Rice Yield Data Research Articles

Integrating solar-induced chlorophyll fluorescence with traditional remote sensing and environmental variables for enhanced rice yield prediction in Nepal using machine learning

Rice is a vital cereal crop providing essential food for Nepal. To mitigate climate risks and ensure food security, especially with climate change and increasing extreme weather events, it is imperative and a prerequisite to have timely and reliable rice yield prediction. While traditional vegetation indices (VIs) and environmental variables are commonly used in rice yield prediction, the potential of solar-induced chlorophyll fluorescence (SIF), despite its greater sensitivity, has been explored in limited studies. Furthermore, integrating SIF with other remote sensing and environmental data for regional crop yield modeling still needs to be explored. This study aims to address these gaps by incorporating SIF to enhance rice yield prediction in Nepal, offering a novel approach to improving model accuracy. This study uses multi-source data and machine learning algorithms to study 22 major rice-growing districts in the Terai region of Nepal. Random forest (RF), support vector machine (SVR), gradient boosting regressor (GBR), and least absolute shrinkage and selection operator (LASSO) algorithms were employed to predict crop yield using various combinations of input features. Machine learning models were trained and tested over four different time windows within the growing period of rice, using 15 years of rice yield data from 2003 to 2017. Our analysis reveals that GBR out-performed other machine learning algorithms when using input features from July to September, achieving root mean square error (RMSE) and the ratio of RMSE (RRMSE) to observed mean values of 250 kg/ha and 8.6%, respectively. The findings also reveal that incorporating SIF alongside traditional remote sensing and environmental variables significantly enhances the model's prediction accuracy. These findings are valuable for ensuring food security, optimizing agricultural resource management, and supporting decision-making for farmers, policymakers, and supply chain stakeholders.

Feasibility of machine learning-based rice yield prediction in India at the district level using climate reanalysis and remote sensing data

CONTEXTYield forecasting, the science of predicting agricultural productivity before the crop harvest occurs, helps a wide range of stakeholders make better decisions around agricultural planning. OBJECTIVEThis study aims to investigate whether machine learning-based yield prediction models can capably predict Kharif season rice yields at the district level in India several months before the rice harvest takes place. METHODOLOGYThe methodology involved training 19 machine learning models such as CatBoost, LightGBM, Orthogonal Matching Pursuit, and Extremely Randomized Trees on 20 years of climate, satellite, and rice yield data across 247 of India's rice-producing districts. In addition to model-building, a dynamic dashboard was built understand how the reliability of rice yield predictions varies across district. RESULTS AND CONCLUSIONSThe results of the proof-of-concept machine learning pipeline demonstrated that rice yields can be predicted with a reasonable degree of accuracy, with out-of-sample R2, MAE, and MAPE performance of up to 0.82, 0.29, and 0.16 respectively. This performance outperformed test set performance reported in related literature on rice yield modelling in other contexts and countries. In addition, SHAP value analysis was conducted to infer both the importance and directional impact of the climate and remote sensing variables included in the model. Important features driving rice yields included temperature, soil water volume, and leaf area index. In particular, higher temperatures in August correlate with increased rice yields, particularly when the leaf area index in August is also high. Building on the results, a proof-of-concept dashboard was developed to allow users to easily explore which districts may experience a rise or fall in yield relative to the previous year. The dashboard show that the model may perform better in some regions than in others. For instance, the absolute percentage error for predicted versus actual yields ranged from an average of 7.1 % in districts in Uttarakhand to an average of 14.7 % in Uttar Pradesh. SIGNIFICANCEThis study underscores the potential for policymakers to consider scaling and operationalizing machine learning approaches to rice yield prediction in the context of agricultural early warning systems to deliver timely crop yield forecasts on a rolling basis throughout the season, thereby equipping agricultural decision-makers with the ability to make informed choices on irrigation scheduling, fertilizer application, and harvest planning to optimize crop output and resource use.

