Estimating the timing of rice flowering is useful for predicting maturity, which subsequently informs management decisions such as field drainage and harvest timing that optimize yield and quality. This study utilized satellite-based remote sensing to identify the highest measured vegetation index (VI) value, or peak of the season (PoS), which was correlated with observed rice flowering dates. Sentinel-2 (S2) and daily Planet Fusion (PF) data were used to derive four VIs and evaluate their effectiveness in determining rice flowering dates using an offset-adjusted peak-VI approach. Our methodology was tested in the major rice-growing regions of New South Wales, Australia, including commercial and experimental fields, multiple varieties, during the 2021–22, 2022–23, and 2023–24 seasons. Additionally, we assessed the extent to which variations in flowering dates induced by nitrogen content differences can be estimated. The results show that the chlorophyll index green (CIG) derived from PF provided a consistent relationship between PoS and observed flowering dates, with an offset of approximately 13 days and a root mean squared error (RMSE) of 6.56 and 7 days for the seasons 2021–22 (n = 114 sample sites) and 2022–23 (n = 134 sample sites), respectively. Furthermore, the study found that the method was able to describe the variation in flowering dates due to differences in nitrogen uptake at most sites, with R^2 values ranging from 0.09 to 0.97; higher explained variance was observed at sites where there was a wider range of flowering dates. The methodology facilitates the estimation of flowering dates and is scalable from pixel, to field, to regional level analysis.

Timely and accurate assessment of rice productivity, encompassing grain yield, aboveground biomass (AGB), and harvest index (HI), is essential for harvest planning, supply chain coordination, and food security. This study evaluates the feasibility of predicting all three productivity components using satellite and weather time series data while examining trade-offs between forecast accuracy and operational lead time. Five machine learning models (CatBoost, Gaussian Process Regression (GPR), Random Forest, Ridge regression, and TabPFN) were compared across six in-season prediction windows (December to May) using Sentinel-2 vegetation indices (Normalized Difference Vegetation Index (NDVI), Chlorophyll Index Red Edge 2 (CIRE2), Land Surface Water Index (LSWI)), weather variables (minimum and maximum temperature and radiation), and agronomic records from 256 commercial and experimental rice fields in southern New South Wales, Australia, over four growing seasons (2022–2025) using leave-one-year-out cross-validation. Rolling in-season forecasts were evaluated across December–May; March was selected for further analysis as a practical window that balances accuracy and timeliness for decision-making, with minimal additional error reduction in later months closer to harvest. TabPFN had the lowest RMSE for yield prediction (RMSE = 1.85 t ha−1, r=0.72), Ridge had the lowest RMSE for AGB (RMSE = 3.05 t ha−1, r=0.77), while tree-based models yielded the lowest RMSE for derived HI (RMSE ≈ 0.07). HI prediction showed weak regional relationships, with direct prediction yielding |r|≈0.24 and derived HI (predicted yield divided by predicted AGB) showing r≈0. Although strong correlations (r>0.9) between HI and vegetation indices were observed within individual site-seasons, consistent with other studies, these relationships were highly variable across site-seasons, reflecting the difficulty of inferring HI from canopy reflectance when biotic and/or abiotic stresses decouple AGB accumulation from grain filling. Both direct and derived HI approaches yielded comparable errors, indicating that satellite and weather data lack information content for regional-scale HI prediction. These findings support satellite-based yield and AGB forecasting for operational use.

