Multi-temporal Methods Research Articles

Convolutional Neural Network-Driven Improvements in Global Cloud Detection for Landsat 8 and Transfer Learning on Sentinel-2 Imagery

A stable and reliable cloud detection algorithm is an important step of optical satellite data preprocessing. Existing threshold methods are mostly based on classifying spectral features of isolated individual pixels and do not contain or incorporate the spatial information. This often leads to misclassifications of bright surfaces, such as human-made structures or snow/ice. Multi-temporal methods can alleviate this problem, but cloud-free images of the scene are difficult to obtain. To deal with this issue, we extended four deep-learning Convolutional Neural Network (CNN) models to improve the global cloud detection accuracy for Landsat imagery. The inputs are simplified as all discrete spectral channels from visible to short wave infrared wavelengths through radiometric calibration, and the United States Geological Survey (USGS) global Landsat 8 Biome cloud-cover assessment dataset is randomly divided for model training and validation independently. Experiments demonstrate that the cloud mask of the extended U-net model (i.e., UNmask) yields the best performance among all the models in estimating the cloud amounts (cloud amount difference, CAD = −0.35%) and capturing the cloud distributions (overall accuracy = 94.9%) for Landsat 8 imagery compared with the real validation masks; in particular, it runs fast and only takes about 41 ± 5.5 s for each scene. Our model can also actually detect broken and thin clouds over both dark and bright surfaces (e.g., urban and barren). Last, the UNmask model trained for Landsat 8 imagery is successfully applied in cloud detections for the Sentinel-2 imagery (overall accuracy = 90.1%) via transfer learning. These prove the great potential of our model in future applications such as remote sensing satellite data preprocessing.

Mapping of Burnt surfaces in the Context of Protected Areas in the Sudanian Savannahs of Northern Cameroon: Contribution of Time Series of Sentinel-2 Images

In Africa, bushfires are common practices in savannah regions. Their impact, is the subject of many scientific debates. The protected areas, of North Cameroon are subjected to ambivalent effects of these fires. Using high-resolution remote sensing data, this work aims at assessing burned areas at the interannual and intraannual scale in order to explain the complex environmental dynamics of landscapes, to contribute to fire management in a context of protected areas. To achieve this, time series of sentinel-2 images from the Multi Spectral Instrument from the year 2015-2020 have been used. By the unitemporal and multitemporal methods, the development of spectral indices has made it possible to locate the burned surfaces and to determine the periods of the fires. Observations and field surveys aimed to understand the factors of fires using, their typology as well as the roles of the actors. The findings show that each spectral index has its own ability to detect burnt surfaces: The NDVI can detect fires only at the start of the dry season, the SAVI is suitable for identifying intermediate and late fires, while the NBR is best suited for separating burnt and unburnt surfaces throughout the season, regardless of the method employed. Depending on their periods of occurrence, the fires spread more in parks with a fairly extensive grass cover compared to overexploited ordinary areas. They then depend on socio-environmental factors, namely the state of land used, the continuity of the herbaceous cover and the social practices in place. The mapping and monitoring of burned surfaces by remote sensing therefore constitute a tool and a method for diagnosing the state of the herbaceous plant cover and for managing natural resources.

SENTINEL-2 SURFACE REFLECTANCE PRODUCTS GENERATED BY CNES AND DLR: METHODS, VALIDATION AND APPLICATIONS

Abstract. To allow for a robust and automatic exploitation of Sentinel-2 data, Analysis Ready Data (ARD) products are requested by most users. The processors of ARD products take care of the common burdens necessary for most applications, that include precise orthorectification, cloud detection and atmospheric correction steps, as well as the generation of periodic syntheses of cloud free surface reflectances. The French Theia land data center, and the German Earth Observation Center (EOC) started delivering Sentinel-2 surface reflectance products to users in 2016 in France and 2019 in Germany respectively. Both centers produce and distribute these data sets in near real time, over large regions requested by French users such as Western Europe, Maghreb, Sahel, Madagascar… Theia’s and EOC products include an instantaneous surface reflectance product (Level-2A), and a monthly cloud free synthesis of surface reflectance (Level-3A). This article shortly describes the methods used to generate the Level-2A products with the MAJA processor, and the Level-3A products with theWASP processor. The MAJA processor is based on multi-temporal methods, that use the slow variation of surface reflectance to detect clouds and estimate aerosol depth, while WASP, thanks to the quality of MAJA cloud mask, calculates a weighted average of all the cloud free observations over 45 days, every month. The article also provides validation results for Level-2A and Level-3A products, resulting from comparison with in-situ data and with other methods. A last section gives first insights from the monitoring of user uptake of the distributed products.

