In this study, a convolutional neural network (CNN) is developed that predicts crop type in the US great plains using optical and active microwave remote sensing, historical crop type data and economic commodity crop price information. It includes a Markov chain model that predicts the prior probability of crop types from historical crop type sequences for each pixel. This prior and the remotely sensed imagery is fed through a CNN to produce the final crop type. The CNN is trained using imagery and crop type from the previous year. The field-scale accuracy of this method is found to be 8̃5% using imagery from just the first 90 growing days.

In this study, a machine learning based algorithm is developed for dynamically classifying and determining area of agricultural land-use using optical and active microwave remote sensing. It includes a gradient boosted machine trained in the US using labels derived from the United States Department of Agriculture National Agricultural Statistics Service crop land data layers (CDL). This is then applied to the state of Mato Grosso in Brazil with a hidden Markov model applied to temporally stabilize the land-cover predictions. High resolution remote sensing products such as land surface temperature, normalized difference vegetation index, nadir adjusted bidirectional reflectance, land-cover and radar backscatter were used to develop the classification model. The results in the US were validated using a hold-out set with CDL labels. The results in Brazil were validated using the harvested area statistics reported by Companhia Nacional de Abastecimento.

Land cover and land use affect a wide range of regional-scale to global-scale ecosystem processes, and many Earth system models rely on accurate land cover information. However, multitemporal land cover products often show unrealistically high levels of year-to-year label change, particularly at coarse spatial resolution (i.e., 300-500 m). Much of this apparent land cover change arises from errors in classification and does not indicate real change in land cover or land use. In this paper, we present a novel framework that uses a hidden Markov model (HMM) to help distinguish real land cover change from spurious land cover changes in classification time series. We apply the HMM as a postprocessing step to supervised classification, and we solve for the optimal label sequence using existing HMM algorithms. Our results demonstrate that the HMM provides a rigorous framework for capturing temporal context and likelihood of land cover change at each pixel. We evaluated our approach using the MODIS Collection 5.1 Land Cover Type product (MCD12Q1), focusing on areas that have experienced little change over the MODIS time series, and areas that have experienced well characterized change (e.g., deforestation). We show that the HMM method provides label sequences that are more accurate and that exhibit less year-to-year variability than label sequences produced by ensemble-decision-tree classification or by postprocessing heuristics that have been used in recent versions of MCD12Q1 product. The framework that we present offers an improvement over conventional multitemporal land cover classification methods, and it is widely applicable to problems in multitemporal land cover and land use monitoring.

Active Label Correction (ALC) is an interactive method that cleans an established training set of mislabeled examples in conjunction with a domain expert. ALC presumes that the expert who conducts this review is either more accurate than the original annotator or has access to additional resources that ensure a high quality label. A high-cost re-review is possible because ALC proceeds iteratively, scoring the full training set but selecting only small batches of examples that are likely mislabeled. The expert reviews each batch and corrects any mislabeled examples, after which the classifier is retrained and the process repeats until the expert terminates it. We compare several instantiations of ALC to fully-automated methods that attempt to discard or correct label noise in a single pass. Our empirical results show that ALC outperforms single-pass methods in terms of selection efficiency and classifier accuracy. We evaluate the best ALC instantiation on our motivating task of detecting mislabeled and poorly formulated sites within a land cover classification training set from the geography domain.

Phenological information can be provided globally using remote sensing based time-series vegetation indices. Basic differences in the data and methods used can yield different results. This study analysed such differences in the phenological information, mainly onset of greenness (OG), estimated using the Enhanced Vegetation Index (EVI) from Moderate Resolution Imaging Spectroradiometer (MODIS) data and the Terrestrial Chlorophyll Index (MTCI) from Medium Resolution Imaging Spectrometer (MERIS) data. The two datasets were processed independently using different techniques to provide weekly estimates. Differences in the OG results were analysed for two years (2003 & 2006) and at four levels: a) full study area, b) within land cover classes, c) within core zones of each class and d) at the edge zones of each class. It was found that the trend of OG estimated from MODIS and MERIS were spatially similar, although not the same. From 15 Biome classes found in the study area the classes with the greatest differences were evergreen needle leaf, mixed forest and cropland. The differences were mainly due to the characteristic nature of the indices and also, to some extent, due to false internal flags in the algorithms.

