skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Quantifying uncertainty in land-use land-cover classification using conformal statistics
Land-use land-cover (LULC) change is one of the most important anthropogenic threats to biodiversity and ecosystems integrity. As a result, the systematic generation of annual regional, national, and global LULC map products derived from the classification of satellite imagery data have become critical inputs for multiple scientific disciplines. The importance of quantifying pixel-level uncertainty to improve the robustness of downstream analyses has long been acknowledged but this practice is still not widely adopted in the generation of these LULC products. The lack of uncertainty quantification is likely due to the fact that most approaches that have been put forward for this task are too computationally intensive for large-scale analysis (e.g., bootstrapping). In this article, we describe how conformal statistics can be used to quantify pixel-level uncertainty in a way that is not computationally intensive, is statistically rigorous despite relying on few assumptions, and can be used together with any classification algorithm that produces class probabilities. Our simulation results show how the size of the predictive sets created by conformal statistics can be used as an indicator of classification uncertainty at the pixel level. Our analysis based on data from the Brazilian Amazon reveals that both forest and water have high certainty whereas pasture and the “natural (other)” category have substantial uncertainty. This information can guide additional ground-truth data collection and the resulting raster combining the LULC classification with the uncertainty results can be used to communicate in a transparent way to downstream users which classified pixels have high or low uncertainty. Given the importance of systematic LULC maps and uncertainty quantification, we believe that this approach will find wide use in the remote sensing community.  more » « less
Award ID(s):
2040819
PAR ID:
10504965
Author(s) / Creator(s):
; ;
Publisher / Repository:
ELSEVIER
Date Published:
Journal Name:
Remote Sensing of Environment
Volume:
295
Issue:
C
ISSN:
0034-4257
Page Range / eLocation ID:
113682
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; (, Zenodo)
    Through integrating multi-source data including high-resolution remote sensing image-based land use and land cover (LULC) data, model-based land use products, and historical land archives, we reconstructed historical LULC at an annual time scale and 1 km x 1 km resolution in the contiguous United States (CONUS) from 1630 to 2020. Compared to other historical LULC datasets, our data can capture the major characters of LULC as well as provide more accurate information with higher spatial and temporal resolution. The LULC data can be used for regional studies in a wide range of topics including LULC impacts on the ecosystem, biodiversity, water resource, carbon and nitrogen cycles, and greenhouse gas emissions.</p> 
    more » « less
  2. ; ; ; ; ; ; ; ; ; (, Frontiers in Water)
    Land use and land cover (LULC) can significantly alter river water, which can in turn have important impacts on downstream coastal ecosystems by delivering nutrients that promote marine eutrophication and hypoxia. Well-documented in temperate systems, less is known about the way land cover relates to water quality in low-lying coastal zones in the tropics. Here we evaluate the catchment LULC and the physical and chemical characteristics of six rivers that contribute flow into a seasonally hypoxic tropical bay in Bocas del Toro, Panama. From July 2019 to March 2020, we routinely surveyed eight physical and chemical characteristics (temperature, specific conductivity, salinity, pH, dissolved oxygen (DO), nitrate and nitrite, ammonium, and phosphate). Our goals were to determine how these physical and chemical characteristics of the rivers reflect the LULC, to compare the water quality of the focal rivers to rivers across Panama, and to discuss the potential impacts of river discharge in the Bay. Overall, we found that the six focal rivers have significantly different river water characteristics that can be linked to catchment LULC and that water quality of rivers 10 s of kilometers apart could differ drastically. Two focal catchments dominated by pristine peat swamp vegetation in San San Pond Sak, showed characteristics typical of blackwater rivers, with low pH, dissolved oxygen, and nutrients. The remaining four catchments were largely mountainous with >50% forest cover. In these rivers, variation in nutrient concentrations were associated with percent urbanization. Comparisons across Panamanian rivers covered in a national survey to our focal rivers shows that saltwater intrusions and low DO of coastal swamp rivers may result in their classification by a standardized water quality index as having slightly contaminated water quality, despite this being their natural state. Examination of deforestation over the last 20 years, show that changes were <10% in the focal catchments, were larger in the small mountainous catchments and suggest that in the past 20 years the physical and chemical characteristics of river water that contributes to Almirante Bay may have shifted slightly in response to these moderate land use changes. (See supplementary information for Spanish-language abstract). 
