Average estimated yields and associated CV values for current (2018) model runs. Based on work done by Harsimran Kaur et al in 2017. The following is from her thesis: Agro-ecological classes (AECs) of dryland cropping systems in the inland Pacific Northwest have been predicted to become more dynamic with greater use of annual fallow under projected climate change. At the same time, initiatives are being taken by growers either to intensify or diversify their cropping systems using oilseed and grain legume crops. The main objective of this study was to use a mechanistic model (CropSyst) to provide yield and soil water forecasts at regional scales which could compare fallow versus spring crop choices (flex/opportunity crop). Model simulations were based on historic weather data (1981-2010) as well as combined with actual year weather data for simulations at pre-planting dates starting in Dec. for representative years. Yield forecasts of spring pea, canola and wheat were compared to yield simulations using only weather of the representative year via linear regression analysis to assess pre-plant forecasts. Crop yield projections on pre-plant forecast date of Feb 1st had higher R2 with yield simulated using actual years weather data and lower CVs across the region as compared to forecasts based on historic weather data and other pre-season forecast dates (Dec. 1st and Jan. 1st). Therefore, Feb. 1st was considered the most reliable time to predict yield and other relevant outputs such as available water forecasts on a regional scale. Regional forecast maps of predicted spring crop yields and CVs showed ranges of 1 to 4367 kg/ha and 11 to 293% for spring canola, 72 to 2646 kg/ha and 11 to 143% for spring pea and 39 to 5330 kg/ha and 11 to 158% for spring wheat across study region for a representative year. These data combined with predicted available water after fallow and following spring crop yield as well as estimates of winter wheat yield reduction would collectively serve as information contributing to decisions related to crop intensification and diversification. Resources in this dataset:Resource Title: GeoData catalog record. File Name: Web Page, url: https://geodata.nal.usda.gov/geonetwork/srv/eng/catalog.search#/metadata/459d2dba-a346-4e54-9750-ef3178c18f38

Analysis of ‘Crop Yield Prediction Dataset’ provided by Analyst-2 (analyst-2.ai), based on source dataset retrieved from https://www.kaggle.com/patelris/crop-yield-prediction-dataset on 28 January 2022.

--- Dataset description provided by original source is as follows ---

Context

The science of training machines to learn and produce models for future predictions is widely used, and not for nothing. Agriculture plays a critical role in the global economy. With the continuing expansion of the human population understanding worldwide crop yield is central to addressing food security challenges and reducing the impacts of climate change.

Crop yield prediction is an important agricultural problem. The Agricultural yield primarily depends on weather conditions (rain, temperature, etc), pesticides and accurate information about history of crop yield is an important thing for making decisions related to agricultural risk management and future predictions.

Acknowledgements

All dataset(publicly available dataset) here are taken form FAO (Food and Agriculture Organization) and World Data Bank. http://www.fao.org/home/en/ https://data.worldbank.org/

--- Original source retains full ownership of the source dataset ---

This dataset includes sample data for the Netherlands to run the machine learning baseline as described in the paper titled Machine learning for large-scale crop yield forecasting, accessible at https://doi.org/10.1016/j.agsy.2020.103016. The software implementation of the machine learning baseline is available at: https://github.com/BigDataWUR/MLforCropYieldForecasting.

Notes:

The NUTS classification (Nomenclature of territorial units for statistics) is a hierarchical system for dividing up the economic territory of the EU and the UK (see Eurostat, 2016) for more details).

Data

The dataset consists of 11 CSV files. They are formatted to work as sample inputs to the machine learning baseline.