Recent advancements in multi-temporal methods applied to new generation SAR systems and applications in South America

Detection and continuous monitoring of Earth's ground surface changes, triggered by natural phenomena or induced by human activities, is nowadays possible using Earth Observation (EO) technologies. Indeed, the exploitation of remotely sensed data collected by constellations of new-generation satellite platforms, complemented with in-situ measurements and ground-based observation systems, represents a well-established practice to get valuable information on Earth's crust and subsurface dynamics. The effects of extreme natural or man-induced events (e.g., earthquakes, volcanic eruptions, flooding phenomena, sea-level rise, big fires, etc.) have severe societal and economic impacts. In particular, the technologies based on the use of Synthetic Aperture Radar (SAR) images reached significant improvements in the last decade due to the growing availability of vast amounts of data collected by multiple-satellite sensors operating at different frequency bands and with complementary viewing angles, polarization and acquisition modes. Accordingly, to process a large amount of SAR data in a timely fashion, up-to-date high-performance computing (HPC) methods and tools are required. This paper addresses the state-of-the-art of SAR technologies for the analysis of long sequences of multiple sets of SAR images and provides a perspective on the forthcoming improvements of these technologies. In particular, the emphasis is placed on novel interferometric SAR and change detection methods, giving an overview of how those processing techniques have been used for investigating sites located in South and Central America. Moreover, an overview of the new generation of SAR sensors' observational capability, especially in the field of ground deformation analysis for mitigating the risk associated with natural and human-induced hazards, is provided. COSMO-SkyMed, ALOS, Sentinel-1, and SAOCOM data are exploited to show how natural and human-induced terrain displacement phenomena can be detected and investigated in different portions (X-, L- and C-band) of the microwave spectrum using SAR technologies.

Thick Clouds Removing From Multitemporal Landsat Images Using Spatiotemporal Neural Networks

Landsat images have played an important role in the field of Earth observation and geoinformatics. However, optical Landsat images are frequently contaminated by cloud cover, especially in tropical and subtropical regions, which limits the utilization of these images. To improve the utilization of Landsat images, in this study, we propose a novel spatiotemporal neural network with four modules: a cloud detection module, a spatial–temporal learning module, a spatial–temporal feature fusion module, and a reconstruction module. The results of the experiments demonstrate that the proposed method is quantitatively effective (root mean square error < 0.0179) and can achieve a better result for reconstructing Landsat images than some of the widely used existing deep learning methods and multitemporal methods. The proposed neural network method provides an effective tool for the removal of contiguous, thick clouds from satellite images, so as to improve the quality of subsequent remote sensing mapping and geoinformation extraction.

Episodic ground deformation associated with geothermal energy production at the Guayabo Caldera, Costa Rica

The ground response to the production of energy in geothermal fields can be observed in the form of ground displacement and seismic activity. Changes in the temperature, volume, pressure of the reservoir can culminate in surface deformation and in seismicity at the reservoir edges and/or near the production wells. Identifying the origin of the ground response enables controlled management of the reservoir and can help prevent damage to structures. Satellite remote sensing (InSAR) permits the study of the deformation at these sites with high spatial and temporal resolution. However, Central America has a dynamic atmosphere with considerable water vapour and constant growth of vegetation all year long, which leads to decorrelation between SAR images. Here we study the Dr. Alfredo Mainieri Protti geothermal field, located near Miravalles Volcano in the Northern Costa Rica volcanic arc. We measure the deformation associated with geothermal energy production using data from two satellites (ALOS-2 and Sentinel-1) between January 2015 and January 2017. The ALOS-2 data was used to obtain the extent of the affected area, and two StaMPS multi-temporal methods (Persistent Scatter (PS) InSAR, and Small Baselines(SB)) were applied to the Sentinel-1 data, to obtain the temporal pattern of the ground displacements. Results show a direct response of the ground to the extraction of fluids. The pattern of deformation obtained with ALOS-2 is spatially correlated with the geothermal field, and with the recorded seismicity. Between April 2015 to October 2016, periods of uplift and subsidence were observed with values between ~2.0 to 3.0 cm every 120 days, linked to changes in the amount of fluids extracted from the reservoir. The results also show the ability of the reservoir to recover rapidly following a decrease in extraction rate. This work has shown that even in this tropical region it is possible to monitor anthropogenic ground motion using InSAR. This is beneficial for the Central American region, as satellite images are becoming more frequently acquired and available, and can be used to boost the monitoring of seismic regions, volcanic activity and anthropogenic activities.