    more » « less
  3. ; ; ; ; ; (, Land)
    Land is a natural resource that humans have utilized for life and various activities. Land use/land cover change (LULCC) has been of great concern to many countries over the years. Some of the main reasons behind LULCC are rapid population growth, migration, and the conversion of rural to urban areas. LULC has a considerable impact on the land-atmosphere/climate interactions. Over the past two decades, numerous studies conducted in LULC have investigated various areas of the field of LULC. However, the assemblage of information is missing for some aspects. Therefore, to provide coherent guidance, a literature review to scrutinize and evaluate many studies in particular topical areas is employed. This research study collected approximately four hundred research articles and investigated five (5) areas of interest, including (1) LULC definitions; (2) classification systems used to classify LULC globally; (3) direct and indirect changes of meta-studies associated with LULC; (4) challenges associated with LULC; and (5) LULC knowledge gaps. The synthesis revealed that LULC definitions carried vital terms, and classification systems for LULC are at the national, regional, and global scales. Most meta-studies for LULC were in the categories of direct and indirect land changes. Additionally, the analysis showed significant areas of LULC challenges were data consistency and quality. The knowledge gaps highlighted a fall in the categories of ecosystem services, forestry, and data/image modeling in LULC. Core findings exhibit common patterns, discrepancies, and relationships from the multiple studies. While literature review as a tool showed similarities among various research studies, our results recommend researchers endeavor to perform further synthesis in the field of LULC to promote our overall understanding, since research investigations will continue in LULC. 
    more » « less
  4. ; (, BioScience)
    Abstract Deforestation due to land-use and land-cover (LULC) change has been linked to increased emerging zoonotic disease risk despite limited local level data on such outbreaks. This Forum reevaluates this risk inference using newly released data on zoonotic disease outbreaks, accounting for Structural One Health features, including socioeconomic development and armed conflict covariates. Event and time series data on disease and forest coverage anomalies at the 0.5-degree level for every month between January 2003 and December 2018 are used to estimate the relationship between LULC and zoonosis using Poisson generalized additive and generalized linear models. Once adjusted for Structural One Health features, outbreak risk is 7%–200% higher in areas that experienced forest cover reversion. These results highlight the importance of accounting for Structural One Health factors when analyzing complex socioecological phenomena such as the LULC–infectious disease nexus. 
    more » « less
  5. ; ; ; ; ; (, Remote Sensing)
    Accurate estimation of land use/land cover (LULC) areas is critical, especially over the semi-arid environments of the southwestern United States where water shortage and loss of rangelands and croplands are affecting the food production systems. This study was conducted within the context of providing an improved understanding of New Mexico’s (NM’s) Food–Energy–Water Systems (FEWS) at the county level. The main goal of this analysis was to evaluate the most important LULC classes for NM’s FEWS by implementing standardized protocols of accuracy assessment and providing bias-corrected area estimates of these classes. The LULC data used in the study was based on National Land Cover Database (NLCD) legacy maps of 1992, 2001, 2006, 2011, and 2016. The analysis was conducted using the cloud-based geospatial processing and modeling tools available from System for Earth Observation Data Access, Processing, and Analysis for Land Monitoring (SEPAL) of the Food and Agricultural Organization. Accuracy assessment, uncertainty analysis, and bias-adjusted area estimates were evaluated by collecting a total of 11,428 reference samples using the Open Foris Collect Earth tool that provided access to high spatial and temporal resolution images available in Google Earth. The reference samples were allocated using a stratified random sampling approach. The results showed an overall accuracy that ranged from 71%–100% in all six study counties. The user’s and producer’s accuracy of most LULC classes were about or above 80%. The obtained bias-adjusted area estimates were higher than those based on pixel counting. The bias-adjusted area estimates simultaneously showed decreasing and increasing trends in grassland and shrubland, respectively in four counties that include Curry, Roosevelt, Lea, and Eddy during the 1992–2016 period. Doña Ana county experienced increasing and decreasing trends in grassland and shrubland areas, respectively. San Juan county experienced decreasing trends in both grassland and shrubland areas. Cultivated cropland areas showed decreasing trends in three counties in southeast NM that rely on groundwater resources including Curry, Roosevelt, and Lea. Similarly, cultivated cropland areas showed increasing trends in the other three counties that rely on surface water or conjunctive use of surface and groundwater resources including San Juan, Doña Ana, and Eddy. The use of SEPAL allowed for efficient assessment and production of more accurate bias-adjusted area estimates compared to using pixel counting. Providing such information can help in understanding the behavior of NM’s food production systems including rangelands and croplands, better monitoring and characterizing NM’s FEWS, and evaluating their behavior under changing environmental and climatic conditions. More effort is needed to evaluate the ability of the NLCD data and other similar products to provide more accurate LULC area estimates at local scales. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email