Estimation of rice yield using multivariate analysis techniques based on meteorological parameters

This study aims to develop predictive models for rice yield by applying multivariate techniques. It utilizes stepwise multiple regression, discriminant function analysis and logistic regression techniques to forecast crop yield in specific districts of Haryana. The time series data on rice crop have been divided into two and three classes based on crop yield. The yearly time series data of rice yield from 1980–81 to 2020–21 have been taken from various issues of Statistical s of Haryana. The study also utilized fortnightly meteorological data sourced from the Agrometeorology Department of CCS HAU, India. For comparing various predictive models' performance, evaluation of measures like Root Mean Square Error, Predicted Error Sum of Squares, Mean Absolute Deviation and Mean Absolute Percentage Error have been used. Results of the study indicated that discriminant function analysis emerged as the most effective to predict the rice yield accurately as compared to logistic regression. Importantly, the research highlighted that the optimum time for forecasting the rice yield is 1 month prior to the crops harvesting, offering valuable insight for agricultural planning and decision-making. This approach demonstrates the fusion of weather data and advanced statistical techniques, showcasing the potential for more precise and informed agricultural practices.

Impact of water resources variation on winter–spring rice yield in the upper Vietnamese Mekong Delta: A case study of An Giang Province

AbstractNatural and socio‐economic development under climate change has affected the distribution of surface water resources (SWRs) in many deltas worldwide. The dynamics in the SWRs has negatively impacted agricultural production, which is one of the growing issues in the Vietnamese Mekong Delta (VMD) as the national rice bowl. Our study aims to assess SWR dynamics related to its effects on rice production in the upper VMD. We applied Mann–Kendall, linear regression and Pettit tests to statistically determine trends of SWRs in terms of water level, total water volume and rainfall in 1996–2020. The findings show that high water level trends decrease significantly after 2011, and this reduction has been due to climate variability and dam development upstream of the Mekong River. Moreover, rice yield data analysis from 1996 to 2020 indicated a decreasing trend in the yield of the winter–spring rice crop in the An Phu district of An Giang Province after 2011. Our interviews with 33 water experts and 90 local farmers revealed 80% agreement that the decline in water resources has affected rice yield. This study provides empirical evidence and a platform for knowledge sharing and learning across different deltas and initiatives worldwide.

Feasibility of mechanical rice transplanting in organic Italian rice system

Abstract Rice cultivation in Italy is usually performed by direct seeding in flooded or in dry fields. Mechanical rice transplanting is a technique that can help control weeds and improve rice competition. To test the feasibility of the technique for different rice varieties in Italy, a study was conducted in the Lombardy region (northwest Italy), from 2016 to 2018. The study also evaluated the efficacy of hoeing in transplanted rice fields. The experiment consisted of two studies, a ‘field plot experiment’ (conducted from 2017 to 2018) and an ‘on-farm transplanting trial’ (conducted from 2016 to 2018). The ‘field plot experiment’ was carried out using a split plot design to determine the optimal transplanting distances within the row (12 or 17 cm) and the most suitable rice varieties for transplanting. Hoeing was performed once in 2017 and twice in 2018. Weed infestation was assessed by counting the number of weeds within a randomly placed square frame in the interrow areas, both before and after hoeing. Rice plant density, panicle density, yield, and yield components were also assessed. The ‘on-farm transplanting trial’ tested transplanting in several farms over the years. Different rice varieties were transplanted using the same machines, and hoeing was performed according to a predetermined schedule. Transplanter performance was assessed as well as rice yield data in all fields, while four fields were selected each year to assess hoeing efficacy against weeds. ANOVAs were used to test the differences in weed control and rice parameters among varieties and transplanting distances. The ‘field plot experiment’ showed that the transplanting distance did not affect weeds or rice variety. Transplanting at 12 cm within the row resulted in a higher plant density compared to 17 cm, however tillering compensated for the difference in the number of panicles. Carnaroli consistently recorded the lowest yield, less than 2 t ha−1, while Selenio, Spillo, and Laser seemed to be better suited for transplanting achieving the highest yield in 2018 (about 7 ha−1). In the ‘on-farm transplanting experiment’ hoeing was effective in controlling weeds, although the machineries used were not always able to function properly in saturated soil. Most of the transplanted field yielded approximately 3 to 5 t ha−1. Varieties with round grain exhibited the greatest yield variability among fields. The study suggests that to achieve a high yield in organic rice, the transplanting technique should be combined with an effective interrow tillage to control weeds.

Temporal variation of the relationships between rice yield and climate variables since 1925.

