Cropland Data Layer Research Articles

Transfer Learning for Crop classification with Cropland Data Layer data (CDL) as training samples

Training samples is fundamental for crop mapping from remotely sensed images, but difficult to acquire in many regions through ground survey, causing significant challenge for crop mapping in these regions. In this paper, a transfer learning (TL) workflow is proposed to use the classification model trained in contiguous U.S.A. (CONUS) to identify crop types in other regions. The workflow is based on fact that same crop growing in different regions of world has similar temporal growth pattern. This study selected high confidence pixels across CONUS in the Cropland Data Layer (CDL) and corresponding 30-m 15-day composited NDVI time series generated from harmonized Landat-8 and Sentinel-2 (HLS) data as training samples, trained the Random Forest (RF) classification models and then applied the models to identify crop types in three test regions, namely Hengshui in China (HS), Alberta in Canada (AB), and Nebraska in USA (NE). NDVI time series with different length were used to identify crops, the effect of time-series length on classification accuracies were then evaluated. Furthermore, local training samples in the three test regions were collected and used to identify crops (LO) for comparison. Results showed that overall classification accuracies in HS, AB and NE were 97.79%, 86.45% and 94.86%, respectively, when using TL with NDVI time series of the entire growing season for classification. However, LO could achieve higher classification accuracies earlier than TL. Because the training samples were collected across USA containing multiple growth conditions, it increased the potential that the crop growth environment in test regions could be similar to those of the training samples; but also led to situation that different crops had similar NDVI time series, which caused lower TL classification accuracy in HS at early-season. Generally, this study provides new options for crop classification in regions of training samples shortage.

Recovering Forecast Distributions of Crop Composition: Method and Application to Kentucky Agriculture

This paper proposes a novel application of the multinomial logit (MNL) model using Cropland Data Layer and field-level boundaries to estimate crop transition probabilities, which are used to generate forecast distributions of total acreage for five major crops produced in the state of Kentucky. These forecasts distributions have a wide range of applications that, besides providing interim acreage estimates ahead of the June Acreage Survey, can inform the ability of producers to incorporate new crops in the land-use rotation, investments in location-specific capital and input distribution as well informing the likelihood of adverse water quality events from nutrient run-off.

AgKit4EE: A toolkit for agricultural land use modeling of the conterminous United States based on Google Earth Engine

Google Earth Engine (GEE) is an ideal platform for large-scale geospatial agricultural and environmental modeling based on its diverse geospatial datasets, easy-to-use application programming interface (API), rich reusable library, and high-performance computational capacity. However, using GEE to prepare geospatial data requires not only the skills of programming languages like JavaScript and Python, but also the knowledge of GEE APIs and data catalog. This paper presents the AgKit4EE toolkit to facilitate the use of the Cropland Data Layer (CDL) product over the GEE platform. This toolkit contains a variety of frequently used functions for use of CDL including crop sequence modeling, crop frequency modeling, confidence layer modeling, and land use change analysis. The experimental results suggest that the proposed software can significantly reduce the workload for modelers who conduct geospatial agricultural and environmental modeling with CDL data as well as developers who build the GEE-enabled geospatial cyberinfrastructure for agricultural land use modeling of the conterminous United States. AgKit4EE is an open source and it is free to use, modify, and distribute. The latest release of AgKit4EE can be imported to any modeling workflow developed using GEE Code Editor (https://code.earthengine.google.com/?accept_repo=users/czhang11/agkit4ee). The source code, examples, documentation, user community, and wiki pages are available on GitHub (https://github.com/czhang11/agkit4ee).

