Dynamics of the relationship between NDVI and SWIR32 vegetation indices in southern Africa: implications for retrieval of fractional cover from MODIS data

Hill, Michael J.; Zhou, Qiang; Sun, Qingsong; Schaaf, Crystal B.; Southworth, Jane; Mishra, Niti B.; Gibbes, Cerian; Bunting, Erin; Christiansen, Thomas B.; Crews, Kelley A.

doi:10.1080/01431161.2016.1154225

Cited by 17 publications

(4 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second most influential variable was the median of SWIR21, the ratio SWIR band 2/SWIR band 1 (see Table 1). Sentinel 2-derived SWIR21 is analogous to the MODIS-derived SWIR32 (ratio of MODIS SWIR bands 3 and 2), which has been found to be correlated with cellulose absorption index (CAI), derived from hyperspectral data and used primarily in remote sensing of dry/senescent biomass (Guerschman et al, 2009;Hill et al, 2016Hill et al, , 2017. We therefore postulate that SWIR21 in our model correlates to the abundance and persistence of (dry) herbaceous biomass, and that its relatively high importance for predicting %WCC arises indirectly from the competitive interactions between trees and grasses in mesic savannas (Scholes and Archer, 1997;Dohn et al, 2013;Kahiu and Hanan, 2018a).…”

Section: Variable Importancementioning

confidence: 99%

Toward Operational Mapping of Woody Canopy Cover in Tropical Savannas Using Google Earth Engine

Anchang¹,

Prihodko²,

Ji³

et al. 2020

Front. Environ. Sci.

View full text Add to dashboard Cite

Savanna woody plants can store significant amounts of carbon while also providing numerous other ecological and socio-economic benefits. However, they are significantly under-represented in widely used tree cover datasets, due to mapping challenges presented by their complex landscapes, and the underestimation of woody plants by methods that exclude short stature trees and shrubs. In this study, we describe a Google Earth Engine (GEE) application and present test case results for mapping percent woody canopy cover (%WCC) over a large savanna area. Relevant predictors of %WCC include information derived from radar backscatter (Sentinel-1) and optical reflectance (Sentinel-2), which are used in conjunction with plot level %WCC measurements to train and evaluate random forest models. We can predict %WCC at 40 m pixel resolution for the full extent of Senegal with a root mean square error of ∼8% (based on independent sample evaluation). Further examination of model results provides insights into method stability and potential generalizability. Annual median radar backscatter intensity is determined to be the most important satellite-based predictor of %WCC in savannas, likely due to its relatively strong response to non-leaf structural components of small woody plants which remain mostly constant across the wet and dry season. However, the best performing model combines radar backscatter metrics with optical reflectance indices that serve as proxies for greenness, dry biomass, burn incidence, plant water content, chlorophyll content, and seasonality. The primary use of GEE in the methodology makes it scalable and replicable by end-users with limited infrastructure for processing large remote sensing data.

show abstract

Section: Variable Importancementioning

confidence: 99%

Toward Operational Mapping of Woody Canopy Cover in Tropical Savannas Using Google Earth Engine

Anchang¹,

Prihodko²,

Ji³

et al. 2020

Front. Environ. Sci.

View full text Add to dashboard Cite

show abstract

“…In case that pure NPV and soil are spectrally too similar and thus do not form a triangular feature space, it is likely that the framework cannot be transferred. Hill et al (2016) and Hill et al (2017) estimated PV, NPV, and soil fractional cover in two savanna ecosystems in southern Africa and Brazil, respectively. Their results suggested high uncertainties for estimating NPV and soil fractional cover in heterogenous vegetation types with a complex phenology.…”

Section: Transferability Of Regression-based Unmixing For Drought Analysismentioning

confidence: 99%

Quantifying drought effects in Central European grasslands through regression-based unmixing of intra-annual Sentinel-2 time series