1. Crop Area Fractions (NUTS2, NUTS1): We aggregated the predictions of the machine learning baseline from NUTS2 to national (NUTS0) level by weighting them on the modeled crop area. Cerrani and López Lozano (2017) have described in detail the algorithm used to model crop areas for different NUTS levels. The data comes from the MARS Crop Yield Forecasting System (MCYFS) of European Commission's Joint Research Centre (JRC) (see Lecerf et al., 2019).
2. Centroids (NUTS2): Data includes latitude, longitude and distance to coast of the centroids of NUTS2 regions.
3. Meteo Daily Data and Meteo Dekadal Data (NUTS2): The data comes from MCYFS (see EC-JRC, 2020). By default, the implementation uses daily data.
4. Remote Sensing Data (NUTS2, see Copernicus Global Land Service, 2020): Data includes fraction of absorbed photosynthetically active radiation (FAPAR) aggregated to NUTS2.
5. Soil Data: Data includes soil moisture information that can be used to calculate soil water holding capacity. The data comes from MCYFS (see Lecerf et al., 2019).
6. WOFOST data (NUTS2): The World Food Studies (WOFOST) crop model (van Diepen et al., 1989; Supit et al., 1994; de Wit et al. 2019) is a simulation model for the quantitative analysis of the growth and production of annual field crops. It is a mechanistic, dynamic model that explains daily crop growth on the basis of the underlying processes, such as photosynthesis, respiration and how these processes are influenced by environmental conditions. The crop simulation is fed by weather, soil and crop data. Observed meteorological data is interpolated on a regular 25 km grid using a method based on the distance, altitude and climatic region similarity between the center of grid cells and weather stations (see Van der Goot, 1998). WOFOST runs on the intersection between the 25 km meteorological grid and soil units based on the European soil map (http://esdac.jrc.ec.europa.eu/). In order to have the output data aggregated to administrative regions such as countries or provinces, simulation units are further intersected with the boundaries of these regions. The outputs at soil unit (STU) level are aggregated to grid level in an area weighted manner. Gridded simulations are aggregated to lowest NUTS level 3 considering the arable land area of each grid, derived from GLOBCOVER and CORINE Land Cover (Cerrani and Lopez Lozano, 2017). From NUTS3 to higher levels, crop area fractions for the current year, retrieved from Eurostat, are used to weight and aggregate the output (Cerrani and Lopez Lozano, 2017).
7. National yield statistics (NUTS0): These are the official Eurostat national yield statistics (Eurostat, 2020a). We used these yield statistics as reference to compare the machine learning predictions aggregated to NUTS0 and the actual MCYFS forecasts (see van der Velde and Nisini, 2019).
8. Regional yield statistics (NUTS2): We used NUTS2 yield statistics as labels to train and evaluate machine learning algorithms. We got NUTS2 yield statistics from The Central Bureau of Statistics (CBS) of the Netherlands (NL-CBS, 2020).
9. Past MCYFS Yield Forecasts (NUTS0): These are actual forecasts made by MCYFS in the past (see van der Velde and Nisini, 2019). We used the official Eurostat national yield statistics (see point 7 above) as the reference to compare the machine learning predictions aggregated to NUTS0 and MCYFS forecasts.

Crop ID and name mapping

2 : grain maize

6 : sugar beets

7 : potatoes

90 : soft wheat

93 : sunflower

95 : spring barley

Acknowledgements

We would like to thank S. Niemeyer from the European Commission’s Joint Research Centre (JRC) for the permission to provide open access to the Netherlands data. Similarly, we would like to thank M. van der Velde, L. Nisini and I. Cerrani from JRC for sharing with us past MCYFS forecasts and Eurostat national yield statistics.

Drone-based Agricultural Dataset for Crop Yield Estimation

This repository contains a comprehensive dataset of cashew, cocoa and coffee images captured by drones, accompanied by meticulously annotated labels. To facilitate object detection, each image is paired with a corresponding text file in YOLO format. The YOLO format file contains annotations, including class labels and bounding box coordinates. The dataset was collected by teams from Ghana (KaraAgro AI) and Uganda (Makerere… See the full description on the dataset page: https://huggingface.co/datasets/KaraAgroAI/Drone-based-Agricultural-Dataset-for-Crop-Yield-Estimation.

We used historical data for crop yield in 27 Indian states and 3 Union Territories of India, covering the years 1997 to 2020. The dataset consists of 19,689 data points, each with ten features including Crop, Season, Crop_Year, State, Annual_Rainfall, Area, Production, Pesticide, Fertilizer, and Yield. The dataset encompasses 55 different types of crops cultivated across India. The crop yield dataset was used to prediction of crop yield using regression with stacking ensemble model. The dataset is split into training 80% and testing 20%.