REFINEMENT OF CROPLAND DATA LAYER USING MACHINE LEARNING

Abstract. As the most widely used crop-specific land use data, the Cropland Data Layer (CDL) product covers the entire Contiguous United States (CONUS) at 30-meter spatial resolution with very high accuracy up to 95% for major crop types (i.e., Corn, Soybean) in major crop area. However, the quality of early-year CDL products were not as good as the recent ones. There are many erroneous pixels in the early-year CDL product due to the cloud cover of the original Landsat images, which affect many follow-on researches and applications. To address this issue, we explore the feasibility of using machine learning technology to refine and correct misclassified pixels in the historical CDLs in this study. An end-to-end deep learning-based framework for restoration of misclassified pixels in CDL image is developed and tested. By feeding the CDL time series into the artificial neural network, a crop sequence model is trained and the misclassified pixels in an original CDL map can be restored. In the experiment with the 2005 CDL data of the State of Illinois, the misclassified pixels over Agricultural Statistics Districts (ASD) #1760 were corrected with a reasonable accuracy (> 85%). The findings suggest that the proposed method provides a low-cost and reliable way to refine the historical CDL data, which can be potentially scaled up to the entire CONUS.

Deep Learning Classification for Crop Types in North Dakota

Recently, agricultural remote sensing community has endeavored to utilize the power of artificial intelligence (AI). One important topic is using AI to make the mapping of crops more accurate, automatic, and rapid. This article proposed a classification workflow using deep neural network (DNN) to produce high-quality in-season crop maps from Landsat imageries for North Dakota. We use historical crop maps from the agricultural department and North Dakota ground measurements as training datasets. Processing workflows are created to automate the tedious preprocessing, training, testing, and postprocessing workflows. We tested this hybrid solution on new images and received accurate results on major crops such as corn, soybean, barley, spring wheat, dry beans, sugar beets, and alfalfa. The pixelwise overall accuracy in all three test regions is over 82% for all land types (including noncrop land), which is the same level of accuracy as the U.S. Department of Agriculture Cropland Data Layer. The texture of DNN maps is more consistent with fewer noises, which is more comfortable to read. We find DNN is better on recognizing big farmlands than recognizing the scattered wetlands and suburban regions in North Dakota. The model trained on multiple scenes of multiple years and months yields higher accuracy than any of the models trained only on a single scene, a single month, or a single year. These results reflect that DNN can produce reliable in-season maps for major crops in North Dakota big farms and could provide a relatively accurate reference for the minor crops in scattered wetland fields.

Temporal and spatial variations of irrigation water use for commercial corn fields in Central Nebraska

The increasing pressure on water resources in Nebraska-US and other agricultural areas requires the implementation of innovative tools and solutions for the governance of water resources and the analysis of water use efficiency. In this vein, this paper presents the application of a remote sensing based soil water balance for the study of water use in agricultural areas. The specific objectives were the identification of the temporal and spatial behavior of the irrigation water use based on the quantification of the water use deviation (irrigation water applied minus irrigation water requirements), as the main indicator and the comparative analysis of the irrigation productivity (crop yield under irrigated field minus crop yield under rainfed condition per volume of water applied by irrigation), WPi, water productivity (harvestable grain per total volume of water applied considering precipitation plus irrigation), WP, and finally water productivity based on evapotranspiration (harvested grain per total volume of water evapotranspired), WPET, in the various management zones analyzed. Additionally, we examined the impact of soil types, local weather and irrigation system (center pivot and furrow irrigation) on these indicators. The study was carried out in three Natural Resources District (Tri-Basin, Central Platte and Lower Niobrara) across Central Nebraska for the period 2004–2012 and comprised over 2000 irrigated corn fields per year. Crop water requirements were estimated using the reflectance-based crop coefficient approach developed in previous research (see Campos et al., 2017) and the field data were reported for each field monitored through cropland data layer by National Agricultural Statistics Service of USDA. The difference between modeled irrigation water requirements and field level irrigation application was significant (p < 0.001) being the water use deviation in generally positive (over irrigation). These results were consistently higher for furrow irrigated fields during the whole analyzed period, reaching up to three times more water applied compared to the required amount. This was expected as surface irrigation systems typically require a higher application depth. This trend changed for the central pivot irrigated fields depending on the climatic conditions, especially in dry years. The analysis of the water use deviation with respect to soil types and weather conditions revealed that the water use deviation is not justified by the biophysical conditions alone. The estimated values of WP and WPi for furrow system was lower compared to center pivot in both NRD’s reaching the maximum value of 1.37 kg m-3 and 3.06 kg m-3 for WP and WPi in Tri-basin respectively for center pivot. In general, the results suggested potential to improve water management in these NRDs in Central Nebraska and reduce pumping potentially saving groundwater resources for drought years and other uses monitoring soil type, weather data and switching to sprinklers system.

