Composite Time Series Research Articles

Time Series approach to map areas of Agricultural Plastic Waste generation

Abstract. The escalating presence of plastics in agriculture has raised concerns regarding Agricultural Plastic Waste (APW). Hitherto, a lack of comprehensive plasticulture data impedes effective waste management strategies, potentially resulting in plastic pollution and contributing to microplastic formation. APW locations and quantities are pivotal for territorial planning and the formulation of public policies on waste management and land use. Remote detection of agri-plastics has garnered increased consideration, particularly in mapping plastic-mulched farmlands (PMFs) dispersed in extensive regions. This study investigates whether the use of satellite image time series can map PMFs accurately. We assessed pixel-based classification using a 16-day composite time series of Sentinel-2 imagery (S2-16D) obtained from the Brazil Data Cube project. The Plastic-Mulched Landcover Index (PMLI) was joined into S2-16D bands because it is an index focused on PMF detection. Four classifiers (RF, MLP, L-TAE, and TCNN) were compared through agreed classification metrics. The most promising outcome showed an overall accuracy of 100% employing L-TAE but with visible noise in the map. The time series enhanced accuracy while minimizing background confusion, offering a viable solution for PMF detection. The PMF map presented herein represents an initial stride toward fostering circularity in plasticulture throughout South America.

An Efficient Algorithm for Exact Segmentation of Large Compositional and Categorical Time Series

ABSTRACTChange‐point detection, also known as signal segmentation, is an essential preprocessing step in many applications, ranging from industrial monitoring to bioinformatics. In short, it consists in finding the temporal boundaries of homogeneous regimes in long and non‐stationary time series. While this area of research is active, most existing methods are designed for Euclidean data. However, in many practical scenarios, the collected time series are compositional, meaning that each observation belongs to the probability simplex (the set of non‐negative vectors whose components sum to one). In this work, we propose an algorithm detecting change‐points in large compositional signals with an underlying piecewise stationary model. We cast the change‐point detection task as a discrete optimization problem, whose solution is shown to converge to the true change‐points. We introduce a new and time‐efficient dynamic programming algorithm that solves exactly this problem. To limit the number of operations, we describe a novel pruning rule that allows us to reduce the set of candidate change‐point indices. Our method is tested on a thorough simulation study, which confirms its efficiency. Additionally, we apply our method to a human activity segmentation task, highlighting the necessity for such novel techniques compared to standard algorithms.

Long-term investigation on the daily variability of microplastic concentration and composition - Monitoring in the effluent of a wastewater treatment plant

The discharge of treated wastewater into receiving waters by wastewater treatment plants (WWTPs) is considered as one of the main pathways for microplastics (MPs) to enter the environment. To gain a better understanding of the temporal variability of MPs in WWTP effluents, this study investigated the concentration and composition (size, shape and polymer type) of MPs in the effluent of a German WWTP over the course of one month. 24-hour mixed samples were collected daily by a custom built automated sampling unit for MPs. Particles and fibers ≥11 μm were analyzed using Fourier transform infrared microspectroscopy. The MP concentration showed large daily fluctuations and ranged between 9.64 103 m−3 and 8.44 104 m−3 MPs over the study period. However, there was no significant correlation between the MP concentration and the precipitation or discharge from the WWTP. In contrast to the MP concentration, the MP composition in terms of size and shape was consistent over the study period. There were strong correlations between the time series of the polymer types polypropylene, polyethylene, polystyrene, polyester, ethylene-vinyl-acetate, cellulose artificially modified as well as the polymer cluster acrylates/polyurethanes/varnish. The time series of polyamide showed no significant correlation with the time series of any other polymer type. This study established a one month long high-resolution time series of MP concentration and composition, and thus provides a valuable basis for future research on the temporal variability of MP inputs into the environment from WWTP effluents.