This dataset contains detailed information on crop yields across various states in India for the year 1997. It includes data on different crops, their production, area under cultivation, season of cultivation, and state-specific information. Additionally, the dataset provides supplementary details such as annual rainfall, fertilizer use, pesticide use, and yield for each crop. This comprehensive dataset can be used for agricultural analysis, trend prediction, and studying the impact of various factors on crop yields in India.

The global crop yield forecasting market is experiencing robust growth, driven by increasing demand for efficient agricultural practices and the need to mitigate risks associated with climate change and resource scarcity. The market, valued at approximately $2.5 billion in 2025, is projected to exhibit a Compound Annual Growth Rate (CAGR) of 15% from 2025 to 2033. This substantial growth is fueled by several key factors. Technological advancements in remote sensing, artificial intelligence (AI), and machine learning (ML) are enabling the development of increasingly accurate and sophisticated forecasting models. The integration of these technologies into both software and service offerings is revolutionizing agricultural planning, enabling farmers to optimize resource allocation, improve yields, and reduce waste. Furthermore, government initiatives promoting sustainable agriculture and precision farming are providing further impetus to market expansion. The rising adoption of precision agriculture techniques, coupled with the growing awareness among farmers about the benefits of data-driven decision-making, is further contributing to the market's upward trajectory. The market segmentation reveals a strong preference for crop yield forecasting software, owing to its scalability and potential for integration with existing farm management systems. The commercial sector dominates the application segment, reflecting the higher adoption rate among large-scale agricultural businesses seeking to enhance operational efficiency and profitability. Geographically, North America and Europe currently hold significant market share, driven by early adoption of advanced technologies and well-established agricultural infrastructure. However, rapidly developing economies in Asia-Pacific, particularly India and China, represent lucrative growth opportunities, with increasing investment in agricultural modernization and technological advancements. The market's growth, while promising, faces certain restraints including the high initial investment costs associated with advanced technologies and the need for reliable internet connectivity and digital literacy among farmers, particularly in developing regions. Overcoming these barriers will be crucial to unlocking the full potential of the crop yield forecasting market in the coming years.

Global Machine Learning For Crop Yield Prediction market size is expected to reach $2.58 billion by 2029 at 26.6%, rising demand for sustainable agriculture driving the growth of the market due to environmental and food security concerns

This dataset provides detailed logs comparing projected and actual crop yields at the field level, including forecast methods, timing, and variance calculations. It enables precise analysis of forecast accuracy, supports agricultural planning, and helps identify patterns or anomalies in yield prediction. The data is ideal for farm management optimization, agronomic research, and improving predictive models.

This dataset is designed to help researchers and data scientists predict crop yield based on key agricultural factors. It contains columns representing the amount of fertilizer used, temperature, and soil nutrients (nitrogen, phosphorus, and potassium), along with the corresponding crop yield. This dataset can be used for machine learning projects focusing on agricultural optimization, yield forecasting, and resource management. The data is structured and ready for analysis, making it ideal for training regression models or conducting exploratory data analysis.

File Information: Number of Rows: 2,596 Number of Columns: 6 Data Types: Fertilizer: float64 temp: float64 N (Nitrogen): float64 P (Phosphorus): float64 K (Potassium): float64 yeild (Yield): float64 Missing Values: None (0 missing values in all columns) Summary Statistics: Fertilizer: Mean: 66.49, Min: 49.75, Max: 80.22 Temperature: Mean: 33.85, Min: 23.77, Max: 40.27 Nitrogen (N): Mean: 69.52, Min: 58.84, Max: 80.22 Phosphorus (P): Mean: 20.71, Min: 17.72, Max: 25.16 Potassium (K): Mean: 17.81, Min: 14.70, Max: 22.06 Yield: Mean: 8.53, Min: 5.15, Max: 12.34 This dataset is complete, with no missing values, and provides diverse statistics for various features important for crop yield prediction.