Stochastic ensemble methods for multi-SAR-mission soil moisture retrieval

The recent and projected investments across the world on radar satellite missions (e.g., Sentinel-1, SAOCOM, BIOMASS and NISAR) provide a great opportunity for operational radar soil moisture mapping with high spatial and temporal resolution. However, there is no retrieval algorithm that can make complementary use of the multi-frequency data from those missions, due to the large uncertainties in observations collected by the different sensors, different validity regions of the forward models, and the fact that inversion algorithms have been designed for specific data sources. In this study, the principle of ensemble learning was introduced to provide two general soil moisture retrieval frameworks accounting for these issues. Instead of trying to find an optimal global solution, multiple soil moisture retrievals (termed sub-retrievals) with moderate performance were first obtained using different channels and/or time instances randomly selected from the available data, with the retrieved ensemble of results being the final output. The ensemble retrievals, taking one existing snapshot method and two multi-temporal methods as the base retrieval algorithms, were evaluated using a synthetic data set with the effectiveness confirmed under various uncertainty sources. An evaluation using the Fifth Soil Moisture Active Passive Experiment (SMAPEx-5) data set showed that the ensemble retrieval outperformed the non-ensemble retrieval in most cases, with a decrease of 0.004 to 0.014 m3/m3 in Root Mean Square Error (RMSE) and an increase of 0.01 to 0.16 in correlation coefficient (R). Weakly biased and correlated sub-retrievals were confirmed to be the basic requirement of an effective ensemble retrieval, being consistent with use of ensemble learning in other applications.

Application of Time-Series Sentinel-1A for Land Deformation in Central Aceh, Indonesia

Several decorrelation phenomena of Interferometric Synthetic Aperture Radar (InSAR) have led researchers to develop various multitemporal InSAR (MT-InSAR) techniques with the application of time series/stack of images. In this study, we present the land surface movement monitoring using MT-InSAR techniques in the Central of Aceh-Indonesia, and focus on the temporal and spatial pattern of uplift and subsidence by using multi InSAR methods such as Quasi-Persistent Scatterer (Q-PS) and Small Baseline Subset (SBAS). A total of 18 scenes of Sentinel-1A(S-1A) and 14 scenes of ALOS PALSAR-1(PALSAR-1) images were acquired between 2018 and 2019, as well as 2007 and 2010, where then the multitemporal methods and techniques were applied sequentially to a set of those data. The results showed that the either the S-1A and PALSAR-1 velocity subsidence at Nunang and Musara Alun villages were range from 2.4 to 5.7 and 0.6 to 2.3 mm/year, respectively which corresponded to the results obtained by other research publication. The deformation in Central Aceh needs continuous monitoring using InSAR as the place is a landslide prone area because of the nature of the terrain.

BIG DATA PROCESSING AND ANALYSIS USING MULTI-TEMPORAL SENTINEL 1 DATA TO MANAGE FLOODS IN NORTH OF MOROCCO

Abstract. Floods are the natural hazards that produce the highest number of casualties and material damage in the Western Mediterranean, especially in Morocco. An improvement in flood risk assessment and study of a possible increase in flooding occurrence are therefore needed. Earth Observation big data such as the ones acquired by the Copernicus programme are providing unprecedented opportunities to detect changes and assess economic impacts in case of disasters. This article present the different results obtained by the multi-temporal methods using the Synthetic Aperture Radar images. The spaceborne Synthetic Aperture Radar (SAR) systems are suitable tools for flood mapping thanks to their daytime and nighttime and almost all-weather imaging capability, in addition to their sensitivity to surface roughness and to Flood monitoring. The method has been developed to exploit Sentinel-1 data. It has been tested for the 2018 flood of Tetouan (Morocco).

