[Estimation of sparse vegetation coverage in arid region based on hyperspectral mixed pixel decomposition]

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Ecology

Context:

• Assessing sparse vegetation coverage (< 40%) is crucial for understanding arid and semi-arid ecosystems.
• The Minqin oasis-desert transitional zone in Gansu Province faces ecological challenges.
• Accurate vegetation mapping is vital for land management and conservation efforts in these fragile environments.

Purpose:

• To estimate sparse vegetation coverage in the Minqin oasis-desert transitional zone using hyperspectral imagery.
• To compare the effectiveness of fully constrained and non-constrained linear spectral mixture models (LSMM) for vegetation fraction estimation.
• To validate the accuracy of the LSMM methods against field-measured vegetation coverage.

Summary:

• Hyperion hyperspectral data were utilized with shifting sand, false-Gobi, and sparse vegetation spectra as endmembers.
• A fully constrained linear spectral mixture model (LSMM) accurately estimated sparse vegetation distribution, with errors less than 5% and an RMSE of 3.0681.
• The non-constrained LSMM showed poor correlation (R2 = 0.5855) and underestimated vegetation coverage, indicating its lower reliability for this application.

Impact:

• The fully constrained LSMM provides a more accurate and reliable method for estimating sparse vegetation coverage in arid transitional zones.
• This approach has significant potential for future applications in ecological monitoring and land resource management.
• Findings offer improved insights into vegetation dynamics in challenging desert-bordering environments.