The global crop yield forecasting market is experiencing robust growth, driven by the increasing need for efficient agricultural practices and enhanced food security in a changing climate. The market, estimated at $2.5 billion in 2025, is projected to witness a Compound Annual Growth Rate (CAGR) of 15% from 2025 to 2033, reaching an estimated value of $8 billion by 2033. This expansion is fueled by several key factors. Technological advancements in remote sensing, data analytics, and artificial intelligence are enabling the development of sophisticated forecasting models, providing farmers with more accurate and timely predictions of crop yields. Furthermore, the rising adoption of precision agriculture techniques, coupled with the growing awareness of climate change's impact on crop production, is further propelling market growth. The increasing demand for higher crop yields to meet the rising global food demand also contributes significantly to this expansion. The market segmentation reveals strong growth across both software and service offerings, with the commercial application segment dominating due to its larger scale operations and greater investment capacity in advanced technologies. Geographic growth is anticipated to be particularly strong in regions like Asia-Pacific and North America, driven by higher technology adoption rates and significant agricultural sectors. However, challenges remain, including the high initial investment costs associated with implementing these technologies, a lack of digital literacy among some farming communities, and potential data security concerns. Despite these restraints, the long-term outlook for the crop yield forecasting market remains positive, with continued innovation and technological advancements expected to drive further market expansion in the coming years.

The use of ML in agronomy has been increasing exponentially since the start of the century, including data-driven predictions of crop yields from farm-level information on soil, climate and management. However, little is known about the effect of data partitioning schemes on the actual performance of the models, in special when they are built for yield forecast. In this study, we explore the effect of the choice of predictive algorithm, amount of data, and data partitioning strategies on predictive performance, using synthetic datasets from biophysical crop models. We simulated sunflower and wheat data using OilcropSun and Ceres-Wheat from DSSAT for the period 2001-2020 in 5 areas of Spain. Simulations were performed in farms differing in soil depth and management. The data set of farm simulated yields was analyzed with different algorithms (regularized linear models, random forest, artificial neural networks) as a function of seasonal weather, management, and soil. The analysis was performed with Keras for neural networks and R packages for all other algorithms. Data partitioning for training and testing was performed with ordered data (i.e., older data for training, newest data for testing) in order to compare the different algorithms in their ability to predict yields in the future by extrapolating from past data. The Random Forest algorithm had a better performance (Root Mean Square Error 35-38%) than artificial neural networks (37-141%) and regularized linear models (64-65%) and was easier to execute. However, even the best models showed a limited advantage over the predictions of a sensible baseline (average yield of the farm in the training set) which showed RMSE of 42%. Errors in seasonal weather forecasting were not taken into account, so real-world performance is expected to be even closer to the baseline. Application of AI algorithms for yield prediction should always include a comparison with the best guess to evaluate if the additional cost of data required for the model compensates for the increase in predictive power. Random partitioning of data for training and validation should be avoided in models for yield forecasting. Crop models validated for the region and cultivars of interest may be used before actual data collection to establish the potential advantage as illustrated in this study.

Georeferenced crop yield prediction is a valuable tool for agronomists and policymakers. One challenge with many existing datasets is that of location accuracy. GPS locations for fields can end up offset from the true location due to sensor inaccuracies or from locations being collected at the edges of fields rather than the field centers. This makes it harder to connect remote-sensed data to the yield values. The goal of this project was to produce a method that can help correct these location offsets by finding the most probable field center given an input location. We prepared and hosted a competition on Zindi (https://zindi.africa) where competitors model the problem using state-of-the-art data science techniques. We provided the competitors with satellite images of fields along with their corresponding manually annotated correct centers. Additionally, we also provided approximate plot size and measured yield in case these help with creating their solutions. Original positions are considered images' centers as (0,0) and a displacement vector for each field in the training set was provided. The goal of the competition was to predict these vectors for each vector in the test set. This dataset includes the locations of 18,481 crop fields across Kenya, Tanzania, and Rwanda, collected in 2016-2017 with mixed qualities and their error-corrected ones from the winning solution using Planet satellite imagery and the Yolo v5 object detection model.