Long-term time-series datasets of crop yield and climate variables are necessary to study the temporal variation of climate effects on crops. The aim of this study was to broadly assess assessment of the effects of climate on rice, and the associated temporal variations of the effects during the long-term period. We conducted field experiments in Taiwan from 1925 to 2019 to collect and analyze rice yield data and evaluate the impacts of changes in average temperature, diurnal temperature range (DTR), rainfall, and sunshine duration on rice yield during cool and warm cropping seasons. We then estimated the relationships between annual grain yield and the climate variables using the time series of their first difference values. We also computed the total relative and annual actual yield changes using regression coefficients for each climate variable for the intervals 1925-1944, 1945-1983, and 1996-2019 to reveal the impacts of climate change on yields and the associated temporal variations during the overall experimental period. The annual daily average temperature calculated from the trend of the regression lines increased by 0.94-1.03 °C during the 95-year period. The maximum temperature remained steady while the minimum temperature increased, leading to decreased DTR. The total annual rainfall decreased by 237-352 mm and the annual total sunshine duration decreased by 93.9-238.9 h during the experimental period. We observed that during the cool cropping season, yield response to temperature change decreased, while that to DTR and rainfall changes increased. During the warm cropping season, all the yield responses to temperature, DTR, and rainfall changes were negative throughout the experimental period. In recent years (1996-2019) the estimated annual actual rice yield changes during the cool cropping season were negatively affected by climate variables (except for sunshine duration), and slightly positively affected (except for temperature) during the warm cropping season. Compared to the effects of temperature and DTR, those of rainfall and sunshine duration on rice yield changes were weak. This study contributes to provide impacts of climate change on rice yield and associated long-term temporal variations over nearly a century.

Response of lowland rice varieties to integrated nutrient management in a derived Savannah agro-ecology

In the inland valley (FADAMA) of the Directorate of University Farms (DUFARMS) of the Federal University of Agriculture, Abeokuta (FUNAAB) in Nigeria, an experiment was conducted successfully. The objective was to see how integrated nutrient sources influenced the performance of two lowland varieties of rice. A split-plot two-factorial experiment was used in the study, and it was replicated three times in a randomized complete block design. Rice seedlings were transplanted 12 days after being planted in the nursery to a depth of 3 cm and a spacing of 20 cm × 20 cm, one seedling per hill. Two varieties and six nutrient sources were used for the research. Rice yield and growth data were collected. ANOVA (Analysis of Variance) was used on data, and the significant means were separated using Tukey HSD (Honestly Significant Difference) at a probability level of 5%. ARICA 3 grew taller than NERICA L-34 as a result of the research. ARICA (Advanced Rice Varieties for Africa) 3 exceeded NERICA (The New Rice for Africa) L-34 in terms of shoot dry weight (4 WAP (Weeks After Planting) (1.53) and 6 WAP (1.53 g), respectively (5.60 g). Rice cultivars produced the highest grain when 25 kg N ha-1 (NPK 20:10:10) and 75 kg N/ha (Poultry manure) were applied.

Rice Yield Forecasting: A Comparative Analysis of Multiple Machine Learning Algorithms

Agriculture plays a crucial role in Nigeria's economy, serving as a vital source of sustenance and livelihood for numerous Nigerians. With the escalating impact of climate change on crop yields, it becomes imperative to develop models that can effectively study and predict rice output under varying climatic conditions. This study collected rice yield data from Katsina state, spanning the years 1970 to 2017, sourced from the Nigeria Bureau of Statistics. Additionally, climatic data for the same period were obtained from the World Bank Climate Knowledge portal. Logistic Regression (LR), Artificial Neural Network (ANN), Random Forest (RF), Random Trees (RT), and Naïve Bayes (NB) were employed to develop rice yield prediction models utilizing this dataset. The findings reveal that random forest and random trees exhibited superior classification performance for yield prediction. The developed models offer a promising tool for predicting future rice yields, facilitating proactive measures to ensure food security for the people of the state.

Effect of climate change-induced water-deficit stress on long-term rice yield.

