days fished, gear units), catch in kg (total, sold, consumed), economics in Ghana cedis (revenue, costs of fuel, ice, food and other items, profit), derived ratios such as catch per unit effort (CPUE), and fisher-history fields computed from each fisher's previous trips. The history fields are missing for each fisher's first trip (160 missing values per field); all other fields are complete. Median catch is 31.2 kg per trip and mean CPUE is 11.2 kg per fisher-day. The data were used to develop and test the CATCH-GH catch forecasting model. Costs exclude equipment depreciation and crew pay, and the records have not been verified against official statistics.