Phenology prediction is critical for optimizing the timing of rice crop management operations such as fertilization and irrigation, particularly in the face of increasing climate variability. This study aimed to estimate three key developmental stages in the temperate irrigated rice systems of Australia: panicle initiation (PI), flowering, and harvest maturity. Extensive and diverse field observations (n≈302) were collected over four consecutive seasons (2022–2025) from the rice-growing regions of the Murrumbidgee and Murray Valleys in southern New South Wales, encompassing six varieties and three sowing methods. The extent of data available allowed a number of traditional and emerging machine learning (ML) models to be directly compared to determine the most robust strategies to predict Australian rice crop phenology. Among all models, Tabular Prior-data Fitted Network (TabPFN), a pre-trained transformer model trained on large synthetic datasets, achieved the highest precision for PI and flowering predictions, with root mean square errors (RMSEs) of 4.9 and 6.5 days, respectively. Meanwhile, long short-term memory (LSTM) excelled in predicting harvest maturity with an RMSE of 5.9 days. Notably, TabPFN achieved strong results without the need for hyperparameter tuning, consistently outperforming other ML approaches. Across all stages, models that integrated remote sensing (RS) and weather variables consistently outperformed those relying on single-source input. These findings underscore the value of hybrid data fusion and modern time series modeling techniques for accurate and scalable phenology prediction, ultimately enabling more informed and adaptive agronomic decision-making.

Remote sensing enables real-time insights into crop progress, aiding farmer decision-making on crop management factors such as fertilization and irrigation. However, artifacts in satellite image time series (SITS), including clouds, sun and sensor geometry variations, intersensor biases, and atmospheric effects, cause reflectance and vegetation index fluctuations, complicating interpretation. To improve SITS temporal consistency, this study developed empirical models that correct these artifacts. The focus scenario was rice monitoring, including 7266 rice fields from four years in New South Wales, Australia. However, the method generalized to corn and cotton crops with promising results. Masking cloud-affected data using Cloud Score+ provided a better trade-off between image frequency and time-series smoothness compared with other cloud masking methods. Self-supervised LightGBM models outperformed linear models in correcting these deviations. The importance of key features including solar and view angles, reflectances in bands sensitive to atmospheric effects and the satellite sensor (differences between Sentinel-2A and -2B, especially in a red edge band), were indicated by SHAP analyses. Surprisingly, top of atmosphere time series had similar or better consistency than harmonized Landsat Sentinel surface reflectance time series after correcting the deviations. Our method reduced the root mean squared deviation of a red edge chlorophyll index by 39.7% and the normalized difference vegetation index by 21.4%. These corrections enhance SITS interpretability for near real-time crop monitoring, improving data in farmer decision-support tools.

Optimizing the timing of rice paddy drainage and harvest is crucial for maximizing yield and quality. These decisions are guided by rice grain moisture content (GMC), which is typically determined by destructive plant samples taken at point locations. Providing rice farmers with predictions of GMC will reduce the time burden of gathering, threshing and testing samples. Additionally, it will reduce errors due to samples being taken from unrepresentative areas of fields, and will facilitate advanced planning of end-of-season drain and harvest timing. This work demonstrates consistent relationships between rice GMC and indices derived from Sentinel-2 satellite imagery, particularly those involving selected shortwave infrared and red edge bands (r=0.84, 1620 field samples, 3 years). A methodology was developed to allow forecasts of grain moisture past the latest image date to be provided, by fusing remote sensing and accumulated weather data as inputs to machine learning models. The moisture content predictions had root mean squared error between 1.6 and 2.6 R^2 of 0.7 with forecast horizons from 0 to 28 days. Time-series grain moisture dry-down predictions were summarized per field to find the optimal harvest date (22

Deep learning reduced macadamia planting year error at a national scale, achieving a pixel-level Mean Absolute Error (MAE) of 1.2 years and outperforming a vegetation index threshold baseline (MAE 1.6 years) and tree-based models—Random Forest (RF; MAE 3.02 years) and Gradient Boosted Trees (GBT; MAE 2.9 years). Using Digital Earth Australia Landsat annual geomedians (1988–2023) and block-level, industry-supplied planting year data, models were trained and evaluated at the pixel level under a strict Leave-One-Region-Out cross-validation (LOROCV) protocol; a secondary block-level random split (80/10/10) is reported only to illustrate the more optimistic setting, where shared regional conditions yield lower errors (0.6–0.7 years). Predictions reconstruct planting year retrospectively from the full historical record rather than providing real-time forecasts. The final model was then applied to all Australian Tree Crop Map (ATCM) macadamia orchard polygons to produce wall-to-wall planting year estimates. The approach enables fine-grained mapping of planting patterns to support yield forecasting, resource allocation, and industry planning. Results indicate that sequence-based deep models capture informative temporal dynamics beyond thresholding and conventional machine learning baselines, while remaining constrained by regional and temporal data sparsity. The framework is scalable and transferable, offering a pathway to planting year mapping for other perennial crops and to more resilient, data-driven agricultural decision-making.