Thinking Like a Grassland: Challenges and Opportunities for Biodiversity Conservation in the Great Plains of North America

Fauna of North America’s Great Plains evolved strategies to contend with the region’s extreme spatiotemporal variability in weather and low annual primary productivity. The capacity for large-scale movement (migration and/or nomadism) enables many species, from bison to lark buntings, to track pulses of productivity at broad spatial scales (> 1 000 km2). Furthermore, even sedentary species often rely on metapopulation dynamics over extensive landscapes for long-term population viability. The current complex pattern of land ownership and use of Great Plains grasslands challenges native species conservation. Approaches to managing both public and private grasslands, frequently focused at the scale of individual pastures or ranches, limit opportunities to conserve landscape-scale processes such as fire, animal movement, and metapopulation dynamics. Using the US National Land Cover Database and Cropland Data Layers for 2011−2017, we analyzed land cover patterns for 12 historical grassland and savanna communities (regions) within the US Great Plains. On the basis of the results of these analyses, we highlight the critical contribution of restored grasslands to the future conservation of Great Plains biodiversity, such as those enrolled in the Conservation Reserve Program. Managing disturbance regimes at larger spatial scales will require acknowledging that, where native large herbivores are absent, domestic livestock grazing can function as a central component of Great Plains disturbance regimes if they are able move at large spatial scales and coexist with a diverse array of native flora and fauna. Opportunities to increase the scale of grassland management include 1) spatial prioritization of grassland restoration and reintroduction of grazing and fire, 2) finding creative approaches to increase the spatial scale at which fire and grazing can be applied to address watershed to landscape-scale objectives, and 3) developing partnerships among government agencies, landowners, businesses, and conservation organizations that enhance cross-jurisdiction management and address biodiversity conservation in grassland landscapes, rather than pastures.

Land Surface Phenologies and Seasonalities in the US Prairie Pothole Region Coupling AMSR Passive Microwave Data with the USDA Cropland Data Layer

Land surface phenologies and seasonalities in the US Prairie Pothole Region (PPR) were characterized using land surface variables derived from the coarse spatial resolution (25 km) Advanced Microwave Scanning Radiometer (AMSR) blended data for 2003 to 2016 linked with the optically based USDA NASS Crop Data Layer (CDL) at a much finer spatial resolution. Two transects of AMSR pixels—one in east-central North Dakota and the other in eastern South Dakota—were selected for analysis. The AMSR data were grouped earlier (2003–2005, 2007) and later (2013–2016) to emphasize temporal change and to avoid data discontinuity in 2011–2012 and a major drought in 2006. The nonparametric Mann-Kendall trend test on the CDL data revealed that area in grasslands and wetlands strongly decreased in both transects, while corn and soybean coverage strongly increased. In crop-dominated sites, the AMSR vegetation optical depth (VOD) time series caught the early spring growth, ploughing, and crop growth and senescence. In contrast, the VOD time series at grassland dominated sites exhibited a lower peak but extended growth period. Crop-dominated sites had significantly higher amplitude VODs in both periods and transects. Based on the paired two-sample t-test, neither the peak VOD amplitude nor the peak VOD timing measured in accumulated growing degree-days was significantly different between temporal groups in the North Dakota transect. In contrast, in South Dakota, both the peak VOD amplitude and its timing were significantly different with shifts to later peak timing during the 2013–2016 period. In addition, in South Dakota but not North Dakota, there were significantly earlier shifts in the timing of peak growing degree-days and peak precipitation water vapor. Both spatial and temporal changes in AMSR land surface variables are linked to shifts in land cover along the South Dakota transect as revealed in the CDL data. More research is required to understand the dynamics evident in the passive microwave time series.