Soil moisture retrieval from time series multi-angular radar data using a dry down constraint

Multi-angular and multi-temporal methods have been developed and accepted as two promising strategies for reliable soil moisture retrieval from radar data. However, the way to combine time series multi-angular data acquired from both descending and ascending orbits with different imaging modes (e.g., ScanSAR and Stripmap) remains unresolved. Consequently, a multi-temporal algorithm is proposed for soil moisture retrieval at the pixel – paddock scale (25–500 m) using time series multi-angular L-band (1.26 GHz) radar data. The method assumes time-invariant roughness and vegetation for the retrieval period together with a soil moisture dry-down constraint for noise reduction, while utilizing multi-angular data without incidence angle normalization. The Numerical Maxwell Model of Three-Dimensional simulations and distorted Born approximation (NMM3D-DBA) were used to build a set of landcover specific multi-angular look up tables (LUTs). Effective isotropic roughness values were assumed suitably able to account for the periodic features in cultivated surfaces, with values determined as part of the soil moisture retrieval. A genetic algorithm was used to minimize the difference between LUTs and time series multi-angular radar observations with the dry-down constraint. Evaluation based on the Fifth Soil Moisture Active Passive Experiment dataset (SMAPEx-5) has shown an acceptable overall root mean square error (RMSE) of 0.070 m3/m3 at the 25-m pixel scale and 0.056 m3/m3 at the paddock (field) scale (~0.1–0.5 km). Further investigations on the effect of polarization combination and time interval of radar data have confirmed the effectiveness of the proposed method for irregularly collected data with different imaging modes.

A cloud detection algorithm for satellite imagery based on deep learning

Reliable detection of clouds is a critical pre-processing step in optical satellite based remote sensing. Currently, most methods are based on classifying invidual pixels from their spectral signatures, therefore they do not incorporate the spatial patterns. This often leads to misclassifications of highly reflective surfaces, such as human made structures or snow/ice. Multi-temporal methods can be used to alleviate this problem, but these methods introduce new problems, such as the need of a cloud-free image of the scene. In this paper, we introduce the Remote Sensing Network (RS-Net), a deep learning model for detection of clouds in optical satellite imagery, based on the U-net architecture. The model is trained and evaluated using the Landsat 8 Biome and SPARCS datasets, and it shows state-of-the-art performance, especially over biomes with hardly distinguishable scenery, such as clouds over snowy and icy regions. In particular, the performance of the model that uses only the RGB bands is significantly improved, showing promising results for cloud detection with smaller satellites with limited multi-spectral capabilities. Furthermore, we show how training the RS-Net models on data from an existing cloud masking method, which are treated as noisy data, leads to increased performance compared to the original method. This is validated by using the Fmask algorithm to annotate the Landsat 8 datasets, and then use these annotations as training data for regularized RS-Net models, which then show improved performance compared to the Fmask algorithm. Finally, the classification time of a full Landsat 8 product is 18.0 ± 2.4 s for the largest RS-Net model, thereby making it suitable for production environments.

Visual Analysis of Satellite Landsat Images Multitemporal and GPS as a Geographic Information System for Mapping of Nugmet Plantations in Tapaktuan

Geographical information systems (GIS) are needed in planning and controlling land use as a basis for anticipating policy decisions in planning and structuring Tapaktuan territory. This study seeks to integrate GPS technology and remote sensing technology products such as Landsat-8 OLI satellite imagery, Landsat-7 ETM + using multitemporal methods using the supervised (Maximum Likehood) classification method. Based on the analysis of the processing of satellite images of nutmeg plantation area for several years, there has been an increasing spread of land use for nutmeg plantations.

An improved algorithm for mapping burnt areas in the Mediterranean forest landscape of Morocco

The identification of burnt forests and their monitoring provide essential information for the suitable management and conservation of these ecosystems. This research focuses on the use of remote sensing with MODIS sensor data in a Mediterranean environment, precisely in the Rif region known for its high occurrence of forest fires and the largest burnt areas in Morocco. It mapped the burnt areas during the summer of 2016 using spectral indices from MODIS images, namely the Normalized Burn Ratio (NBR) and the Burnt Area Index for MODIS (BAIM). Two field surveys were used to calibrate spectral indices and validate the maps. First, a monotemporal analysis using a single pre-fire image determined the appropriate threshold of the spectral indices (BAIM and NBR) for burn detecting. Secondly, a multitemporal method was applied based on dBAIM and dNBR images which represented pre-fire and postfire differences of the BAIM and NBR images, respectively. The results show that separate use of monotemporal postfire and multitemporal methods produced an overestimation of the burnt areas. Finally, we propose a new algorithm combining both methods for burnt area mapping that we name Burnt Area Algorithm. MCD45A1 and MCD64A1 MODIS burnt area products were compared to the proposed algorithm. Validation of the estimated burnt areas using reference data of the Moroccan High Commission for Water, Forests and Fight against Desertification showed satisfactory results using the proposed algorithm, with a determination coefficient of 0.68 and a root mean square error of 44.0 ha.