Effective nitrogen (N) management is critical for optimizing rice production. This study utilized Sentinel-2 (S2) spectral bands and vegetation indices (VIs) to predict N uptake at panicle initiation (PI) in Australian rice fields using data from 2022 to 2024 (n = 238). We employed various regression models-including linear, ridge, lasso, elastic net, random forest, and XGBoost-to validate predictions across seasons, locations, and field types (experimental and commercial). The results indicated that the normalized difference red edge index (NDRE1 + NDRE2)2 was an effective predictor when combined with the linear regression model, achieving root mean square errors (RMSEs) ranging from 18.6 to 28.2 kg/ha (R-2: 0.6-0.75) across the three test years. Notably, the linear regression model performed slightly better than all other machine learning models tested, demonstrating both simplicity and robustness. Models trained on commercial fields predicted experimental fields with an RMSE of 22.36 kg/ha (R-2 = 0.67), while the reverse yielded an RMSE of 30.98 kg/ha (R-2 = 0.36). Location-wise validation showed RMSEs ranging from 23.88 kg/ha to 26.13 kg/ha, with corresponding R-2 values of -0.18 to 0.70. These findings underscore that (NDRE1 + NDRE2)2, combined with linear regression, can be effectively utilized for field-scale N monitoring.

Monitoring agricultural crops in near real-time using remote sensing data can aid optimized decision-making to achieve productivity and sustainability goals. For example, deciding how much fertilizer to apply to rice during the panicle initiation (PI) growth stage is supported by timely provision of vegetation indices (VIs) that are related to nitrogen status, such as the red edge chlorophyll index (CIRE). However, cloud during satellite overpasses can prevent this. This work aims to estimate important VIs derived from Sentinel-2 multispectral data, using a combination of Sentinel-1 synthetic aperture radar (SAR) data and field dates (sowing, ponding and PI) to fill data gaps. We used 7,336 fields harvested in the 2018-2023 seasons for training, and 1,804 fields from 2024 for testing. The LightGBM algorithm produced the best results compared with random forest, ridge and lasso regression. Day of season (DOS) alone outperformed SAR data (CIRE RMSE 10% vs 13%). However the best results (RMSE 6%) were obtained by fusing DOS, field dates and SAR data. The model was able to explain 92% of CIRE variability across the whole season, and 56% at the critical PI growth stage.

Rice yield depends on factors including variety, weather, field management, nutrient and water availability. We analyzed important drivers of yield variability at the field scale, and developed yield forecast models for crops in the temperate irrigated rice growing region of Australia. We fused a time-series of Sentinel-1 and Sentinel-2 satellite remote sensing imagery, spatial weather data and field management information. Rice phenology was predicted using previously reported models. Higher yields were associated with early flowering, higher chlorophyll indices and higher temperatures around flowering. Successive rice cropping in the same field was associated with lower yield (p<0.001). After running a series of leave-one-year-out cross validation experiments, final models were trained using 2018–2022 data, and were applied to predicting the yield of 1580 fields (43,700 hectares) from an independent season with challenging conditions (2023). Models which aggregated remote sensing and weather time-series data to phenological periods provided more accurate predictions than models that aggregated these predictors to calendar periods. The accuracy of forecast models improved as the growing season progressed, reaching RMSE=1.6 t/ha and Lin’s concordance correlation coefficient (LCCC) of 0.67 30 days after flowering at the field level. Explainability was provided using the SHAP method, revealing the likely drivers of yield variability overall, and of individual fields.