Machine-learned prediction of annual crop planting in the U.S. Corn Belt based on historical crop planting maps

An accurate crop planting map can provide essential information for decision support in agriculture. The method of post-season and in-season crop mapping has been widely studied in the land use and land cover community. However, it remains a challenge to predict the spatial distribution of crop planting before the growing season. This paper is the first attempt to use machine learning approach on the prediction of field-level annual crop planting from historical crop planting maps. We present an end-to-end machine learning framework for crop planting prediction using Cropland Data Layer (CDL) time series as reference data and multi-layer artificial neural network as prediction model. The proposed framework was first tested at Lancaster County of Nebraska State, then scaled up to the U.S. Corn Belt. According to the experiment results from 53 Agricultural Statistics Districts, we found the machine-learned crop planting map was expected to reach 88% agreement with the future CDL. Meanwhile, the crop acreage estimates derived from the machine-learned prediction were highly correlated (R2 > 0.9) with the crop acreage estimates of CDL and official statistics by the U.S. Department of Agriculture National Agricultural Statistics Service. This study provides a low-cost and efficient way to predict annual crop planting map, which can be used to support many agricultural applications and decision makings before the beginning of a growing season.

Deep learning based winter wheat mapping using statistical data as ground references in Kansas and northern Texas, US

Winter wheat is a major staple crop and it is critical to monitor winter wheat production using efficient and automated means. This study proposed a novel approach to produce winter wheat maps using statistics as the training targets of supervised classification. Deep neural network architectures were built to link remotely sensed image series to the harvested areas of individual administrative units. After training, the resultant maps were generated using the activations on a middle layer of the deep model. The direct use of statistical data to some extent alleviates the shortage of ground samples in classification tasks and provides an opportunity to utilize a wealth of statistical records to improve land use mapping. The experiments were carried out in Kansas and Northern Texas during 2001–2017. For each study area the goal was to create winter maps that are consistent with USDA county-level statistics of harvested areas. The trained deep models automatically identified the seasonal pattern of winter wheat pixels without using pixel-level reference data. The winter wheat maps were compared with the Cropland Data Layer (CDL) for years when the CDL is available. In Kansas where the winter wheat extent of the CDL has high reported accuracy and agrees well with county statistics, the maps produced from the deep model was evaluated using the CDL as an independent test set. Northern Texas was selected as an example where the winter wheat area of the CDL is very different from official statistics, and the maps by the deep model enabled a map-to-map comparison with the CDL to highlight the areas of discrepancy. Visual representation of the deep model behaviors and recognized patterns show that deep learning is an automated and robust means to handle the variability of winter wheat seasonality without the need of manual feature engineering and intensive ground data collection. Showing the possibility of generating maps solely from regional statistics, the proposed deep learning approach has great potential to fill the historical gaps of conventional sample-based classification and extend applications to areas where only regional statistics are available. The flexible deep network architecture can be fused with various statistical datasets to fully employ existing sources of data and knowledge.

Modelling the impacts of grassland to cropland conversion on river flow regimes in Skunk Creek watershed, Upper Midwest United States

AbstractConversion of grassland to cultivated cropland has been linked to downstream alteration of flow regimes. This study used the Soil and Water Assessment Tool (SWAT) to construct seven “what if” scenarios for quantifying the impacts of grassland to cropland and vice versa conversion (i.e., replacement of grassland with selected agricultural crops) on river flow regimes in Skunk Creek watershed. The Cropland Data Layer for the year 2011 in conjunction with historical climate data was used to create SWAT models for scenario simulations over 19 years, from 1996 to 2014. The model developed for the historical climate records (baseline) was compared with the scenarios examined using stream flow metrics for a range of flow regimes, including magnitude, duration, frequency, and timing of annual low‐ and high‐flow conditions. The simulation results suggest that grassland conversion to cultivated cropland would generally increase river flows compared with conversion of cultivated cropland to grassland, which may reduce flows in the watershed. Low and moderate flows increased by 2–8% from the baseline scenario with conversion of grass crop and by 1–20% decreases with crop–grass conversion. High flows increased by 3–7% and decreased by 1–18% when grass is converted to crop and crop to grass, respectively. The analysis also suggests that grassland establishment may attenuate the peaks of prolonged small floods and shorter but earlier large floods.