The water requirements of crops should be investigated to improve the efficiency of water use in irrigated agriculture. The main objective of the study was to assess the effects of water deficit stress on rice yields throughout the major cropping seasons. We analyzed rice yield data from field experiments in Taiwan over the period 1925-2019 to evaluate the effects of water-deficit stress on the yield of 12 rice cultivars. Weather data, including air temperatures, humidity, wind speed, sunshine duration, and rainfall were used to compute the temporal trends of reference evapotranspiration and crop water status (CWS) during rice growth stages. A negative CWS value indicates that the crop is water deficient, and a smaller value represents a lower water level (greater water-deficit stress) in crop growth. The CWS on rice growth under the initial, crop development, reproductive, and maturity stages declined by 96.9, 58.9, 24.7, and 198.6 mm in the cool cropping season and declined by 63.7, 18.1, 8.6, and 3.8 mm in the warm cropping season during the 95 years. The decreasing trends in the CWSs were used to represent the increases in water-deficit stress. The total yield change related to water-deficit stress on the cultivars from 1925-1944, 1945-1983, and 1996-2019 under the initial, crop development, reproductive, and maturity stages are -56.1 to 37.0, -77.5 to -12.3, 11.2 to 19.8, and -146.4 to 39.1 kg ha-1 in the cool cropping season and -16.5 to 8.2, -12.9 to 8.1, -2.3 to 9.0, and -9.3 to 8.0 in the warm cropping season, respectively. Our results suggest that CWS may be a determining factor for rice to thrive during the developmental stage, but not the reproductive stage. In addition, the effect of water-deficit stress has increasingly affected the growth of rice in recent years.

Rice Yield Prediction in Hubei Province Based on Deep Learning and the Effect of Spatial Heterogeneity

Timely and accurate crop yield information can ensure regional food security. In the field of predicting crop yields, deep learning techniques such as long short-term memory (LSTM) and convolutional neural networks (CNN) are frequently employed. Many studies have shown that the predictions of models combining the two are better than those of single models. Crop growth can be reflected by the vegetation index calculated using data from remote sensing. However, the use of pure remote sensing data alone ignores the spatial heterogeneity of different regions. In this paper, we tested a total of three models, CNN-LSTM, CNN and convolutional LSTM (ConvLSTM), for predicting the annual rice yield at the county level in Hubei Province, China. The model was trained by ERA5 temperature (AT) data, MODIS remote sensing data including the Enhanced Vegetation Index (EVI), Gross Primary Productivity (GPP) and Soil-Adapted Vegetation Index (SAVI), and a dummy variable representing spatial heterogeneity; rice yield data from 2000–2019 were employed as labels. Data download and processing were based on Google Earth Engine (GEE). The downloaded remote sensing images were processed into normalized histograms for the training and prediction of deep learning models. According to the experimental findings, the model that included a dummy variable to represent spatial heterogeneity had a stronger predictive ability than the model trained using just remote sensing data. The prediction performance of the CNN-LSTM model outperformed the CNN or ConvLSTM model.

Pre-harvest Forecasting Models for Kharif Rice Yield in South Gujarat Using Composite Weather Indices

The pre-harvest forecasting models were developed for Kharif rice yield forecasts for Navsari, Surat and Tapi districts respectively. The data of rice yield and the weather parameters from 1995 to 2017 was used for developing statistical models for three districts of south Gujarat. The weather indices like Z41, Z50, Z231, Z241, Z341, Z351 and Time were able to forecast the yield of rice for Navsari district. Similarly, Z41, Z51, Z131, Z141, Z231, Z241, Z351 and Time were found to be most efficient predictors for Surat district. Only two variables i.e., Z241 and Z251 were found to be able to forecast the rice yield in Tapi district. Models were validated with 3 years (2018, 2019 and 2020) data. Results indicated that models explained 53 to 93 percent variations for rice yield during F1 stage, 54 to 92 percent variation for rice yield during F2 stage and 52 to 93 percent variations during F3 stage for rice yield in three districts. Hence these models can be used to some extent for forecast the yield in different districts of south Gujarat a few months before harvest.

Downscaling Global Gridded Crop Yield Data Products and Crop Water Productivity Mapping Using Remote Sensing Derived Variables in the South Asia.

The online version contains supplementary material available at 10.1007/s42106-022-00223-2.

