A Dataset for Pasture Parameter Estimation based on Satellite Remote Sensing and Weather Variables

During 12 months in two paddocks with Brachiaria brizantha cv, Marandu forage, forage collection was carried out with gardening shears. The data were collected from April 6, 2022, until March 1, 2023, in a paddock with animals (4 Nelore breed animals) and another paddock without animals. Two collection types were performed: every 15 days in a square of 1m² randomly released in both paddocks and containment cages of 1 cubic meter, collected every 30 days, and placed in the paddock with animals. In each sample, coordinates (latitude and longitude) were collected using GPS.

The laboratory chemical analysis was performed on all samples to estimate forage parameters: Crude Protein (CP), Acid Detergent Fiber (ADF), Neutral Detergent Fiber (NDF), Dry Matter content (DM), Biomass Content, Mineral Metter (MM) and, Total Digestible Nutrients (TDN).

For each GPS coordinate, satellite multispectral and weather data were acquired. The multispectral data were collected using Google Earth Engine API, based on Sentinel-2 multispectral images. Eleven Sentinel-2 bands (B01, B02, B03, B04, B05, B06, B07, 8A, B09, B11, and B12) were acquired, and eight well-known spectral indices (NDVI, NDWI, EVI, LAI, DVI, GCI, GEMI, and SAVI) were calculated and integrated into the dataset.

Weather data were acquired in two ways: using two APIs (Open Wheater MAP and Open-Meteo) and data from an existing meteorological station at the sample collection site. The weather data acquired based on APIs are Maximum and Minimum Temperature Acquired during the day (TEMP_MAX, TEMP_MIN), Average of Solar Radiation during the day (RAD_SOL), Average average registered during the day (RAIN), Average wind speed registered during the day (WIND_SPD), Average evapotranspiration estimated of the soil during the day (EVAPOT), Average Atmospheric Pressure registered during the day (PREST_ATM) and, Average Relative humidity registered during the day (HUM_REL).

Finally, data related to the day of collection (date and Day of Year - DOY), sample coordinates (latitude and longitude), and sample type (ID that identifies the type of sample):
Paddock with animals:
- Q1 - Q4: square 1 - 4,
- G1 - G4: cage 1 - 4,
Paddock without animals:
- S1 or S2: square 1 or 2

Were integrated into the complete dataset.

In the Folder called "data" there are four files: “Table 1 - Field_Experiment_Data.csv”, “Table 2 - Multispectral_data.csv”, "Table 3 - Weather_Data.csv" and “Complete_DataSet.csv”. The first file, contains just pasture chemical analysis values; in the second, hyperspectral data; on the third file are the weather data acquired on APIs and the meterological station, and in the last, all the data has been integrated.

In the "src" folder, the Python script to acquire hyperspectral and weather data from APIs is called “Search_Images_and_Weather_Data.ipynb”. The file "weatherapi.py" receives data to estimate weather information from both APIs.

本数据集采集周期为12个月，实验场地为两块种植臂形草属马兰多品种（Brachiaria brizantha cv. Marandu）牧草的试验田，采用园艺剪开展牧草样本采集工作。采样时段为2022年4月6日至2023年3月1日，试验田分为两组：一组放牧有4头内洛尔（Nelore）肉牛，另一组无放牧牲畜。

本次实验采用两种采样方案：其一为每15天在两块试验田中随机布设1m²样方进行采集；其二为布设1立方米截留采样笼并放置于有牲畜的试验田内，每30天完成一次采样。所有样本均通过GPS采集经纬度坐标。

对所有采集的样本开展实验室化学分析，以估算牧草核心参数：粗蛋白质（Crude Protein, CP）、酸性洗涤纤维（Acid Detergent Fiber, ADF）、中性洗涤纤维（Neutral Detergent Fiber, NDF）、干物质含量（Dry Matter content, DM）、生物量含量、矿物质含量（Mineral Matter, MM）及总可消化养分（Total Digestible Nutrients, TDN）。

针对每个GPS坐标点，同步获取卫星多光谱数据与气象数据。多光谱数据基于Sentinel-2多光谱影像，通过谷歌地球引擎（Google Earth Engine, GEE）API采集，共获取11个Sentinel-2波段（B01、B02、B03、B04、B05、B06、B07、8A、B09、B11及B12），并计算并整合8种经典光谱指数：归一化植被指数（Normalized Difference Vegetation Index, NDVI）、归一化水指数（Normalized Difference Water Index, NDWI）、增强型植被指数（Enhanced Vegetation Index, EVI）、叶面积指数（Leaf Area Index, LAI）、差值植被指数（Difference Vegetation Index, DVI）、绿色叶绿素指数（Green Chlorophyll Index, GCI）、全球环境监测指数（Global Environmental Monitoring Index, GEMI）及土壤调节植被指数（Soil Adjusted Vegetation Index, SAVI）。

气象数据通过两种途径获取：一是调用OpenWeatherMap与Open-Meteo两个API接口；二是采集采样点所在区域现有气象站的观测数据。通过API获取的气象参数包括：日间最高气温（TEMP_MAX）、日间最低气温（TEMP_MIN）、日间平均太阳辐射（RAD_SOL）、日间平均降雨量（RAIN）、日间平均风速（WIND_SPD）、日间土壤日均蒸散量（EVAPOT）、日间平均大气压强（PREST_ATM）及日间平均相对湿度（HUM_REL）。

最终，完整数据集还整合了以下元数据：采样相关信息（采集日期及年中日（Day of Year, DOY））、样本坐标（经纬度）及样本类型标识ID：
- 有牲畜试验田样本：Q1~Q4代表1~4号1m²样方样本，G1~G4代表1~4号截留采样笼样本；
- 无牲畜试验田样本：S1或S2代表1号或2号1m²样方样本。

在名为"data"的文件夹中包含4个数据文件："Table 1 - Field_Experiment_Data.csv"、"Table 2 - Multispectral_data.csv"、"Table 3 - Weather_Data.csv"及"Complete_DataSet.csv"。其中第一个文件仅包含牧草化学分析数据；第二个文件包含多光谱数据；第三个文件包含通过API及气象站获取的气象数据；最后一个文件整合了全部数据集。

在名为"src"的文件夹中，用于从API获取多光谱及气象数据的Python脚本为"Search_Images_and_Weather_Data.ipynb"；文件"weatherapi.py"用于获取并估算两个API提供的气象信息。