Crop phenology modeling often involves determining variety-specific growing degree day thresholds, or parameterizing mechanistic crop models. In this work, we used machine learning methods to develop models that provide daily predictions of the probability that rice (Oryza sativa) crops had reached the panicle initiation and flowering growth stages. These per-date classifications were summarized into per-paddock growth stage transition dates, which were then compared with field-sampled reference data, encompassing 15 rice varieties, 10 years, and 380 sites. Leave-one-year-out cross validation was used to provide realistic estimates of model errors. Compared with more complex and computationally intensive algorithms, logistic regression produced competitive results (mean cross-season validation RMSE 3.9 and 5.2 days for panicle initiation and flowering, respectively). Logistic regression had additional advantages: providing confidence of growth stage predictions at each date (as it is a probabilistic algorithm), and straightforward explainability (as model parameters directly indicated how the various input variables contributed to growth stage predictions). Input variables included accumulated weather, rice variety, and sowing methods. The models were applied to forecasting phenology transition dates of the rice crops planted throughout the Murray and Murrumbidgee valleys. In addition, recommendations for optimal sowing dates were developed, using simulations involving more than 40 years of weather data, with the goal of minimizing the risk of cold-temperatures during the microspore growth phase, which can severely degrade yield in temperate rice growing regions.

Tree- and block-level prediction of mango yield is important for farm operations, but current manual methods are inefficient. Previous research has identified the accuracies of mango yield forecasting using very-high-resolution (VHR) satellite imagery and an ’18-tree’ stratified sampling method. However, this approach still requires infield sampling to calibrate canopy reflectance and the derived block-level algorithms are unable to translate to other orchards due to the influences of abiotic and biotic conditions. To better appreciate these influences, individual tree yields and corresponding canopy reflectance properties were collected from 2015 to 2021 for 1958 individual mango trees from 55 orchard blocks across 14 farms located in three mango growing regions of Australia. A linear regression analysis of the block-level data revealed the non-existence of a universal relationship between the 24 vegetation indices (VIs) derived from VHR satellite data and fruit count per tree, an outcome likely due to the influence of location, season, management and cultivar. The tree-level fruit count predicted using a random forest (RF) model trained on all calibration data produced a percentage root mean squared error (PRMSE) of 26.5% and a mean absolute error (MAE) of 48 fruits/tree. The lowest PRMSEs produced from RF-based models developed from location, season and cultivar subsets at the individual tree level ranged from 19.3% to 32.6%. At the block level, the PRMSE for the combined model was 10.1% and the lowest values for the location, seasonal and cultivar subset models varied between 7.2% and 10.0% upon validation. Generally, the block-level predictions outperformed the individual tree-level models. Maps were produced to provide mango growers with a visual representation of yield variability across orchards. This enables better identification and management of the influence of abiotic and biotic constraints on production. Future research could investigate the causes of spatial yield variability in mango orchards.

Macadamia yield forecast models were trained with a large set of commercial yield data (10 years, 1,156 records). Predictors included remote sensing and weather data, aggregated spatially to macadamia block boundaries, and temporally to quarterly intervals. Errors were typically around 23% at the block level, and 10% at the region level. Much of the yield variability yield was predicted even for orchards excluded from training data. At least 400-500 training data points were needed to minimize error. Best results were obtained with a fusion of weather and remote sensing data, aggregated over 8 quarterly periods from 2 years before harvest.