An adaptive composite time series forecasting model for short-term traffic flow

Short-term traffic flow forecasting is a hot issue in the field of intelligent transportation. The research field of traffic forecasting has evolved greatly in past decades. With the rapid development of deep learning and neural networks, a series of effective methods have been proposed to address the short-term traffic flow forecasting problem, which makes it possible to examine and forecast traffic situations more accurately than ever. Different from linear based methods, deep learning based methods achieve traffic flow forecasting by exploring the complex nonlinear relationships in traffic flow. Most existing methods always use a single framework for feature extraction and forecasting only. These approaches treat all traffic flow equally and consider them contain same attribute. However, the traffic flow from different time spots or roads may contain distinct attributes information (such as congested and uncongested). A simple single framework usually ignore the different attributes embedded in different distributions of data. This would decrease the accuracy of traffic forecasting. To tackle these issues, we propose an adaptive composite framework, named Long-Short-Combination (LSC). In the proposed method, two data forecasting modules(L and S) are designed for short-term traffic flow with different attributes respectively. Furthermore, we also integrate an attribute forecasting module (C) to forecast the traffic attributes for each time point in future time series. The proposed framework has been assessed on real-world datasets. The experimental results demonstrate that the proposed model has excellent forecasting performance.

Usable observations over Europe: evaluation of compositing windows for Landsat and Sentinel-2 time series

ABSTRACT Landsat and Sentinel-2 data archives provide ever-increasing amounts of satellite data. However, the availability of usable observations greatly varies spatially and temporally. Pixel-based compositing that generates temporally equidistant cloud-free synthetic images can mitigate temporal variability, by constructing uninterrupted time series using different compositing windows. Here, we evaluated the feasibility of using compositing windows ranging from five days to one year for 1984–2021 Landsat and 2015–2021 Sentinel 2 time series to derive uninterrupted time series across Europe. We considered separate and joint use of both data archives and analyzed the spatio-temporal availability of composites during each calendar year and pixel-specific growing season across a variety of time windows and hypothesizing data interpolation. Our results demonstrated opportunities and limitations in the available data records to support medium- and long-term analyses requiring uninterrupted time series of composites with sub-annual temporal resolution. Spatial disparities across different compositing windows provide guidance on the feasibility of workflows relying on different data densities and on the challenges in wall-to-wall analyses. The feasibility of consistent time series based on composites with sub-monthly aggregation periods was mostly limited to the combined Landsat and Sentinel-2 archives after 2015, yet in some geographies requires interpolation of up to 50% of data.

Look over there. Where? A compositional approach to the modeling of public opinion on the most important problem

AbstractObjectiveThis study aims to test whether the American public is polarized and/or parallel in its assessments of the most important problem.MethodsWe use compositional time series models and new data on public opinion to test for differences between subgroups.ResultsWe find inconsistent evidence of polarization for some issue areas but not others and remarkably robust evidence of parallel reactions across subgroups to economic and international shocks.ConclusionThe U.S. public is remarkably consistent in terms of its assessments of the most important problem and in how subgroups shift their perceptions of issue importance in reaction to changing circumstances.

Local reflects global: Life stage-dependent changes in the phenology of coastal habitat use by North Sea herring.

Climate warming is affecting the suitability and utilization of coastal habitats by marine fishes around the world. Phenological changes are an important indicator of population responses to climate-induced changes but remain difficult to detect in marine fish populations. The design of large-scale monitoring surveys does not allow fine-grained temporal inference of population responses, while the responses of ecologically and economically important species groups such as small pelagic fish are particularly sensitive to temporal resolution. Here, we use the longest, highest resolution time series of species composition and abundance of marine fishes in northern Europe to detect possible phenological shifts in the small pelagic North Sea herring. We detect a clear forward temporal shift in the phenology of nearshore habitat use by small juvenile North Sea herring. This forward shift might be linked to changes in water temperatures in the North Sea. We next assessed the robustness of the effects we found with respect to monitoring design. We find that reducing the temporal resolution of our data to reflect the resolution typical of larger surveys makes it difficult to detect phenological shifts and drastically reduces the effect sizes of environmental covariates such as seawater temperature. Our study therefore shows how local, long-term, high-resolution time series of fish catches are essential to understand the general phenological responses of marine fishes to climate warming and to define ecological indicators of system-level changes.

Research on Multistate Variable Prediction of a Master‐Slave Electric‐Hydraulic Hybrid Vehicle Based on CNN‐LSTM Model with Attention Mechanism

This paper utilizes a convolutional neural network and long short‐term memory with attention mechanism (CLA) prediction model for time series forecasting. An innovative master‐slave electric‐hydraulic hybrid vehicle (MSEHV) is introduced. It simplifies the drivetrain structure by utilizing planetary gears for energy transfer and coupling to transform electrical, mechanical, and hydraulic energy. This paper conducts simulation verification using existing control strategies. It combines the velocity time series information with the hydraulic accumulator pressure time series information to create a unified time series. Some of the more popular machine learning and neural network algorithms are considered, and the mean square error, mean absolute error, and coefficient of determination evaluation metrics are calculated for validation. The research presents a developed battery state of charge regression prediction model based on genetic algorithm‐backpropagation by combining the predicted results of composite time series with the current discharge depth of the battery. The validation results indicate that the CLA time series prediction model proposed in this paper for MSEHV has higher prediction accuracy compared to other prediction models.