Design of feed rate monitoring system and estimation method for yield distribution information on combine harvester

Real-time monitoring of harvester feed rate is of great significance for guiding harvesting work and improving harvesting efficiency. In this study, a feed rate monitoring system is designed and developed. The system consists of a torque sensor for measuring the header power shaft torque, an angel sensor for measuring header height, a humidity sensor for measuring grains moisture content, a Global Navigation Satellite System receiver, and a vehicle industrial computer. Data, such as the moisture content of grains and the working position of the harvester, are collected by the system during harvest. The height of the header, the moisture content of grains, and the torque of the header power shaft are selected as the input quantity, and feed rate is selected as the output quantity. A calculation model for feed rate is established using a particle swarm optimization–back propagation neural network, and then real-time monitoring of feed rate is realized. On the basis of the monitoring of feed rate, an estimation method for yield distribution information is proposed. Through analyzing the rice samples in the experimental area, the method establishes a binary linear model between the header height, the grain moisture content, and the stem-to-grain ratio of the rice fed to the harvester. The model is combined with the monitoring results of feed rate to obtain the rice yield data fed to the harvester. Then, the rice yield data are matched with the positional information at the time of harvesting to obtain a yield distribution map of the harvested area. Field experiments are conducted in a rice-growing area in Northeast China, and the average relative error of the feed rate monitoring system is 7.69%. The field sample verification accuracy of the stem-to-grain ratio model is higher than 90%, and the correlation coefficient R2 is 0.93. The estimation accuracy of the yield distribution is higher than 90%, which provides a scientific basis for agricultural decision making.

Advances in rice yield estimation using Neural Networks

Timely and reliable estimation of crop yield is an important dimension in agriculture as it aids in effective and timely policy decisions. Weather is a the most important factor, affecting crop yield in the agricultural domain and rice crop is no exception. The present study has been taken up to identify the effect of weekly weather parameters namely bright sunshine hours, maximum temperature, minimum temperature, morning relative humidity, evening relative humidity, and weekly total rainfall on rice crop yield being an important staple food of India. In order to suggest a suitable neural network model for rice yield estimation, Ranga Reddy District of Telangana state was chosen and weekly averages of weather variables from the 30th to 47th meteorological standard weeks (MSWs) of 31 years and rice yield data from 1988-89 to 2018-19 were considered in the study. Back A back propagation neural network and two activation functions namely logistic sigmoid and linear were employed in the neural network model. The proposed neural network model “F” (Input Neurons =11, Hidden Neurons=12, Output Neuron=1, Train Data Size = 80 % and Test data Size=20%) exhibited better results with the low MAE and AEER% while estimating rice yields. All the estimated yields of respective years were close to the actual yields as the multiple correlation coefficients (R) values for train and test data were also close to 1. The errors of simulated estimation of rice yield ranged between -8.1 to -3.8 % for the proposed neural networks model. Thus, better rice yield was estimated by using the proposed neural network model “F”.

Climate Change and Farmers' Adaptation Strategies: The Case of Rice Producers in Nghe An Province, Vietnam

This research aimed at assessing climate trends and adaptation strategies to the impacts of climate change on rice producers in Nghe An province. For the first objective, multiple graphs, the Mann–Kendall test, and Sen’s Slope test were employed to analyze the secondary data of rainfall, temperature, and rice yield in Nghe An from 1995 to 2019. The results showed that the rice productivity of Nghe An increased over the 25-year period, annual rainfall decreased in the winter-spring season, and annual temperature increased in the summer-autumn season. For the second objective, primary data were collected by conducting household surveys of 396 households in Nghe An province. Sixty-two percent of farmers did not adopt any adaptive strategies, while other strategies included changing the planting schedule, growing stress-tolerant crops, improving irrigation practices, and utilizing fertilizer and other good practices. Multivariate probit regression (MVP) was employed to analyze the factors affecting the adaptation strategies of households in the study area. The results of MVP showed that age, education, gender, household size, farm size, cooperative, farmer organization, extension services, access to credit, access to irrigation, rainfall, and living in Quynh Luu or Dien Chau significantly affected the farmers’ choice. The results implied that farmers can obtain advanced technology, social learning, and community practices by participating in farmer organizations and cooperatives. In specified area, local authorities should invest in improving irrigation systems, practicing land consolidation, practicing special farming techniques, and providing timely weather forecasts for enhancing farmers’ resilience.