Two aspects are crucial when constructing any real world supervised classification task: the set of classes whose distinction might be useful for the domain expert, and the set of classifications that can actually be distinguished by the data. Often a set of labels is defined with some initial intuition but these are not the best match for the task. For example, labels have been assigned for land cover classification of the Earth but it has been suspected that these labels are not ideal and some classes may be best split into subclasses whereas others should be merged. This paper formalizes this problem using three ingredients: the existing class labels, the underlying separability in the data, and a special type of input from the domain expert. We require a domain expert to specify an L × L matrix of pairwise probabilistic constraints expressing their beliefs as to whether the L classes should be kept separate, merged, or split. This type of input is intuitive and easy for experts to supply. We then show that the problem can be solved by casting it as an instance of penalized probabilistic clustering (PPC). Our method, Class-Level PPC (CPPC) extends PPC showing how its time complexity can be reduced from O(N2) to O(NL) for the problem of class re-definition. We further extend the algorithm by presenting a heuristic to measure adherence to constraints, and providing a criterion for determining the model complexity (number of classes) for constraint-based clustering. We demonstrate and evaluate CPPC on artificial data and on our motivating domain of land cover classification. For the latter, an evaluation by domain experts shows that the algorithm discovers novel class definitions that are better suited to land cover classification than the original set of labels.

This paper presents two machine learning techniques that greatly reduce the number of person-hours required to generate high-quality training data for land cover classification. The first technique uses active learning to guide the generation of training data by selecting only the most informative examples for labeling. The second technique identifies and mitigates the impact of mislabeled instances. Both techniques are tested on data from NASA's Moderate Resolution Imaging Spectroradiometer (MODIS), which has required thousands of person hours to label. Our results shows that the active learning method requires fewer labeled examples than random sampling to produce a high quality classifier. Our results on class noise mitigation show that if mislabelings occur, we can further improve classifier accuracy, and that weighting instances by their label confidence outperforms an analogous method that discards suspected mislabelings. If combined, these methods have the potential to make training data generation a more efficient and reliable process.

Although only a small percentage of global land cover, urban areas significantly alter climate, biogeochemistry, and hydrology at local, regional, and global scales. Despite the fact that three billion people live in cities, information related to global geographic patterns in urban extent and density is poor. Here we present results from efforts to map the global distribution of urban land use at 463 m spatial resolution using remotely sensed data from MODIS. Our approach uses a supervised decision tree classification algorithm that we process using region-specific parameters. The resulting map shows that urban areas occupy only 0.5-1.0 percent of the Earth's land area. Because resource and energy consumption is concentrated in urban areas, the geographic distribution and intensity of urban land use has important implications for models and related studies that require information related to role of humans in the global environment.

This paper addresses the task of analyzing the correlation between two related domains X and Y. Our research is motivated by an Earth Science task that studies the relationship between vegetation and precipitation. A standard statistical technique for such problems is Canonical Correlation Analysis (CCA). A critical limitation of CCA is that it can only detect linear correlation between the two domains that is globally valid throughout both data sets. Our approach addresses this limitation by constructing a mixture of local linear CCA models through a process we name correlation clustering. In correlation clustering, both data sets are clustered simultaneously according to the data's correlation structure such that, within a cluster, domain X and domain Y are linearly correlated in the same way. Each cluster is then analyzed using the traditional CCA to construct local linear correlation models. We present results on both artificial data sets and Earth Science data sets to demonstrate that the proposed approach can detect useful correlation patterns, which traditional CCA fails to discover.

A new global land-water mask at 1-km resolution provides a more accurate delineation of ocean shorelines and large lake coastlines than the Earth observing system (EOS) digital elevation model (DEM) land-water mask. The mask, which is derived from the moderate resolution imaging spectrometer (MODIS) Nadir BRDF-adjusted reflectance (NBAR) and land cover products, is an update to the EOS DEM land-water mask currently used for MODIS processing. The new land-water mask exploits the superior geolocation accuracy of the Terra MODIS instrument and is based primarily on twelve months of reprocessed (V004) 1-km MODIS NBAR data from 2001 as processed by the MODIS land cover classification algorithm. Evaluation of the new mask reveals errors of omission in delineation of narrow rivers and small lakes.