Multi-rule combination prediction of compositional data time series based on multivariate fuzzy time series model and its application

Compositional data (CoDa) is a type of relative data reflecting the proportion of different units in an integrated system, which can reveal the internal structure of observed objects and has been widely used to analyze practical economic and management problems. However, the summation constraint limits the prediction of compositional data time series. Especially, with small samples and poor information characteristics of the data set, the difficulties are further exacerbated. Compared with real-valued time series analysis, fuzzy time series (FTS) mainly transforms initial observed data to be different status so that the historical evolution rules can be derived. As a result, forms of initial observed data can be concealed. Correspondingly, this paper aims to establish a multi-rule combination prediction of compositional data based on a multivariate fuzzy time series model, named GA-CoDa-MFTS. Firstly, with the interaction of components, the adjustment rules between components are explored by using the fuzzy logical relationship (FLR). Secondly, the Genetic algorithm (GA) is used to find the optimal weight of the predicted values under different rules and derive the combined predicted values. Finally, the prediction effectiveness of the model is verified by CRMSE and CMAPE through a numerical study.

Ordinal compositional data and time series

There are several real applications where the categories behind compositional data (CoDa) exhibit a natural order, which, however, is not accounted for by existing CoDa methods. For various application areas, it is demonstrated that appropriately developed methods for ordinal CoDa provide valuable additional insights and are, thus, recommended to complement existing CoDa methods. The potential benefits are demonstrated for the (visual) descriptive analysis of ordinal CoDa, for statistical inference based on CoDa samples, for the monitoring of CoDa processes by means of control charts, and for the analysis and modelling of compositional time series. The novel methods are illustrated by a couple of real-world data examples.

Seasonal dynamics of phytoplankton community assembly at the Blanes Bay Microbial Observatory (BBMO), NW Mediterranean Sea

The dynamics of phytoplankton biomass and community composition is important for the functioning of marine ecosystems and ocean biogeochemical cycles. However, there is a shortage of studies addressing the interannual seasonal patterns of phytoplankton community assembly due to sampling limitations. Here we study the seasonal dynamics of eight major phytoplankton groups over a 12 year period (2006 to 2018) using a time-series of taxonomic composition from the Blanes Bay Microbial Observatory (BBMO) in the North Western Mediterranean Sea: dinoflagellates, diatoms, coccolithophores, Prochlorococcus, Synechococcus, picoeukaryotes, nanoeukaryotes, and photosynthetic nanoflagellates. We combine the analysis of biotic factors (primary production, phytoplankton taxa, cell abundance, cell size, chlorophyll-a concentration, and phytoplankton biomass) and abiotic factors (nutrients, temperature, and irradiance) to provide a coherent picture of the observed seasonal patterns of phytoplankton community assembly. The BBMO ecosystem is seasonally heterogeneous in community composition, displaying large fluctuating alternations in phytoplankton group dominance throughout the year. The seasonal succession of phytoplankton groups tends to repeat itself every year in a regular fashion, being the seasonal variability of the phytoplankton groups larger than their interannual variability. We compute α-diversity, a measure of the effective richness of phytoplankton groups. The seasonality of α-diversity shows that it is lowest during winter and highest during summer. We compute temporal β-diversity, a measure of compositional heterogeneity of the phytoplankton community. The data show a sinusoidal behavior of β-diversity as a function of the temporal distance between samples, with a period of one year. We use the mirror index (1 - β-diversity) at a temporal distance of one month to compute the phytoplankton group turnover. The seasonality of turnover shows that it is highest during spring and autumn. To evaluate the validity of niche and neutral theories in predicting the interannual sinusoidal behavior of β-diversity, we performed numerical simulations using a mechanistic model. The results provide support to the niche theory for marine phytoplankton ecology and community assembly. The phytoplankton groups appear to follow their specific ecological niches, tracking the seasonal changes in environmental conditions. The ecological implications of these findings are that marine phytoplankton groups appear to fill distinct environmental niches and thus may have different functional roles in the ecosystem. Furthermore, they may be predictable using both mechanistic and species distribution modeling approaches. The climatological time-series presented here can be an excellent testing ground for evaluating the performance of these marine ecosystem models having an explicit representation of different phytoplankton groups.