Using the Landsat archive to map crop cover history across the United States

Landsat Thematic Mapper has been collecting multispectral imagery at 30 m resolution globally since 1984. One utility of the data has been for detailed mapping of agricultural regions and seasonal identification of crops grown within them. However, the ability to do so has only been applied sporadically and eluded widespread adoption due to cost of the imagery, burdensome preprocessing requirements, and computing not being up to the task. These hurdles have become much reduced of recent with the free and open distribution of the Landsat imagery, emphasis on ready-to-use surface reflectance data products, and distributed high performance computing infrastructures available online in “the cloud.” As such, this work leverages these aspects and investigates the ability to retrospectively map summer crops over the United States (US) annually from 1984 to 2007. Google's Earth Engine Internet-based analytical platform containing the historical Landsat archive in surface reflectance format was used as a foundation for the classification work. Robust 30 m Cropland Data Layer (CDL) information from US Department of Agriculture (USDA) for years 2008 through 2011 were leveraged to train rule-based classifiers which were applied back through time to each year 1984 through 2007. Focus crops were corn, soybeans, and winter wheat – the three largest by area in the US. A large sampling of highly intensive counties throughout the country were prototyped for generation of the 24 years of historical crop cover. For validation, crop area statistics were calculated for each county-year and compared to survey-based information existing from the USDA. Results were muted overall with the average crop area coefficient of determination (R2) correlations for the years 1984–2007 found to be 0.192, 0.159, and 0.142 for corn, soybeans, and winter wheat, respectively. Furthermore, the standard deviations were variable at 0.132, 0.177, and 0.133, also respectively. While unimpressive, it was found as a benchmark that the R2 between the 2008 through 2017 CDL classifications were only 0.478, 0.686, and 0.726 and thus a suggestion that the USDA area statistics are an imperfect measure of map accuracy. Deletion of approximately one third to one half of the grossest 1984–2007 outlier years from the historical outputs pulled the correlations to the benchmark standard. Qualitatively, most of the remaining years classified looked of high quality and were believed useful as field-level thematic crop area maps. These historical cropland maps could provide the ability to better detail the role farming has played on the broad US landscape over recent decades.

Impact of land cover on groundwater quality in the Upper Floridan Aquifer in Florida, United States

Although agricultural lands are generally assumed to correlate negatively with groundwater quality, the intricate relationship between general land cover and contaminant concentrations present in an aquifer may vary substantially; contingent upon the land type, interacting factors, and scale considered. The Upper Floridan Aquifer (UFA) is a primary source of potable water supply for the state of Florida. The Suwannee River Water Management District (SRWMD), located in northcentral Florida, relies exclusively on the UFA for water supplies. Over much of the SRWMD in the UFA is unconfined, rendering it vulnerable to contamination from surface sources. This study analyses groundwater concentrations of Nitrate-Nitrogen (NO3–N) and Potassium (K) from shallow wells across the SRWMD for assessing the effect of different land covers on groundwater quality over time. Annual potentiometric surface maps were used to delineate semicircular recharge zones of 500 m, 1000 m, and 2000 m radii upstream of sampled well stations. Proportions of agriculture, forest, and urban lands were identified for each buffer zone using USDA Cropland Data Layer. Multivariate regression models were developed to infer relationships between land cover and NO3–N and K concentrations. Results show significant associations among land cover type, water table height, and groundwater quality parameters. Specifically, we find a large proportion of agricultural cover consistently associated with larger increases in groundwater pollutant loads relative to urban or forest cover across all models, after controlling for depth to water table. Our study suggests a need for widespread adoption of cost-effective agricultural best management practices (BMPs) that could help in securing regional water supply.