Kowalski

Okujeni

Brell

et al. 2022

Remote Sensing of Environment

View full text Add to dashboard Cite

Severe droughts caused unprecedented impacts on grasslands in Central Europe in 2018 and 2019. Yet, spatially varying drought impacts on grasslands remain poorly understood as they are driven by complex interactions of environmental conditions and land management. Sentinel-2 time series offer untapped potential for improving grassland monitoring during droughts with the required spatial and temporal detail. In this study, we quantified drought effects in a major Central European grassland region from 2017 to 2020 using a regression-based unmixing framework. The Sentinel-2-based intra-annual time series of photosynthetic vegetation (PV), nonphotosynthetic vegetation (NPV), and soil fractional cover provide easily interpretable quantities relevant for understanding drought effects on grasslands. Fractional cover estimates from Sentinel-2 matched in-situ conditions observed during field visits. The comparison to a multitemporal reference dataset showed the best agreement for PV cover (MAE = 7.2%). Agreement was lower for soil and NPV, but we observed positive relationships between fractional cover from Sentinel-2 and the reference data with MAE = 10.1% and MAE = 15.4% for soil and NPV, respectively. Based on the fractional cover estimates, we derived a Normalized Difference Fraction Index (NDFI) time series contrasting NPV and soil cover relative to PV. In line with meteorological and soil moisture drought indices, and with the Normalized Difference Vegetation Index (NDVI), NDFI time series showed the most severe drought impacts in 2018, followed by less severe, but persisting effects in 2019. Drought-specific metrics from NDFI time series revealed a high spatial variability of onset, duration, impact, and end of drought effects on grasslands. Evaluating drought metrics on different soil types, we found that grasslands on less productive, sandy Cambisols were strongly affected by the drought in 2018 and 2019. In comparison, grasslands on Gleysols and Histosols were less severely impacted suggesting a higher drought resistance of these grasslands. Our study emphasizes that the high temporal and spatial detail of Sentinel-2 time series is mandatory for capturing relevant vegetation dynamics in Central European lowland grasslands under drought.

show abstract

“…For this reason, since the late 1990s, there have been many papers in the remote sensing literature which have applied linear mixture models (LMMs) to data from multispectral sensors to estimate the proportions of PV, NPV and BS [1][2][3][4][17][18][19][20][21][22][23][24]. Okin [25] used the three endmembers PV, NPV and snow, while in [26], they are shade, vegetation and other landforms (including water, rock and sand).…”

Section: Introductionmentioning

confidence: 99%

“…Typically, the endmembers are obtained from the dataset itself. Others obtain the endmembers by using suitable indices, such as NDVI and the SWIR32 vegetation index, to create a suitable two-dimensional space in which to unmix the data [3,4,[21][22][23][24].…”

Section: Introductionmentioning

confidence: 99%

Confidence Intervals and Regions for Proportions under Various Three-Endmember Linear Mixture Models

Berman

2023

Remote Sensing

View full text Add to dashboard Cite

Many studies in recent years have been devoted to estimating the per-pixel proportions of three broad classes of materials (e.g., photosynthetic vegetation, non-photosynthetic vegetation and bare soil) using data from multispectral sensors. Oftentimes, the estimated proportions are used to monitor environmental change in both urban and non-urban environments. Many of these papers use proportion estimation methods based on the linear mixture model. Very few of these papers assess the accuracy of their estimators. This paper shows how to produce confidence intervals (CIs) and joint confidence regions (JCRs) for the proportions associated with various linear mixture models. There are two main models, both of which assume that the coefficients in the model are non-negative. The first model assumes that the coefficients sum to 1. The second does not, but uses rescaling of the estimated coefficients to produce estimated proportions. Three variants of these two models are also analysed. JCRs are shown to be particularly informative, because they are typically better at localising the information than CIs are. The methodology is illustrated using examples from Landsat Thematic Mapper data at 1169 locations across Australia, each of which has associated field observations. There is also a discussion about the extent to which the methodology can be extended to hyperspectral data.

show abstract

Dynamics of the relationship between NDVI and SWIR32 vegetation indices in southern Africa: implications for retrieval of fractional cover from MODIS data

Cited by 17 publications

References 48 publications

Toward Operational Mapping of Woody Canopy Cover in Tropical Savannas Using Google Earth Engine

Toward Operational Mapping of Woody Canopy Cover in Tropical Savannas Using Google Earth Engine

Quantifying drought effects in Central European grasslands through regression-based unmixing of intra-annual Sentinel-2 time series

Confidence Intervals and Regions for Proportions under Various Three-Endmember Linear Mixture Models

Contact Info

Product

Resources

About