The main objective of this research is to improve understanding of the methodological and validation requirements for mapping urban land cover over large areas from coarse resolution remotely sensed data. A technique called boosting is used to improve supervised classification accuracy and provides a means to integrate MODIS data with the DMSP nighttime lights data. Results indicate that fusion of these two data types improves urban classification results by resolving confusion between urban and other classes that occurs when either of the data sets is used alone. Traditional measures of accuracy assessment as well as new, maplet-based methods demonstrate the effectiveness of the methodology for creating maps of cities at continental to global scales.

Independent component analysis (ICA) provides a powerful new method to spatially and temporally deconvolve image sequences into components that capture variability arising from independent physical sources. To do this, ICA uses information contained in higher order cross-moments of multivariate data. We use remotely sensed time series of the normalized difference vegetation index to illustrate the utility of this technique.

In recent years, machine learning and data mining methods have become increasingly common in remote sensing applications. One area in which such techniques are particularly useful is classification of remotely sensed data for land cover and vegetation mapping applications. In this paper, we describe new methods to include available information (domain knowledge) in supervised classification of land cover using high dimensional remote sensing observations. specifically, land cover and vegetation classification schemes are generally designed for ecological or land use applications. As a result, the classes of interest are often poorly separable in the multi-spectral or multi-temporal feature space provided by remote sensing. In many cases, ancillary data sources can provide useful information to help distinguish between problematic classes. However, available methods for including ancillary data sources, such as the use of prior probabilities in maximum likelihood classification, are often problematic in practice. This paper presents a method for incorporating prior probabilities in remote sensing-based land cover classification using a supervised decision tree classification algorithm. The method exploits recent theory from the domain of statistics and machine learning that allows robust estimates of class membership to be estimated using a technique known as boosting. This approach allows poorly separable classes to be distinguished based on ancillary information, but does not penalize rare classes.

Data from MODIS provide a rich source of information for vegetation and land cover mapping applications. In particular, multi-spectral and multi-temporal information supplied by MODIS allows a variety of vegetation properties to be mapped and monitored at global scales in a fashion that was not previously possible. We describe the algorithms and databases being used to produce the MODIS global land cover parameter, focusing particular attention on the suite of vegetation properties that are being mapped from MODIS data as part of this product. Specifically, we describe a set of internally consistent maps, each depicting different vegetation attributes that have wide utility for studies of ecosystem dynamics at continental to global scales. The set of parameters being mapped include permanence of above ground biomass, leaf longevity, leaf type, and phenologic dynamics, among others. The methodology used to create these maps exploits a database of site data that includes detailed information related to ecosystem properties at global scales. To illustrate the utility and information content of the maps being produced from MODIS data, specific examples at regional scales that include a range of climate and ecosystem types are presented.

We investigated methods to identify phenological transition dates and the linkage between phenology and spatial climatic variation in North America between 30.5/spl deg/N and 70/spl deg/N utilizing Moderate-Resolution Imaging Spectroradiometer (MODIS) data. The results show a clear trend in which vegetation greenup gradually migrated northward from 38/spl deg/N to 70/spl deg/N starting in late March, and dormancy spread southward from late September. The observed pattern in phenology was significantly associated with both mean annual temperature and incoming solar radiation. Spatial analysis revealed that the date of greenup advanced by roughly 2.4 days, while the onset of dormancy was delayed by 1.8 days, for each one-degree increase in mean annual temperature.

A time series of globally distributed spectral MODIS NBARs (Nadir Bidirectional-reflectance-distribution-function Adjusted surface Reflectance) was used to determine initial tasseled cap coefficients. An assessment of an annual time series of tasseled cap features indicated their utility for detecting vegetation phenological cycles. The comparison analysis showed that the temporal pattern of NBAR greenness was closely correlated with the Enhanced Vegetation Index (EVI), while NBAR brightness matched MODIS global broadband albedos. Thresholded global NBAR wetnesses appear to relate to MODIS snow and ice presence as determined by the Normalized Difference Snow Index (NDSI).