Rapid Flood Progress Monitoring in Cropland with NASA SMAP

Research in different agricultural sectors, including in crop loss estimation during flood and yield estimation, substantially rely on inundation information. Spaceborne remote sensing has widely been used in the mapping and monitoring of floods. However, the inability of optical remote sensing to cloud penetration and the scarcity of fine temporal resolution SAR data hinder the application of flood mapping in many cases. Soil Moisture Active Passive (SMAP) level 4 products, which are model-driven soil moisture data derived from SMAP observations and are available at 3-h intervals, can offer an intermediate but effective solution. This study maps flood progress in croplands by incorporating SMAP surface soil moisture, soil physical properties, and national floodplain information. Soil moisture above the effective soil porosity is a direct indication of soil saturation. Soil moisture also increases considerably during a flood event. Therefore, this approach took into account three conditions to map the flooded pixels: a minimum of 0.05 m3m−3 increment in soil moisture from pre-flood to post-flood condition, soil moisture above the effective soil porosity, and the holding of saturation condition for the 72 consecutive hours. Results indicated that the SMAP-derived maps were able to successfully map most of the flooded areas in the reference maps in the majority of the cases, though with some degree of overestimation (due to the coarse spatial resolution of SMAP). Finally, the inundated croplands are extracted from saturated areas by Spatial Hazard Zone areas (SHFA) of Federal Emergency Management Agency (FEMA) and cropland data layer (CDL). The flood maps extracted from SMAP data are validated with FEMA-declared affected counties as well as with flood maps from other sources.

Crop type mapping without field-level labels: Random forest transfer and unsupervised clustering techniques

Crop type mapping at the field level is necessary for a variety of applications in agricultural monitoring and food security. As remote sensing imagery continues to increase in spatial and temporal resolution, it is becoming an increasingly powerful raw input from which to create crop type maps. Still, automated crop type mapping remains constrained by a lack of field-level crop labels for training supervised classification models. In this study, we explore the use of random forests transferred across geographic distance and time and unsupervised methods in conjunction with aggregate crop statistics for crop type mapping in the US Midwest, where we simulated the label-poor setting by depriving the models of labels in various states and years. We validated our methodology using available 30 m spatial resolution crop type labels from the US Department of Agriculture's Cropland Data Layer (CDL). Using Google Earth Engine, we computed Fourier transforms (or harmonic regressions) on the time series of Landsat Surface Reflectance and derived vegetation indices, and extracted the coefficients as features for machine learning models. We found that random forests trained on regions and years similar in growing degree days (GDD) transfer to the target region with accuracies consistently exceeding 80%. Accuracies decrease as differences in GDD expand. Unsupervised Gaussian mixture models (GMM) with class labels derived using county-level crop statistics classify crops less consistently but require no field-level labels for training. GMM achieves over 85% accuracy in states with low crop diversity (Illinois, Iowa, Indiana, Nebraska), but performs sometimes no better than random when high crop diversity interferes with clustering (North Dakota, South Dakota, Wisconsin, Michigan). Under the appropriate conditions, these methods offer options for field-resolution crop type mapping in regions around the world with few or no ground labels.

In-Season Major Crop-Type Identification for US Cropland from Landsat Images Using Crop-Rotation Pattern and Progressive Data Classification

Crop type information at the field level is vital for many types of research and applications. The United States Department of Agriculture (USDA) provides information on crop types for US cropland as a Cropland Data Layer (CDL). However, CDL is only available at the end of the year after the crop growing season. Therefore, CDL is unable to support in-season research and decision-making regarding crop loss estimation, yield estimation, and grain pricing. The USDA mostly relies on field survey and farmers’ reports for the ground truth to train image classification models, which is one of the major reasons for the delayed release of CDL. This research aims to use trusted pixels as ground truth to train classification models. Trusted pixels are pixels which follow a specific crop rotation pattern. These trusted pixels are used to train image classification models for the classification of in-season Landsat images to identify major crop types. Six different classification algorithms are investigated and tested to select the best algorithm for this study. The Random Forest algorithm stands out among selected algorithms. This study classified Landsat scenes between May and mid-August for Iowa. The overall agreements of classification results with CDL in 2017 are 84%, 94%, and 96% for May, June, and July, respectively. The classification accuracies have been assessed through 683 ground truth data points collected from the fields. The overall accuracies of single date multi-band image classification are 84%, 89% and 92% for May, June, and July, respectively. The result also shows higher accuracy (94–95%) can be achieved through multi-date image classification compared to single date image classification.