Assessing Seasonal and Inter‐Annual Marine Sediment Climate Proxy Data

AbstractThree recently published papers including Napier et al. (2022, https://doi.org/10.1029/2021PA004355) utilize novel microanalytical approaches with varved marine sediments to demonstrate the potential to reconstruct seasonal and inter‐annual climate variability. Obtaining paleoclimate data at a resolution akin to the observational record is vitally important for improving our understanding of climate phenomena such as monsoons and modes of variability such as the El Niño Southern Oscillation, for which appraisals of past inter‐annual variability is critical. The ability to generate seasonal and inter annual resolution sea surface temperature proxy time series spanning a thousand years or more is revolutionary and has the potential to fill gaps in our knowledge of climate variability. Although generally limited to sediments from regions with oxygen depleted bottom waters, there is great potential to integrate shorter seasonal resolution climate “snap shots” from other archives such as annually banded corals into composite time series. But as paleoceanographic data are used more by the observational and modeling fields, we make the case for conducting a thorough case‐by‐case assessment of the processes that influence the climate signal recovered from proxies, using careful replication to validate new approaches. Understanding or exploring the potential influence of processes which effectively filter the climate signal will lead to more quantitative paleoceanographic data that will better serve the broader climate science community.

Global and Regional Snow Cover Decline: 2000–2022

Snow cover affects the global surface energy balance and, with its high albedo, exerts a cooling effect on the Earth’s climate. Decreases in snow cover alter the flow of solar energy from being reflected away from Earth to being absorbed, increasing the Earth’s surface temperature. To gain a global understanding of snow cover change, in situ measurements are too few and far between, so remotely sensed data are needed. This research used the medium-resolution sensor MODIS on the Terra satellite, which has been observing global snow cover almost daily since the year 2000. Here, the MOD10C2 eight-day maximum value composite time series data from February 2000 to March 2023 were analyzed to detect global and regional trends in snow cover extent for the first 23 years of the 21st century. Trends in snow cover change during different time periods (seasons and snow-year) were examined using the Mann—Kendall test and the univariate differencing analysis. Both methods produced similar results. Globally, snow cover declined two to ten times as much as it increased, depending on the season of analysis, and annually, global snow cover decreased 5.12% (not including Antarctica or Greenland) based on the Mann—Kendall test at the 95th percentile (p < 0.05). Regionally, Asia had the greatest net area decline in snow cover, followed by Europe. Although North America has the second-largest extent of snow cover, it had the least amount of net decreasing snow cover relative to its size. South America had the greatest local decline in snow cover, decreasing 20.60% of its annual (snow-year) snow cover area. The Australia–New Zealand region, with just 0.34% of the global snow cover, was the only region to have a net increase in snow cover, increasing 3.61% of its annual snow cover area.

Global Synthesis of Regional Holocene Hydroclimate Variability Using Proxy and Model Data

AbstractSubstantial changes in terrestrial hydroclimate during the Holocene are recorded in geological archives and simulated by computer models. To identify spatial and temporal patterns during the past 12 ka, proxy records sensitive to changing precipitation and effective moisture (precipitation minus evaporation) were compiled from across the globe (n = 813). Proxy composite timeseries were computed for 30 of the IPCC AR6 regions and compared to two full‐Holocene transient model simulations (TraCE‐21ka and HadCM3) and twelve mid‐Holocene CMIP6 simulations. We find that throughout Northern Hemisphere monsoon regions, proxy and model simulations indicate wetter‐than‐modern conditions during the early and mid‐Holocene while Southern Hemisphere monsoon regions were drier. This insolation driven trend toward modern values began approximately 6,000 years ago, and the clear agreement among proxy records and models may reflect the large magnitude of precipitation change and consistent atmospheric circulation forcing mechanism for these regions. In the midlatitudes, the pattern of change is less certain. Generally, proxy composites show a wetting trend throughout the Holocene for the northern midlatitudes, possibly due to strengthening westerlies from an increasing latitudinal temperature gradient. However, simulations indicate that the magnitude of change was relatively low, and for portions of North America, there is a proxy‐model disagreement. At high latitudes, hydroclimate is positively correlated with temperature in both proxies and models, consistent with projected wetting as temperatures rise. Overall, this large proxy database reveals a coherent pattern of hydroclimate variability despite the challenges associated with reconstructing hydroclimate fields.