The global crop yield forecasting market is experiencing robust growth, driven by the increasing need for precision agriculture and improved food security. The market's expansion is fueled by several key factors, including the rising adoption of advanced technologies like AI, machine learning, and remote sensing, which enable more accurate and timely yield predictions. Furthermore, climate change and its unpredictable impact on agricultural production are pushing farmers and stakeholders to leverage predictive analytics to mitigate risks and optimize resource allocation. The market is segmented based on various factors, including technology (satellite imagery, weather data, soil sensors), application (field-level forecasting, regional forecasting), and crop type. While the precise market size is not provided, considering a typical CAGR (Compound Annual Growth Rate) of 15-20% for such a technology-driven market, a reasonable estimate for the 2025 market size could be in the range of $500 million to $750 million, given the involvement of established players like Pessl Instruments and emerging companies like AgroMetShell and CropProphet. This growth trajectory is projected to continue throughout the forecast period (2025-2033), driven by continuous technological advancements and increasing adoption rates across different regions. Constraints such as high initial investment costs for technology and data accessibility limitations, especially in developing countries, may pose some challenges, but are likely to be offset by the significant economic and environmental benefits offered by accurate crop yield forecasting. The competitive landscape comprises a mix of established players and innovative startups. Companies like EOSDA and Agronomy Insights are leveraging their expertise in data analytics and remote sensing to offer comprehensive solutions, while others focus on niche applications or specific crop types. The market is also witnessing increasing collaboration between agricultural research institutions (like CCAFS and WUR) and technology companies, further accelerating innovation. Geographical expansion, particularly in developing economies with significant agricultural sectors, presents significant growth opportunities. The integration of crop yield forecasting with other precision agriculture technologies, such as variable-rate fertilization and irrigation, promises to further enhance efficiency and sustainability in agricultural practices, strengthening the market's long-term prospects. Overall, the crop yield forecasting market is poised for substantial growth, contributing significantly to enhancing agricultural productivity and addressing global food security concerns.

California Crop Yield Benchmark: Combining Satellite Image, Climate, Evapotranspiration, and Soil Data Layers for County-Level Yield Forecasting of Over 70 Crops

This benchmark offers a comprehensive, unified, and multi-modal dataset for county-level crop yield prediction across California. It integrates diverse data sources, including monthly time series from Landsat satellite imagery, monthly evapotranspiration (ET) data, daily DayMet climate variables, static soil attributes, annual… See the full description on the dataset page: https://huggingface.co/datasets/hkaman/california-crop-yield-benchmark.

CY-Bench: A comprehensive benchmark dataset for sub-national crop yield forecasting

Overview

CY-Bench is a dataset and benchmark for subnational crop yield forecasting, with coverage of major crop growing countries of the world for maize and wheat. By subnational, we mean the administrative level where yield statistics are published. When statistics are available for multiple levels, we pick the highest resolution. The dataset combines sub-national yield statistics with relevant predictors, such as growing-season weather indicators, remote sensing indicators, evapotranspiration, soil moisture indicators, and static soil properties. CY-Bench has been designed and curated by agricultural experts, climate scientists, and machine learning researchers from the AgML Community, with the aim of facilitating model intercomparison across the diverse agricultural systems around the globe in conditions as close as possible to real-world operationalization. Ultimately, by lowering the barrier to entry for ML researchers in this crucial application area, CY-Bench will facilitate the development of improved crop forecasting tools that can be used to support decision-makers in food security planning worldwide.

* Crops : Wheat & Maize
* Spatial Coverage : Wheat (29 countries), Maize (38).
See CY-Bench paper appendix for the list of countries.
* Temporal Coverage : Varies. See country-specific data

Data

Data format

The benchmark data is organized as a collection of CSV files, with each file representing a specific category of variable for a particular country. Each CSV file is named according to the category and the country it pertains to, facilitating easy identification and retrieval. The data within each CSV file is structured in tabular format, where rows represent observations and columns represent different predictors related to a category of variable.

Data content

All data files are provided as .csv.