The progress of operational forest fire monitoring with infrared remote sensing

Forest wildfires pose significant and growing threats to human safety, wildlife habitat, regional economies and global climate change. It is crucial that forest fires be subject to timely and accurate monitoring by forest fire managers and other stake-holders. Measurement by spaceborne equipment has become a practical and appealing method to monitor the occurrence and development of forest wildfires. Here we present an overview of the principles and case studies of forest fire monitoring (FFM) with satellite- and drone-mounted infrared remote sensing (IRRS). This review includes four types of FFM-relevant IRRS algorithms: bi-spectral methods, fixed threshold methods, spatial contextual methods, and multi-temporal methods. The spatial contextual methods are presented in detail since they can be applied easily with commonly available satellite IRRS data, including MODIS, VIIRS, and Landsat 8 OLI. This review also evaluates typical cases of FFM using NOAA-AVHRR, EOS-MODIS, S-NPP VIIRS, Landsat 8 OLI, MSG-SEVIRI, and drone infrared data. To better implement IRRS applications in FFM, it is important to develop accurate forest masks, carry out systematic comparative studies of various forest fire detection systems (known as forest fire products), and improve methods for assessing the accuracy of forest fire detection. Medium-resolution IRRS data are effective for landscape-scale FFM, and the VIIRS 375 m contextual algorithm and RST-FIRES algorithm are helpful for closely tracking forest fires (including small and short-lived fires) and forest-fire early warning.

Cloud masking of MODIS imagery based on multitemporal image analysis

We present multitemporal Bayesian classification of Moderate Resolution Imaging Spectroradiometer (MODIS) mosaic data over the Caspian Sea during winter. The multitemporal analysis methods used were cross-correlation and motion detection based on phase correlation. Our motion estimation algorithm estimates the motion of a target between two adjacent images over the same areas based on finding the maximum correlation with respect to location shift within a given area around each location. The motion detection algorithm also provides a quality estimate for the detection. Because sea ice, unlike clouds, is typically rigid and its motion is significantly slower than the motion and metamorphosis of clouds, drifting sea ice can be distinguished from clouds. Over land and static sea ice, detection of clouds, is easier because the cross-correlation is typically higher for land and ice than for clouds. Also, locating clouds over open water is straightforward because clouds appear significantly brighter than open water. The results show that multitemporal features can be used to distinguish between clouds and clear sky.

Airborne multi-temporal L-band polarimetric SAR data for biomass estimation in semi-arid forests

Using the airborne Polarimetric L-band Imaging Synthetic aperture radar (PLIS) the impact of high revisit cycle and full polarimetric acquisitions on biomass retrieval was investigated by means of backscatter-based multi-temporal methods. Parametric and non-parametric models were used to relate reference biomass levels obtained from field plot measurements and high point density lidar data to backscatter intensities or polarimetric target decomposition components. Single-date retrieval using multiple independent variables provided lower estimation errors when compared to models using one independent variable with errors decreasing by 2% to 15%. The multi-temporal aggregation of daily biomass estimates did not improve the overall retrieval accuracy but provided more reliable estimates with respect to single-date methods. Backscatter intensities improved estimation accuracies up to 10% compared to polarimetric target decomposition components. Using all four polarizations increased the estimation accuracy marginally (2%) when compared to a dual-polarized system. The biomass estimation error was considerably reduced (up to 30%) only by decreasing the spatial resolution and was related to decreasing forest variability with increasing pixel size. These results indicate that, at least in semi-arid areas, future L-band missions would not significantly improve biomass estimation accuracy using backscatter-based modeling approaches despite their better spatial resolution, higher revisit cycles and the availability of fully polarimetric information.