Rice field management around maturity and harvest are some of the most difficult decisions growers face. Field drainage and harvest timing affect quality, yield, and post-harvest drying costs. These decisions are informed by grain moisture content (MC). Over three years, three sites and three varieties, we studied the field dry-down rate and time to optimal harvest MC. We showed that field-specific parameters significantly affected these characteristics, including rice variety, Nitrogen applied (NA), mid-season N uptake (NU) and dry matter (DM). Increased N and DM is associated with increased MC and thus delays time to harvest. We developed models based on linear regression and nonlinear machine learning (ML) algorithms, including parameters describing these field-specific conditions. Cross validation across the three years provided a realistic expectation of model prediction errors. A linear model with the addition of nonlinear predictors achieved competitive performance compared with more complex and less interpretable ML models. When MC was modeled as a function of days since heading, similar or better accuracy was achieved to using accumulated weather parameters. Moisture content was predicted with mean absolute error of 2.1 %. The predicted time from heading to harvest MC was improved by the inclusion of field-specific parameters (N and variety) from mean absolute error of 6.8 days to 5.7 days. The final linear regression model explained 80 % of the moisture variability in the dataset, and provided estimates of dry-down rates, moisture as a function of time, and time to reach harvest moisture. This study shows the importance of including field-specific parameters when estimating of rice harvest timing, and provides methods to model these effects.

Accurate pre-harvest yield forecasting of mango is essential to the industry as it supports better decision making around harvesting logistics and forward selling, thus optimizing productivity and reducing food waste. Current methods for yield forecasting such as manually counting 2–3% of the orchard can be accurate but are very time inefficient and labour intensive. More recent evaluations of technological solutions such as remote (satellite) and proximal (on ground) sensing have provided very encouraging results, but they still require infield in-season sampling for calibration, the technology comes at a significant cost, and commercial availability is limited, especially for vehicle-mounted sensors. This study presents the first evaluation of a ”time series”—based remote sensing method for yield forecasting of mango, a method that does not require infield fruit counts and utilizes freely available satellite imagery. Historic yield data from 2015 to 2022 were sourced from 51 individual orchard blocks from two farms (AH and MK) in the Northern Territory of Australia. Time series measures of the canopy reflectance properties of the blocks were obtained from Landsat 7 and 8 satellite data for the 2015–2022 growing seasons. From the imagery, the following vegetation indices (VIs) were derived: EVI, GNDVI, NDVI, and LSWI, whilst corresponding weather variables (rainfall (Prec), temperature (Tmin/Tmax), evapotranspiration (ETo), solar radiation (Rad), and vapor pressure deficit (vpd)) were also sourced from SILO data. To determine the relationships among weather and remotely sensed measures of canopy throughout the growing season and the yield achieved (at the block level and the farm level), six machine learning (ML) algorithms, namely random forest (RF), support vector regression (SVR), eXtreme gradient boosting (XGBOOST), RIDGE, LASSO and partial least square regression (PLSR), were trialed. The EVI/GNDVI and Prec/Tmin were found to be the best RS and weather predictors, respectively. The block-level combined RS/weather-based RF model for 2021 produced the best result (MAE = 2.9 t/ha), marginally better than the RS only RF model (MAE = 3.4 t/ha). The farm-level model error (FLEM) was generally lower than the block-level model error, for both the combined RS/weather-based RF model (farm = 3.7%, block (NMAE) = 33.6% for 2021) and the RS-based model (farm = 4.3%, block = 38.4% for 2021). Further testing of the RS/weather-based RF models over six additional orchards (other than AH and MK) produced errors ranging between 24% and 39% from 2016 to 2020. Although accuracies of prediction did vary at both the block level and the farm level, this preliminary study demonstrates the potential of a ”time series” RS method for predicting mango yields. The benefits to the mango industry are that it utilizes freely available imagery, requires no infield calibration, and provides predictions several months before the commercial harvest. Therefore, this outcome not only presents a more adoptable option for the industry, but also better supports automation and scalability in terms of block-, farm-, regional, and national level forecasting.