Conventional approaches to accuracy assessment for land cover maps produced from remote sensing use either confusion matrices or the Kappa statistic to quantify map quality. These approaches yield global or class-specific measures of map quality by comparing classification results with independent ground-truth data. In most maps, considerable spatial variation is present in the accuracy of land cover labels that is not captured by these statistics. To date, this issue has rarely been addressed in the land cover remote sensing literature. The authors present a method to estimate pixel-scale land cover classification confidence using nonparametric machine learning methods. The method is based on recent theoretical developments from the domains of statistics and machine learning that explain the machine learning technique known as "boosting" as being equivalent to additive logistic regression. As a result, results from classification algorithms that use boosting can be assigned classification confidences based on probability estimates assigned to them using this theory. they test this approach using three different data sets. Their results demonstrate that classification errors tend to have low classification confidence while correctly classified pixels tend to have higher confidence. Thus, the method described in this paper may be used as a basis for providing spatially explicit maps of classification quality. This type of information will provide substantial additional information regarding map quality relative to more conventional quality measures and should be useful to end-users of map products derived from remote sensing.

We present an algorithm to retrieve land surface temperatures from wideband thermal infrared measurements using the model of Li et al. (1999). Forward simulation and inversion demonstrates the method to be stable in the presence of observation noise. Results from inversions performed using field measurements show that errors are generally less than the uncertainty in the observations.

This paper presents a new approach to identifying and eliminating mislabeled training instances for supervised learning. The goal of this approach is to improve classification accuracies produced by learning algorithms by improving the quality of the training data. Our approach uses a set of learning algorithms to create classifiers that serve as noise filters for the training data. We evaluate single algorithm, majority vote and consensus filters on five datasets that are prone to labeling errors. Our experiments illustrate that filtering significantly improves classification accuracy for noise levels up to 30 percent. An analytical and empirical evaluation of the precision of our approach shows that consensus filters are conservative at throwing away good data at the expense of retaining bad data and that majority filters are better at detecting bad data at the expense of throwing away good data. This suggests that for situations in which there is a paucity of data, consensus filters are preferable, whereas majority vote filters are preferable for situations with an abundance of data.

Classification of land cover from remotely sensed data at continental to global scales requires sophisticated algorithms and feature selection techniques to optimize classifier performance. The authors examine methods to maximize classification accuracies using decision trees to map land cover from multitemporal AVHRR imagery at continental and global scales. As part of their analysis they test the utility of "boosting", a new technique developed to increase classification accuracy by forcing the learning (classification) algorithm to concentrate on those training observations that are most difficult to classify. Their results show that boosting consistently reduces misclassification rates by 20-50% depending on the data set in question, and that most of the benefit gained by boosting is achieved after seven boosting iterations. They also assess the utility of including phenological metrics and geographic position as additional features to the classification algorithm. They find that using derived phenological metrics produces little improvement in classification accuracy relative to using an annual time series of NDVI data, but that geographic position provides substantial power for predicting land cover types at continental and global scales. However, in order to avoid generating spurious classification accuracies using geographic position, training data must be distributed evenly in geographic space.

The conventional definition of emissivity requires the source of radiation to be isothermal in order to compare its thermal emission to that of a blackbody at the same temperature. This requirement is not met for most land surfaces considered in thermal infrared remote sensing. Thus, the effective or equivalent emissivity of nonisothermal surfaces has been a poorly defined but widely used concept for years. Recently, several authors have attempted to define this concept more clearly. Unfortunately, definitions such as ensemble emissivity (e-emissivity) and emissivity derived from the surface bidirectional reflectance distribution function (r-emissivity), J. Norman et al. (1995), do not fully satisfy current needs for estimating true land surface temperature (LST). The present authors suggest the use of an additional term, the "apparent emissivity increment", which considers the effects of geometric optics to explain the directional and spectral dependence in LST caused by the three-dimensional (3D) structure and subpixel temperature distribution of the surface. They define this quantity based upon the emissivity derived from the bidirectional reflectance distribution function (/spl epsi//sub BRDF/) for isothermal surfaces and present a conceptual model of thermal emission from nonisothermal land surfaces. Their study also indicates that an average LST corresponding to the hemispherical wideband /spl epsi//sub BRDF/ Will be useful in remote sensing-based LST modeling and inversion.