Modelling the Impacts of Climate Change on the Yield of Crops

This study aims to improve the understanding of the impact changes being experienced in our climate system will have on the level of crop productivity in the immediate period as well as in the nearest future. Nigeria was used as a case study and an observed climatic dataset was obtained and used alongside collected 20 year cassava, rice and soybean yield data to develop models that were applied to estimate future crop yield. Four statistically downscaled and bias-corrected Global Climate Models (GCMs): NOAA, MIROC5, ICHEC, and NCC performed simulations for the period 1985–2100 under the Representative Concentration Pathway RCP8.5. These were used to predict how the yields of cassava, rice and soybean will be in the years 2020-2050 and 2070-2100 for the 36 states in Nigeria and the FCT. 89 Empirical models were developed to estimate the yields of the three crops earlier mentioned across Nigeria with their coefficient of determination (R2) ranging between 15% - 99%. The result showed an increase of 3.91% (P<0.001), 0.08, 1.79 (P<0.1) and a decrease of 0.93% for cassava yield for ICHEC, MIROC, NOAA and NCC respectively. It also projected an increase in yield of 8.88% (P<0.001), 7.77% (P<0.001), 6.62% (P<0.001) and 8.85% (P<0.001) for Rice yield using climatic data from ICHEC, MIROC, NOAA and NCC respectively. Soybean, increase in yield are 2.81% (P<0.01), 5.84% (P<0.001), 11.38 (P<0.001) and 9.06% (P<0.001) for ICHEC, MIROC, NOAA and NCC respectively.

Pre-harvest forecast of kharif rice yield using PCA and MLR technique in Navsari district of Gujarat

In this paper Principal Components (PC) and Multiple Linear Regression (MLR) Technique were used for development of pre-harvest model for rice yield in the Navsari district of south Gujarat. The weather indices were developed and utilized for development of pre-harvest forecast models. The data of rice yield and weather parameters from 1990 to 2012 were utilized. The cross validation of the developed forecast model were confirmed using data of the years 2013 to 2016. It was observed that value of Adj. R2 varied from 89 to 96. The appropriate forecast model was selected based on high value of Adj. R2. Based on the outcomes in Navsari district, MLR techniques found to be better than PCA for pre harvest forecasting of rice crop yield. The Model-2 found competent to forecast rice yield in Navsari district before eight weeks of actual harvest of crop (37th SMW) i.e during reproductive stage of the crop growth period.

Predicting Rice Production in Central Thailand Using the WOFOST Model with ENSO Impact

The World Food Studies Simulation Model (WOFOST) model is a daily crop growth and yield forecast model with interactions with the environment, including soil, agricultural management, and especially climate conditions. An El Niño–Southern Oscillation (ENSO) phenomenon directly affected climate change and indirectly affected the rice yield in Thailand. This study aims to simulate rice production in central Thailand using the WOFOST model and to find the relationship between rice yield and ENSO. The meteorological data and information on rice yields of Suphan Buri 1 variety from 2011 to 2018 in central Thailand were used to study the rice yields. The study of rice yield found that the WOFOST model was able to simulate rice yield with a Root Mean Square Error (RMSE) value of 752 kg ha−1, with approximately 16% discrepancy. The WOFOST model was able to simulate the growth of Suphan Buri 1 rice, with an average discrepancy of 16.205%, and Suphan Buri province had the least discrepancy at 6.99%. Most rice yield simulations in the central region were overestimated (except Suphan Buri) because the model did not cover crop damage factors such as rice disease or insect damage. The WOFOST model had good relative accuracy and could respond to estimates of rice yields. When an El Niño phenomenon occurs at Niño 3.4, it results in lower-than-normal yields of Suphan Buri 1 rice in the next 8 months. On the other hand, when a La Niña phenomenon occurs at Niño 3.4, Suphan Buri 1 rice yields are higher than normal in the next 8 months. An analysis of the rice yield data confirms the significant impact of ENSO on rice yields in Thailand. This study shows that climate change leads to impacts on rice production, especially during ENSO years.

