Abstract
Estimating pasture parameters is essential for decision-making in the management of livestock and agriculture. Despite that, the time-consuming acquisition of outdoor forage samples and the high cost of laboratory analysis make it infeasible to predict parameters of quality and quantity forage recurrently and with great accuracy. Previous work has shown that multispectral and weather data have correlation with forage parameters, enabling the design of supervised machine learning models to predict forage conditions. Nevertheless, datasets with pasture yield and nutritional parameters, remote sensing and weather information are scarce and rarely available, limiting the design of prediction models. This paper presents a dataset with more than 300 samples of pasture laboratory analyses collected over nearly twelve months from two paddocks. Latitude and longitude coordinates were collected for each sample using GPS coordinates, and this data helped acquire multispectral band signals and eight vegetation index values extracted from Google Earth Engine (Sentinel-2 satellite) for each pixel of each sample. Furthermore, the dataset has weather data from APIs and a meteorological station. These data can also motivate new studies that aim determine pasture behaviour, joining this dataset with larger datasets that have similar information.