This study presents a comprehensive evaluation of seasonal, locational, and varietal variations in canopy reflectance responses in 315 commercial citrus blocks from three major growing regions in Australia. The dataset includes three different citrus types (Mandarin, Navel, Valencia) and 26 varieties. The aim is to utilize this combined information to better understand yield variation and develop improved forecasting models. Landsat satellite data spanning from October 2006 to February 2021 (1419 tiles) were used to derive reflectance values, and calculate four vegetation indices (NDVI, GNDVI, LSWI, and GCVI), for each citrus block. These indices were then analyzed alongside corresponding yield data, which consisted of 3660 individual yield records dating back to 2007. Two temporal resolutions were incorporated as predictors: spatio-temporal vegetation index time series (TS) aggregated every two months and annual time series of historical block-yield records. Six statistical and machine learning algorithms were calibrated using a leave-one-year-out cross-validation approach (LOYO CV) and validated for one-year forward prediction over a five-year period (2017–2021). The results highlight significant yield variations across years, alternate bearing patterns, and spatio-temporal changes in reflectance profiles influenced by seasonal conditions, varietal characteristics, and locations. The support vector machine (SVM) algorithm with a radial basis function kernel consistently outperformed other algorithms, indicating a non-linear relationship between citrus yield and predictors. The SVM model achieved an RMSE of 15.5 T ha−1, R2 of 0.88, MAE of 12.1 T ha−1, and MAPE of 29% in predicting block-yield across farms, varieties, and seasons. These prediction accuracy metrics demonstrate an improvement over current forecasting methods. Notably, the proposed approach utilizes freely available imagery, provides forecasts between two to nine months before harvest, and eliminates the need for infield counting of fruit load for image calibration. This approach provides an improved method for understanding seasonal yield variation and quantifying citrus block-yield, offering valuable insights for growers in harvest logistics, labor allocation, and resource management.

Accurate, non-destructive forecasting of carrot yield is difficult due to its subterranean growing habit. Furthermore, the timing of forecasting usually occurs when the crop is mature, limiting the opportunity to implement alternative management decisions to improve yield (during the growing season). This study aims to improve the accuracy of carrot yield forecasting by exploring time series and multivariate approaches. Using Sentinel-2 satellite imagery in three Australian vegetable regions, we established a time series of carrot phenological stages (PhS) from ‘days after sowing’ (DAS) to enhance prediction timing. Numerous vegetation indices (VIs) were analyzed to derive temporal growth patterns. Correlations with yield at different PhS were established. Although the average root yield (t ha−1) did not significantly differ across the regions, the temporal VI signatures, indicating different regional crop growth trends, did vary as well as the PhS at when the maximum correlation with yield occurred ( PhS_R2_max ) with two of the regions producing a delayed PhS_R2_max (i.e. 90–130 DAS). The best multivariate model was identified at 70 DAS, extending the forecasting window before harvest between 20 to 60 days. The performance of this model was validated with new crops producing an average error of 16.9 t ha−1 (27

Rice yield depends on factors including variety, weather, field management, nutrient and water availability. We analyzed important drivers yield variability at the field scale, and developed yield forecast models for crops in the temperate irrigated rice growing region of Australia. We fused a time-series of Sentinel-1 and Sentinel-2 satellite remote sensing imagery, spatial weather data and field management information, while phenology was predicted using previously reported models. Higher yields were associated with early flowering, higher indices indicating nitrogen status, and higher temperatures around flowering, while successive rice cropping in the same field was associated with lower yield. Rice yield forecast models which aggregated remote sensing and weather time-series data to phenological periods provided more accurate models than aggregating to calendar periods. Yield particularly depended on reflectances and vegetation indices based on red edge and short wave infrared bands just before flowering, and minimum temperature around flowering. Final models were trained using 2018-2022 data, and were applied to predicting the yield of 1580 fields (43,700 hectares) from an independent season with challenging conditions (2023). The accuracy of forecast models improved as the growing season progressed, by 30 days after flowering reaching RMSE=1.6 t/ha and Lin's concordance correlation coefficient (LCCC) of 0.67 at the field level. Explainability was provided using the SHAP method, revealing the likely causes of low yields in individual fields, including nitrogen status and low temperatures during flowering due to late sowing. The ability of the models to predict inter-annual yield variability was further validated using independent 2006-2017 yield data, achieving RMSE=0.62 t/ha, LCCC=0.6 over seasons and sowing methods.