Within-Field Rice Yield Estimation Based on Sentinel-2 Satellite Data

Rice is considered one of the most important crops in the world. According to the Food and Agriculture Organization of the United Nations (FAO), rice production has increased significantly (156%) during the last 50 years, with a limited increase in cultivated area (24%). With the recent advances in remote sensing technologies, it is now possible to monitor rice crop production for a better understanding of its management at field scale to ultimately improve rice yields. In this work, we monitor within-field rice production of the two main rice varieties grown in Valencia (Spain) JSendra and Bomba during the 2020 season. The sowing date of both varieties was May 22–25, while the harvesting date was September 15–17 for Bomba and October 5–8 for JSendra. Rice yield data was collected over 66.03 ha (52 fields) by harvesting machines equipped with onboard sensors that determine the dry grain yield within irregular polygons of 3–7 m width. This dataset was split in two, selecting 70% of fields for training and 30% for validation purposes. Sentinel-2 surface reflectance spectral data acquired from May until September 2020 was considered over the test area at the two different spatial resolutions of 10 and 20 m. These two datasets were combined assessing the best combination of spectral reflectance bands (SR) or vegetation indices (VIs) as well as the best timing to infer final within-field yields. The results show that SR improves the performance of models with VIs. Furthermore, the correlation of each spectral band and VIs with the final yield changes with the dates and varieties. Considering the training data, the best correlation with the yields is obtained on July 4, with R2 for JSendra of 0.72 at 10 m and 0.76 at 20 m resolution, while the R2 for Bomba is 0.87 at 10 m and 0.92 at 20 m resolution. Based on the validation dataset, the proposed models provide within-field yield modelling Mean Absolute Error (MAE) of 0.254 t×ha−1 (Mean Absolute Percentage Error, MAPE, of 3.73%) for JSendra at 10 m (0.240 t×ha−1; 3.48% at 20 m) and 0.218 t×ha−1 (MAPE 5.82%) for Bomba (0.223 t×ha−1; 5.78% at 20 m) on July 4, that is three months before harvest. At parcel level the model’s MAE is 0.176 t×ha−1 (MAPE 2.61%) for JSendra and 0.142 t×ha−1 (MAPE 4.51%) for Bomba. These results confirm the close correlation between the rice yield and the spectral information from satellite imagery. Additionally, these models provide a timeliness overview of underperforming areas within the field three months before the harvest where farmers can improve their management practices. Furthermore, it highlights the importance of optimum agronomic management of rice plants during the first weeks of rice cultivation (40–50 days after sowing) to achieve high yields.

Rice Yield Simulation and Planting Suitability Environment Pattern Recognition at a Fine Scale

Analyzing rice yields and multidimensional environmental factors at a fine scale facilitates the discovery of the planting environment patterns that guide the spatial layout of rice production. This study uses Pucheng County, Fujian Province, a demonstration county of China Good Grains and Oils, as the research area. Using actual rice yield sample data and environment data, a yield simulation model based on random forest regression is constructed to realize a fine-scale simulation of rice yield and its spatial distribution pattern in Pucheng County. On this basis, we construct a method system to identify spatial combination patterns between rice yields and fine-scale multidimensional environmental planting suitability using rice yield data and environmental planting suitability evaluation data. We categorize the areas into four combination model areas to analyze the spatial correlation model of planting suitability, multidimensional environment, and yield: higher-yield and higher-suitability cluster–comprehensive environmental-advantage areas, high-yield and high-suitability cluster–soil condition-limited areas, moderate-yield and moderate-suitability cluster–irrigation and drainage condition-limited areas, and low-yield and low-suitability cluster–site condition-limited areas. The following results are found. (1) The rice yield simulation model, which is based on random forest regression, considers the various complex relationships between yield and natural as well as human factors to realize the refined simulation of rice yields at a county scale. (2) The county rice yield has a strong positive spatial correlation, and the spatial clustering characteristics are obvious; these relationships can provide a basis for effectively implementing intensive rice planting in Pucheng County. (3) We construct a spatial combination pattern recognition method based on rice yield and environmental planting suitability. We can use this method to effectively identify the spatial relationship between yield and planting suitability as well as the shortcomings and advantages of different regions in terms of the climate, soil, irrigation, site, mechanical farming, and similar factors. On this basis, we can provide regional rice planting guidance for Pucheng County. In addition, this method system also provides a new perspective and method for research into spatial combination models and related spatial issues.