The drivers of plant community composition have shifted from external to internal processes over the past 20,000 years

Internal and external factors regulating the past composition of plant communities are difficult to identify in palaeo-vegetation records. Here, we develop an index of relative entropy of community assembly, which applies to changes in the composition of a community over time, measuring disorder in its assembly relative to disassembly. Historical periods of relatively ordered assembly (negative index values) are characteristic of a community undergoing endogenous self-organisation, in contrast to relatively disordered assembly (positive values) characterising periods of exogenous abiotic forcing. We quantified the relative entropy index for a 22,000-year time-series of tundra vegetation obtained in the Polar Urals, based on sedimentary DNA. We find it most positive during the Late Pleistocene characterized by persistent taxa, and most negative during the post-glacial Holocene characterized by more ephemeral floras. Changes in relative entropy coincide with changes in regional temperature as reconstructed from stable oxygen composition of an Arctic ice-core. Our results suggest that temperature strongly influenced community assembly in the Polar Urals until about 9000 years before present, after which endogenous community self-organization prevailed through to the present. We conclude that time-series of community composition can reveal changes in the balance between internal and external influences on taxonomic turnover and resulting diversity.

Increasing Nonfossil Fuel Contributions to Atmospheric Nitrate in Urban China from Observation to Prediction.

China's nitrogen oxide (NOx) emissions have undergone significant changes over the past few decades. However, nonfossil fuel NOx emissions are not yet well constrained in urban environments, resulting in a substantial underestimation of their importance relative to the known fossil fuel NOx emissions. We developed an approach using machine learning that is accurate enough to generate a long time series of the nitrogen isotopic composition (δ15N) of atmospheric nitrate using high-level accuracies of air pollutants and meteorology data. Air temperature was found to be the critical driver of the variation of nitrate δ15N at daily resolution based on this approach, while significant reductions of aerosol and its precursor emissions played a key role in the change of nitrate δ15N on the yearly scale. Predictions from this model found a significant decrease in nitrate δ15N in Chinese megacities (Beijing and Guangzhou as representative cities in the north and south, respectively) since 2013, implying an enhanced contribution of nonfossil fuel NOx emissions to nitrate aerosols (up to 22%-26% in 2021 from 18%-22% in 2013 quantified by an isotope mixing model), as confirmed by the Weather Research and Forecasting model coupled with online chemistry (WRF-Chem) simulation. Meanwhile, the declining contribution in coal combustion (34%-39% in 2013 to 31%-34% in 2021) and increasing contribution of natural gas combustion (11%-14% in 2013 to 14%-17% in 2021) demonstrated the transformation of China's energy structure from coal to natural gas. This approach provides missing records for exploring long-term variability in the nitrogen isotope system and may contribute to the study of the global reactive nitrogen biogeochemical cycle.

A dynamic nonlinear optimization framework for learning data-driven reduced-order microkinetic models

The use of kinetic models is key for analyzing, designing, controlling, and optimizing the manufacturing of high value chemicals. In particular, microkinetic modeling relies on large-scale networks of elementary reaction steps that map the conversion of feed materials to final products by considering a large number of pathways and intermediate species for a given reactor configuration. Intuitively, several of the (often automatically generated) reaction steps might be redundant or insignificant, and thus determining the true governing reaction network is critical to understanding and modeling the underlying chemistry. This work introduces a nonlinear dynamic optimization framework for discovering governing reaction networks from data, whereby both the model structure (the elementary reaction steps) and the model parameters (reaction rate constants, pre-exponential factors, and activation energies) are simultaneously learned from composition time series data. The proposed framework can also achieve dimensionality reduction to produce accurate reduced-order microkinetic models that are computationally parsimonious and thus better suited for applications involving simulation and optimization. Two numerical examples of different dimensions are presented to illustrate the key properties of our approach.