Folder structure

The CY-Bench dataset has been structure at first level by crop type and subsequently by country. For each country, the folder name follows the ISO 3166-1 alpha-2 two-character code. A separate .csv is available for each predictor data and crop calendar as shown below. The csv files are named to reflect the corresponding country and crop type e.g. **variable_croptype_country.csv**.
```
CY-Bench
│
└─── maize
│ │
│ └─── AO
│ │ -- crop_calendar_maize_AO.csv
│ │ -- fpar_maize_AO.csv
│ │ -- meteo_maize_AO.csv
│ │ -- ndvi_maize_AO.csv
│ │ -- soil_maize_AO.csv
│ │ -- soil_moisture_maize_AO.csv
│ │ -- yield_maize_AO.csv
│ │
│ └─── AR
│ -- crop_calendar_maize_AR.csv
│ -- fpar_maize_AR.csv
│ -- ...
│
└─── wheat
│ │
│ └─── AR
│ │ -- crop_calendar_wheat_AR.csv
│ │ -- fpar_wheat_AR.csv
│ │ ...
```

Example : CSV data content for maize in country X

```
X
└─── crop_calendar_maize_X.csv
│ -- crop_name (name of the crop)
│ -- adm_id (unique identifier for a subnational unit)
│ -- sos (start of crop season)
│ -- eos (end of crop season)
│
└─── fpar_maize_X.csv
│ -- crop_name
│ -- adm_id
│ -- date (in the format YYYYMMdd)
│ -- fpar
│
└─── meteo_maize_X.csv
│ -- crop_name
│ -- adm_id
│ -- date (in the format YYYYMMdd)

│ -- tmin (minimum temperature)
│ -- tmax (maximum temperature)
│ -- prec (precipitation)
│ -- rad (radiation)
│ -- tavg (average temperature)
│ -- et0 (evapotranspiration)
│ -- cwb (crop water balance)
│
└─── ndvi_maize_X.csv
│ -- crop_name
│ -- adm_id
│ -- date (in the format YYYYMMdd)
│ -- ndvi
│
└─── soil_maize_X.csv
│ -- crop_name
│ -- adm_id
│ -- awc (available water capacity)
│ -- bulk_density
│ -- drainage_class
│
└─── soil_moisture_maize_X.csv
│ -- crop_name
│ -- adm_id
│ -- date (in the format YYYYMMdd)
│ -- ssm (surface soil moisture)
│ -- rsm ()
│
└─── yield_maize_X.csv
│ -- crop_name
│ -- country_code
│ -- adm_id
│ -- harvest_year
│ -- yield
│ -- harvest_area
│ -- production

Data access

The full dataset can be downloaded directly from Zenodo or using the ```zenodo_get``` library

License and citation

We kindly ask all users of CY-Bench to properly respect licensing and citation conditions of the datasets included.

The Machine Learning for Crop Yield Prediction report features an extensive regional analysis, identifying market penetration levels across major geographic areas. It highlights regional growth trends and opportunities, allowing businesses to tailor their market entry strategies and maximize growth in specific regions.

Context

A number of research is underway in the agricultural sector to better predict crop yield using machine learning algorithms. Many machine learning algorithms require large amounts of data in order to give useful results. One of the major challenges in training and experimenting with machine learning algorithms is the availability of training data in sufficient quality and quantity remains a limiting factor. In the paper, “Wild blueberry yield prediction using a combination of computer simulation and machine learning algorithms”, we used dataset generated by the Wild Blueberry Pollination Model, a spatially explicit simulation model validated by field observation and experimental data collected in Maine USA during the last 30 years. The blueberry yields predictive models require data that sufficiently characterize the influence of plant spatial traits, bee species composition, and weather conditions on production. In a multi-step process, we designed simulation experiments and conducted the runs on the calibrated version of the blueberry simulation model. The simulated dataset was then examined, and important features were selected to build four machine-learning-based predictive models. This simulated data provides researchers who have actual data collected from field observation and those who wants to experiment the potential of machine learning algorithms response to real data and computer simulation modelling generated data as input for crop yield prediction models.

Acknowledgements

Qu, Hongchun; Obsie, Efrem; Drummond, Frank (2020), “Data for: Wild blueberry yield prediction using a combination of computer simulation and machine learning algorithms”, Mendeley Data, V1, doi: 10.17632/p5hvjzsvn8.1

Dataset

This dataset was created by Aya Shabbar

Contents

On average, AI in agriculture was estimated to significantly increase the crop yield prediction accuracy. It was estimated that AI can predict crop yield by ** percent accuracy rate.