Mapping arable field areas is crucial for assessing agricultural productivity but poses challenges in sub-Saharan agroecosystems because of diverse crop calendars, small and irregularly shaped fields, persistent cloud cover, and lack of high-quality model training data. This study proposes several methodological improvements to overcome these challenges. Specifically, it utilizes long-term MODIS data to stratify finer Sentinel-2 reflectance and Sentinel-1 backscatter image features on a per-pixel basis. It also incorporates texture features and employs a machine learning approach with over 300,000 samples. The eastern region of Ghana was stratified into seven seasonal strata exhibiting distinct vegetation seasonality, capturing diversity in crop calendars, using long-term MODIS (2001–2009) normalized difference vegetation index phenology. Three years (2017–2019) of Sentinel-1 and Sentinel-2 original bands at 20 m were composited into dry and wet seasonal features according to the strata, from which spectral, polarimetric, and texture features were extracted. The field boundaries were digitized using PlanetScope images (2018–2019). Random Forest classifier with 10-fold cross-validation and recursive feature elimination was used for feature selection and model building. Including topographic variables, out of 137 image features, only 11 features were found important. Sentinel-2 SWIR-based spectral features were most important, followed by Sentinel-1 polarimetric (VV) and elevation features. Half of the 11 features were variance texture features, followed by spectral features. The Random Forest classifier produced a 0.78 AUC score with overall precision, recall, and F1-score of 0.96, 0.78, and 0.85, respectively. While the precision for both classes was >0.90, the recall rate for arable areas was half that of non-arable areas. Future studies could improve the technical workflow with reliable balanced sampling, narrowband hyperspectral images, and fully polarized SAR images.
Early information on crops spatial extent is useful from food security standpoint. Growing number of remote sensing satellites are making it easier to map the crop areas in a rapid and cost-effective way. Grape being an commercial fruit crop of India, it is useful to map the grape areas for acreage estimation and monitor its health continuously to get optimum yield. The main contribution of this paper is mapping areas under export and non-export grapes and their health estimation based on hybrid classification-clustering approach using Sentinel-2 satellite data. Nashik district of Maharashtra, India was chosen as study area which is known to be a famous grape belt. We have used Sentinel-2 image of 31st December 2016 to carry out the classification. Two step classification approach have been followed wherein first step involves classification of grape and non-grape areas using broadband spectral features. Normalized Difference Vegetation Index (NDVI), Land Surface Water Index (LSWI) and Tasseled Crop Transformation (TCT) components viz. Brightness (BR), Greenness (GR) and Wetness (WT) estimated using Sentinel-2 data and slope estimated from ASTER DEM were used during the first step. We evaluate the performance of classifiers such as Artificial Neural Networks (ANN), Random Forest (RF) and Support Vector Machine (SVM). Result shows RF performs best among all classifiers with overall classification accuracy of 93.57 % on validation dataset. Further, comparison of grape areas estimated using our proposed approach with statistics obtained from district agriculture department shows deviation of only 1.84 % in the total grape area. Second step involves classification of export and non-export grapes. Texture features such as entropy, homogeneity and dissimilarity estimated from red-edge bands (Bands 5,6,7) of sentinel-2 have been fed as an input to the same set of classifiers. Results depict the maximum classification accuracy of 81.82 % using RF classifier. Further, estimated export grape area was compared with district agriculture statistics which shows the deviation of 7.98 %. It is also important to monitor the crop health continuously to get the better yield and returns. We have further carried out the health estimation of non-export grape areas. The objective was to get the clusters of healthy, low stressed and high stressed grape areas. ISODATA clustering has been carried out using the indices such as Difference Vegetation Index (DVI), Inverted Red-Edge Chlorophyll Index (IRECI) and NDVI45, Red-Edge Position (REP) and NDVI. Comparison of clusters with ground truth yield data of 31 fields shows that the fields with high stress have obtained lesser yields than others. We envision to provide health related information to the farmers through our existing mKRISHI(îRplatform for entire crop season using proposed approach.
22 May 2024 in American Journal of Enology and Viticulture
Background and goals In winegrape production, clusters are sampled multiple times per growing season to assess fruit maturation. The economic cost of implementing a previously-developed spatial sampling method using normalized difference vegetation index (NDVI) images to assess block variability was compared to the economic costs of random sampling (R20) and of sampling via the four corners approach (4C). The objective was to compare the travel distance to sample blocks between the spatial sampling protocol (NDVI3), R20, and 4C, and the time and cost associated with the required travel distance of each sampling method. Methods and key findings Travel distances for each sampling method were calculated for six commercial vineyard blocks. The cost of required labor and travel was calculated for each travel distance. Travel distances per block for the R20, 4C, and NDVI3 methods ranged from 3.34 to 13.63 km, 0.68 to 2.41 km, and 0.36 to 1.58 km (all per sampling event), respectively. Total cost savings per sampling event using NDVI3 ranged from $5.54 to $32.40 when compared to R20, and from $0.33 to $4.61 when compared to 4C. Conclusions and significance The cost savings of using NDVI3 for maturation monitoring appears relatively small compared to R20 and 4C, but considering that most blocks are sampled multiple times leading up to harvest, and that larger wine companies have thousands of acres to sample for fruit maturity, savings may become more substantial. We conclude that adoption of the NDVI3 sampling protocol for fruit maturity may substantially reduce the cost of production for winegrapes grown in large blocks, compared to R20 and 4C sampling.
30 September 2023 in Australian Journal of Grape and Wine Research
Mapping and monitoring cluster morphology provides insights for disease risk assessment, quality control in wine production, and understanding environmental influences on cluster shape. During the progression of grapevine morphology, cluster closure (CC) (also called bunch closure) is the stage when berries touch one another. This study used mobile phone images to develop a direct quantification method for tracking CC in three grapevine cultivars (Riesling, Pinot gris, and Cabernet Franc). A total of 809 cluster images from fruit set to veraison were analyzed using two image segmentation methods: (i) a Pyramid Scene Parsing Network (PSPNet) to extract cluster boundaries and (ii) Otsus image thresholding method to calculate CC based on gaps between the berries. PSPNet produced high accuracy (mean accuracy=0.98, mean intersection over union (mIoU)=0.95) with mIoU>0.90 for both cluster and noncluster classes. Otsus thresholding method resulted in <2 falsely classified gap and berry pixels affecting quantified CC. The progression of CC was described using basic statistics (mean and standard deviation) and using a curve fit. The CC curve showed an asymptotic trend, with a higher rate of progression observed in the first three weeks, followed by a gradual approach towards an asymptote. We propose that the X value (in this example, number of weeks past berry set) at which the CC progression curve reaches the asymptote be considered as the official phenological stage of CC. The developed method provides a continuous scale of CC throughout the season, potentially serving as a valuable open-source research tool for studying grapevine cluster phenology and factors affecting CC.