Spherical autoregressive models, with application to distributional and compositional time series

We introduce a new class of autoregressive models for spherical time series. The dimension of the spheres on which the observations of the time series are situated may be finite-dimensional or infinite-dimensional, where in the latter case we consider the Hilbert sphere. Spherical time series arise in various settings. We focus here on distributional and compositional time series. Applying a square root transformation to the densities of the observations of a distributional time series maps the distributional observations to the Hilbert sphere, equipped with the Fisher–Rao metric. Likewise, applying a square root transformation to the components of the observations of a compositional time series maps the compositional observations to a finite-dimensional sphere, equipped with the geodesic metric on spheres. The challenge in modeling such time series lies in the intrinsic non-linearity of spheres and Hilbert spheres, where conventional arithmetic operations such as addition or scalar multiplication are no longer available. To address this difficulty, we consider rotation operators to map observations on the sphere. Specifically, we introduce a class of skew-symmetric operators such that the associated exponential operators are rotation operators that for each given pair of points on the sphere map the first point of the pair to the second point of the pair. We exploit the fact that the space of skew-symmetric operators is Hilbertian to develop autoregressive modeling of geometric differences that correspond to rotations of spherical and distributional time series. Differences expressed in terms of rotations can be taken between the Fréchet mean and the observations or between consecutive observations of the time series. We derive theoretical properties of the ensuing autoregressive models and showcase these approaches with several motivating data. These include a time series of yearly observations of bivariate distributions of the minimum/maximum temperatures for a period of 120 days during each summer for the years 1990-2018 at Los Angeles (LAX) and John F. Kennedy (JFK) international airports. A second data application concerns a compositional time series with annual observations of compositions of energy sources for power generation in the U.S..

Moisture Sources and Pathways Determine Stable Isotope Signature of Himalayan Waters in Nepal

AbstractThe Himalayan mountain range produces one of the steepest and largest rainfall gradients on Earth, with >3 m/yr rainfall difference over a ∼100 km distance. The Indian Summer Monsoon (ISM) contributes more than 80% to the annual precipitation budget of the central Himalayas. The remaining 20% falls mainly during pre‐ISM season. Understanding the seasonal cycle and the transfer pathways of moisture from precipitation to the rivers is crucial for constraining water availability in a warming climate. However, the partitioning of moisture into the different storage systems such as snow, glacier, and groundwater and their relative contribution to river discharge throughout the year remains under‐constrained. Here, we present novel field data from the Kali Gandaki, a trans‐Himalayan river, and use 4‐year time series of river and rain water stable isotope composition (δ18O and δ2H values) as well as river discharge, satellite Global Precipitation Measurement amounts, and moisture source trajectories to constrain hydrological variability. We find that rainfall before the onset of the ISM is isotopically distinct and that ISM rain and groundwater have similar isotopic values. Our study lays the groundwork for using isotopic measurements to track changes in precipitation sources during the pre‐ISM to ISM transition in this key region of orographic precipitation. Specifically, we highlight the role of pre‐ISM precipitation, derived from the Gangetic plain, to define the seasonal river isotopic variability across the central Himalayas. Lastly, isotopic values across the catchment document the importance of a large well‐mixed groundwater reservoir supplying river discharge, especially during the non‐ISM season.

Characterizing Tree Species in Northern Boreal Forests Using Multiple-Endmember Spectral Mixture Analysis and Multi-Temporal Satellite Imagery

Northern boreal forests are characterized by open stands whereby trees, understory background, and shadow are all significant components of the spectral response within a pixels’ spatial footprint. To overcome this mixed pixel problem, accurate spectral characterization of these (endmember) components is necessary for spectral mixture analysis (SMA) to generate forest classifications at the species level. Obtaining these endmember spectra in the field, however, can be difficult or impossible. This study examined whether image endmember spectra can be identified using forest inventory information to derive dominant tree species classifications. This was tested using multiple-endmember SMA (MESMA) and single- and multi-date Landsat imagery of a forested area in the Northwest Territories, Canada. Image classifications (n = 80) were generated based on 20 image-date combinations and four unmixing models. Accuracies of 80% and 82% were achieved for open and medium dense forest stands, respectively using multi-date imagery, which outperformed single-date imagery acquired at peak phenology. The overall accuracy is 72%; lower due to challenges in very open stands. The multi-date MESMA approach was robust for both compositionally pure and mixed stands. The approach merits further investigation, particularly within the context of the increasing availability of regional-scale satellite imagery enabling composite time-series and spectral-temporal image features.