Reflectance-based Model for Soybean Mapping in United States at Common Land Unit Scale with Landsat 8

ABSTRACT The objective of this study is to validate the feasibility of a reflectance-based model for soybean crop area classification in advance of the county scale statistics from the United States Department of Agriculture (USDA). This classification method is named Reflectance-based North American Model (RNAM). It operates through the analysis of the main physically driven characteristics of farm fields and their specific radiometric profile obtained from Operational Land Imager (OLI) onboard Landsat 8. The state area of Illinois/US was selected because it is the largest soybean producer and accounted for nearly 35 percent of the total soybeans production in US. Farm fields within a set of 32 counties were analyzed for six crop years between 2013 to 2018. Results obtained from RNAM were compared to official estimates of USDA at county level. Coefficients R2 ranged between 0.92 and 0.96, indicating good agreement of the estimates. Results from RNAM were also validated with the geospatial reference map Cropland Data Layer (CDL) of soybeans from USDA. The overall map accuracy found was 93.86% with Kappa Index of Agreement of 0.795. Thus, RNAM was considered able to provide timely thematic soybean maps, in late September, in advance of the county scale statistics from USDA.

Characterizing land cover/land use from multiple years of Landsat and MODIS time series: A novel approach using land surface phenology modeling and random forest classifier

Over the last 20 years, substantial amounts of grassland have been converted to other land uses in the Northern Great Plains. Most of land cover/land use (LCLU) assessments in this region have been based on the U.S. Department of Agriculture - Cropland Data Layer (USDA - CDL), which may be inconsistent. Here, we demonstrate an approach to map land cover utilizing multi-temporal Earth Observation data from Landsat and MODIS. We first built an annual time series of accumulated growing degree-days (AGDD) from MODIS 8-day composites of land surface temperatures. Using the Enhanced Vegetation Index (EVI) derived from Landsat Collection 1's surface reflectance, we then fit at each pixel a downward convex quadratic model to each year's progression of AGDD (i.e., EVI = α + β × AGDD − γ × AGDD2). Phenological metrics derived from fitted model and the goodness of fit then are submitted to a random forest classifier (RFC) to characterize LCLU for four sample counties in South Dakota in three years (2006, 2012, 2014) when reference point datasets are available for training and validation. To examine the sensitivity of the RFC to sample size and design, we performed classifications under different sample selection scenarios. The results indicate that our proposed method accurately mapped major crops in the study area but showed limited accuracy for non-vegetated land covers. Although all RFC models exhibit high accuracy, estimated land cover areas from alternative models could vary widely, suggesting the need for a careful examination of model stability in any future land cover supervised classification study. Among all sampling designs, the “same distribution” models (proportional distribution of the sample is like proportional distribution of the population) tend to yield best land cover prediction. RFC used only the most eight important variables (e.g., three fitted parameter coefficients [α, β, and γ]; maximum modeled EVI; AGDD at maximum modeled EVI; the number of observations used to fit CxQ model; and the number of valid observations) have slightly higher accuracy compared to those using all variables. By summarizing annual image time series through land surface phenology modeling, LCLU classification can embrace both seasonality and interannual variability, thereby increasing the accuracy of LCLU change detection.