Observing the Middle Elbe Biosphere in Germany by Means of TerraSAR-X Images

The Lower Saxonian Elbe Valley Biosphere Reserve is part of the UNESCO Biosphere Reserve “Elbe River Landscape”, and used mainly for agriculture. One of tasks of the Biosphere Reserve Administration is to develop sustainable forms of land use which requires comprehensive updated land cover maps. Land use maps are hard to produce because of surveying costs and time. Nevertheless, these large areas need to be monitored. TerraSAR-X images are used to establish agricultural land use maps. In this study, two areas are selected within the Elbe Biosphere Reserve situated around the oxbows Wehninger Werder and Walmsburger Werder. Multi temporal classification methods were used to identify the different crops using maximum likelihood classifier for the years 2010 and 2011. The crop classifications were used to evaluate the effect of the number of images, the necessity of polarizations, and the consequences of some missing images within the crop calendar. These classifications were analyzed to estimate producer accuracy and Kappa index for each crop besides the overall accuracy for each agricultural land use map. The study shows that using dual polarization imagery enhances producer accuracies for many crops over the single polarization imagery, and demonstrates the importance of using frequent images during the cultivation period.

A generalized fuzzy Markov chain-based model for classification of remote-sensing multitemporal images

In this work we propose a conceptual generalization of the cascade classification, fuzzy Markov chain-based method introduced in earlier studies. Such generalization, which is based on the assumption of the invertibility of the fuzzy Markov classification model with respect to time, leads to a model that can classify image objects at two points in time simultaneously. We start by defining two temporal modes of operation. In the forward mode a temporal transformation, supported by a transition possibility matrix T, projects an image object’s fuzzy classification for time t into time t + 1 and fuses the updated membership values with the object’s fuzzy classification for time t + 1. In the backward mode the transition matrix is inverted and the fuzzy classification for time t + 1 is updated backwards, i.e. projected into time t. Furthermore, we tackle a key problem with respect to the application of fuzzy Markov chains in remote-sensing data classification: the estimation of transition possibility values. Previously, transition possibilities estimation in the context of fuzzy Markov chain-based multitemporal classification methods has been carried out with the aid of stochastic methods – specifically, through genetic algorithms. In this work we propose an analytical, least squares-based estimation technique, as a more stable and computational efficient alternative to the stochastic approach. Finally, we report on the application of the multitemporal method in the classification of two different test sites – rural and urban – covered by images produced by medium and high resolution orbital, optical sensor systems.

GeoDMA—Geographic Data Mining Analyst

Remote sensing images obtained by remote sensing are a key source of data for studying large-scale geographic areas. From 2013 onwards, a new generation of land remote sensing satellites from USA, China, Brazil, India and Europe will produce in 1year as much data as 5 years of the Landsat-7 satellite. Thus, the research community needs new ways to analyze large data sets of remote sensing imagery. To address this need, this paper describes a toolbox for combing land remote sensing image analysis with data mining techniques. Data mining methods are being extensively used for statistical analysis, but up to now have had limited use in remote sensing image interpretation due to the lack of appropriate tools. The toolbox described in this paper is the Geographic Data Mining Analyst (GeoDMA). It has algorithms for segmentation, feature extraction, feature selection, classification, landscape metrics and multi-temporal methods for change detection and analysis. GeoDMA uses decision-tree strategies adapted for spatial data mining. It connects remotely sensed imagery with other geographic data types using access to local or remote database. GeoDMA has methods to assess the accuracy of simulation models, as well as tools for spatio-temporal analysis, including a visualization of time-series that helps users to find patterns in cyclic events. The software includes a new approach for analyzing spatio-temporal data based on polar coordinates transformation. This method creates a set of descriptive features that improves the classification accuracy of multi-temporal image databases. GeoDMA is tightly integrated with TerraView GIS, so its users have access to all traditional GIS features. To demonstrate GeoDMA, we show two case studies on land use and land cover change.

Multitemporal spectral analysis for cheatgrass (Bromus tectorum) classification

Operational satellite remote sensing data can provide the temporal repeatability necessary to capture phenological differences among species. This study develops a multitemporal stacking method coupled with spectral analysis for extracting information from Landsat imagery to provide species‐level information. Temporal stacking can, in an approximate mathematical sense, effectively increase the ‘spectral’ resolution of the system by adding spectral bands of several multitemporal images. As a demonstration, multitemporal linear spectral unmixing is used to successfully delineate cheatgrass (Bromus tectorum) from soil and surrounding vegetation (77% overall accuracy). This invasive plant is an ideal target for exploring multitemporal methods because of its phenological differences with other vegetation in early spring and, to a lesser degree, in late summer. The techniques developed in this work are directly applicable for other targets with temporally unique spectral differences.