Rice is unique, in that yields are maximized when it is grown under ponded (or flooded) conditions. This however has implications for water use (an important consideration in water-scarce environments) and greenhouse gas emissions. This work aimed to provide precise predictions of the date when irrigated rice fields were ponded, on a per-field basis. Models were developed using Sentinel-2 data (with the advantage of inclusion of water-sensitive shortwave infrared bands) and Planet Fusion data (which provides daily, temporally consistent, cross-calibrated, gap-free data). Models were trained with data from both commercial farms and research sites in New South Wales, Australia, and over four growing seasons (harvest in 2018–2021). Predictions were tested on the 2022 harvest season, which included a variety of sowing and water management strategies. A time-series method was developed to provide models with features including satellite observations from before and after the date being classified (as ponded or non-ponded). Logistic regression models using time-series features produced mean absolute errors for ponding date prediction of 4.9 days using Sentinel-2 data, and 4.3 days using Planet Fusion data. The temporal frequency of the Planet Fusion data compensated for the lack of spectral bands relative to Sentinel-2.

Regional maps of rice fields provided early in each growing season facilitate production estimates, planning around harvest logistics, marketing and targeted agronomic recommendations. This work develops maps of all irrigated rice fields in New South Wales, Australia. Classification models were trained on reference maps from the 2019 and 2020 harvest seasons. Model predictions were tested against a reference rice map from the 2021 harvest season, covering 60,000 km 2 . The random forest algorithm was used, with features from aggregated time-series of Sentinel-2 imagery. A sequence of maps were generated at intervals of 15 days, from early to late in the growing season, with accuracy assessed at each time. The maps achieved 95% overall accuracy against point samples at 16 January 2021 ( $\approx 80$ days after sowing). Pixel-based F1-scores against the reference map were above 80% for the 1, 16 and 31 January classified maps.

In 2020, mango (Mangifera indica) exports contributed over 40 million tons, worth around US$20 billion, to the global economy. Only 10% of this contribution was made from African countries including Ghana, largely due to lower investment in the sector and general paucity of research into the mango value chain, especially production, quality and volume. Considering the global economic importance of mango coupled with the gap in the use of the remote sensing technology in the sector, this study tested the hypothesis that phenological stages of mango can be retrieved from Sentinel-2 (S2) derived time series vegetation indices (VIs) data. The study was conducted on four mango farms in the Yilo Krobo Municipal Area of Ghana. Seasonal (temporal) growth curves using four VIs (NDVI, GNDVI, EVI and SAVI) for the period from 2017 to 2020 were derived for each of the selected orchards and then aligned with five known phenology stages: Flowering/Fruitset (F/FS), Fruit Development (FRD), Maturity/Harvesting (M/H), Flushing (FLU) and Dormancy (D). The significance of the variation "within" and "between" farms obtained from the VI metrics of the S2 data were tested using single-factor and two-factor analysis of variance (ANOVA). Furthermore, to identify which specific variable pairs (phenology stages) were significantly different, a Tukey honest significant difference (HSD) post-hoc test was conducted, following the results of the ANOVA. Whilst it was possible to differentiate the phenological stages using all the four VIs, EVI was found to be the best related with p < 0.05 for most of the studied farms. A distinct annual trend was identified with a peak in June/July and troughs in December/January. The derivation of remote sensing based 'time series' growth profiles for commercial mango orchards supports the 'benchmarking' of annual and seasonal orchard performance and therefore offers a near 'real time' technology for identifying significant variations resulting from pest and disease incursions and the potential impacts of seasonal weather variations.