Accuracies Achieved in Classifying Five Leading World Crop Types and their Growth Stages Using Optimal Earth Observing-1 Hyperion Hyperspectral Narrowbands on Google Earth Engine

As the global population increases, we face increasing demand for food and nutrition. Remote sensing can help monitor food availability to assess global food security rapidly and accurately enough to inform decision-making. However, advances in remote sensing technology are still often limited to multispectral broadband sensors. Although these sensors have many applications, they can be limited in studying agricultural crop characteristics such as differentiating crop types and their growth stages with a high degree of accuracy and detail. In contrast, hyperspectral data contain continuous narrowbands that provide data in terms of spectral signatures rather than a few data points along the spectrum, and hence can help advance the study of crop characteristics. To better understand and advance this idea, we conducted a detailed study of five leading world crops (corn, soybean, winter wheat, rice, and cotton) that occupy 75% and 54% of principal crop areas in the United States and the world respectively. The study was conducted in seven agroecological zones of the United States using 99 Earth Observing-1 (EO-1) Hyperion hyperspectral images from 2008–2015 at 30 m resolution. The authors first developed a first-of-its-kind comprehensive Hyperion-derived Hyperspectral Imaging Spectral Library of Agricultural crops (HISA) of these crops in the US based on USDA Cropland Data Layer (CDL) reference data. Principal Component Analysis was used to eliminate redundant bands by using factor loadings to determine which bands most influenced the first few principal components. This resulted in the establishment of 30 optimal hyperspectral narrowbands (OHNBs) for the study of agricultural crops. The rest of the 242 Hyperion HNBs were redundant, uncalibrated, or noisy. Crop types and crop growth stages were classified using linear discriminant analysis (LDA) and support vector machines (SVM) in the Google Earth Engine cloud computing platform using the 30 optimal HNBs (OHNBs). The best overall accuracies were between 75% to 95% in classifying crop types and their growth stages, which were achieved using 15–20 HNBs in the majority of cases. However, in complex cases (e.g., 4 or more crops in a Hyperion image) 25–30 HNBs were required to achieve optimal accuracies. Beyond 25–30 bands, accuracies asymptote. This research makes a significant contribution towards understanding modeling, mapping, and monitoring agricultural crops using data from upcoming hyperspectral satellites, such as NASA’s Surface Biology and Geology mission (formerly HyspIRI mission) and the recently launched HysIS (Indian Hyperspectral Imaging Satellite, 55 bands over 400–950 nm in VNIR and 165 bands over 900–2500 nm in SWIR), and contributions in advancing the building of a novel, first-of-its-kind global hyperspectral imaging spectral-library of agricultural crops (GHISA: www.usgs.gov/WGSC/GHISA).

Enhancing Crop Acreage Estimation within a Semiarid Watershed via Statistical Assessments and Techniques

Core Ideas We assessed methods to effectively reduce error in crop acreage estimates. The study compared statistical methods against National Agricultural Statistics Service data at the state and county level. We determined that improvements in crop estimates are largely affected by the input datasets. The error in crop acreage estimates is dependent on the crop, scope of analysis, and growing season. It is necessary to examine if the statistical methods designed to improve Cropland Data Layer (CDL) acreage estimates at a national level improve estimates at the county level. This study determined whether the calibration of CDL by statistical techniques increased the accuracy of crop estimates on a semiarid West Texas watershed. Since there is a substantial amount of agriculture and farming in areas like West Texas, accurate crop estimates of datasets at smaller scales become important. We compared three methods: the pixel counting method of the CDL, linear regression of the CDL against the MODIS Land Cover Type Product (MLCT), and the Bayesian method (posterior probability) on the National Land Cover Dataset (NLCD) against the USDA National Agricultural Statistical Service (NASS) Quick Stats. Four crops [corn (Zea mays L.), cotton (Gossypium hirsutum L.), sorghum (Sorghum bicolor [L.] Moench), and winter wheat (Triticum aestivum L.)] were analyzed at state and county levels during two growing seasons (2009 and 2011). Prediction errors were estimated by comparing computed acreage with the appropriate NASS Quick Stats values. The total error for each assessment method was estimated for all acreage values in that particular assessment category. We found that the Bayesian method improved the estimates of sorghum and winter wheat and linear regression improved acreage estimates for cotton and corn in 2011 and 2009, respectively. Our results indicate that county‐level adjustments are very sensitive. Therefore, one must carefully select the method that best fits the type of crop, geographical location, and growing season being considered.