The strategic management of agricultural lands involves crop field monitoring each year. Crop discrimination via remote sensing is a complex task, especially if different crops have a similar spectral response and cropping pattern. In such cases, crop identification could be improved by combining object-based image analysis and advanced machine learning methods. In this investigation, we evaluated the C4.5 decision tree, logistic regression (LR), support vector machine (SVM) and multilayer perceptron (MLP) neural network methods, both as single classifiers and combined in a hierarchical classification, for the mapping of nine major summer crops (both woody and herbaceous) from ASTER satellite images captured in two different dates. Each method was built with different combinations of spectral and textural features obtained after the segmentation of the remote images in an object-based framework. As single classifiers, MLP and SVM obtained maximum overall accuracy of 88%, slightly higher than LR (86%) and notably higher than C4.5 (79%). The SVM+SVM classifier (best method) improved these results to 89%. In most cases, the hierarchical classifiers considerably increased the accuracy of the most poorly classified class (minimum sensitivity). The SVM+SVM method offered a significant improvement in classification accuracy for all of the studied crops compared to the conventional decision tree classifier, ranging between 4% for safflower and 29% for corn, which suggests the application of object-based image analysis and advanced machine learning methods in complex crop classification tasks.

Remote sensing technology allows monitoring the progress of vegetation and crop phenology in large regions. Seasonal vegetation trends are commonly estimated from high temporal resolution but coarse spatial resolution satellite imagery, e.g., from MODIS-NDVI (Moderate Resolution Imaging Spectroradiometer—Normalized Difference Vegetation Index) time-series, which has usually limited their application to scenarios with few land uses or crops covering areas larger than actual parcel sizes. As an alternative, this paper proposes a general and robust procedure to map crop phenology at the level of individual crop parcels, and validates its feasibility in a complex and diverse cropland area located in central California. A first calibration phase consisted of evaluating the three curve-fitting models implemented in the TIMESAT software (i.e., asymmetric Gaussian (AG), double logistic (DL), and adaptive Savitzky–Golay (SG) filtering) and reporting the model and its settings that best adjusted to the MODIS-NDVI profile of each crop studied. Next, based on the selected crop-specific models and with a crop map previously obtained from ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) multi-temporal images, the procedure mapped four crop calendar events (i.e., start, end, middle, and length of the season) and five phenology-related metrics (i.e., base, maximum, amplitude, derivatives, and integrals of the NDVI values) of the study region by object-based image analysis (OBIA) of the MODIS-NDVI time-series. To mitigate the impact of mixed pixels, the OBIA procedure was designed to automatically apply a restrictive criterion based on the coverage of MODIS-NDVI pixels in each crop parcel: (1) using only the MODIS-NDVI pixels that were placed 100% within each crop parcel (i.e., “pure” pixels); or (2) if no “pure” pixels exist in any crop parcel, using only pixels with coverage percentages greater than 50%, and in such cases, reporting the mixing percentage in the output file. The calibration phase showed that the performance of the SG filtering was superior in most crops, with the exception of rice, while the AG model was intermediate in all of the cases. Differences between the dates of the start and end of the season that were observed in 120 ground-truth fields and the ones estimated by the crop-specific models were in a range of 11 days (for the corn fields) and 22 days (for the vineyard fields) on average. The OBIA procedure was also validated in 240 independent parcels with “pure” MODIS-NDVI pixels, reporting 89% and 82% of accuracy when mapping the start and end of the season, respectively. Our results revealed different growth patterns of the studied crops, especially of the crop calendar events of herbaceous (i.e., corn, rice, sunflower, and tomato) and woody crops (i.e., almond, walnut, and vineyard), of the NDVI derivatives of rice and the other studied herbaceous crops, and of the NDVI integrals of vineyard and the other studied woody crops. The resulting maps and tables provide valuable geospatial information for every parcel over time with several applications in cropland management, irrigation scheduling, and ecosystem modeling.

Fundamental to most classical data collection sampling theory development is the random drawings assumption requiring that each targeted population member has a known sample selection (i.e., inclusion) probability. Frequently, however, unrestricted random sampling of spatially autocorrelated data is impractical and/or inefficient. Instead, randomly choosing a population subset accounts for its exhibited spatial pattern by utilizing a grid, which often provides improved parameter estimates, such as the geographic landscape mean, at least via its precision. Unfortunately, spatial autocorrelation latent in these data can produce a questionable mean and/or standard error estimate because each sampled population member contains information about its nearby members, a data feature explicitly acknowledged in model-based inference, but ignored in design-based inference. This autocorrelation effect prompted the development of formulae for calculating an effective sample size (i.e., the equivalent number of sample selections from a geographically randomly distributed population that would yield the same sampling error) estimate. Some researchers recently challenged this and other aspects of spatial statistics as being incorrect/invalid/misleading. This paper seeks to address this category of misconceptions, demonstrating that the effective geographic sample size is a valid and useful concept regardless of the inferential basis invoked. Its spatial statistical methodology builds upon the preceding ingredients.