Desertification Measured by the Wrong Map

Patagonian steppe covers approximately 800,000 km² of southern Argentina, and an estimated 90% shows some degree of desertification driven by a century of overgrazing by sheep ranching. Measuring the extent and progression of this degradation is essential for management, but standard desertification assessment indices — UNCCD's LADA framework, NDVI-based remote sensing metrics, the MEDALUS approach — were developed for Saharan/Sahelian and Mediterranean ecosystems and produce unreliable or misleading readings in Patagonia. CONICET's IADIZA laboratory has documented that NDVI (Normalized Difference Vegetation Index), the most widely used remote sensing metric for desertification monitoring, fails in Patagonia because the dominant vegetation (low shrubs, cushion plants, and cryptogamic crusts) has naturally low and highly variable NDVI signatures that overlap with the signatures of degraded land. A healthy Patagonian steppe and a severely degraded one can have identical NDVI readings.

Patagonia is one of the world's largest extents of active desertification, but it receives a fraction of the monitoring attention directed at the Sahel, Central Asia, or northern China because its desertification doesn't register on the global indices designed for those regions. This invisibility has policy consequences: Argentina cannot accurately report on UNCCD commitments, grazing management decisions are made without reliable degradation data, and degradation trends go undetected until they become visually obvious (at which point the land is often beyond recoverable degradation thresholds). The sheep ranching industry that drives degradation is also the primary livelihood for remote Patagonian communities — understanding where degradation is accelerating versus where it's stable is essential for sustainable land management, not just conservation.

IADIZA has developed Patagonia-specific indicators using soil crust integrity, patch/inter-patch dynamics (landscape ecology metrics for measuring vegetation distribution patterns), and wind erosion proxies. These work scientifically but require field measurement — they cannot be automated from satellite data, which means they can't be scaled across Patagonia's vast, sparsely inhabited extent. High-resolution satellite imagery (Sentinel-2, Planet) offers better spatial detail than Landsat-era NDVI, but the spectral signatures of Patagonian degradation (loss of biological soil crusts, shift from grass to shrub dominance, exposure of underlying volcanic substrate) haven't been systematically characterized to enable automated classification. Argentine land management agencies use LADA indicators for reporting because UNCCD requires them, even though IADIZA's research demonstrates these indicators are unreliable for Patagonia — creating a reporting-science disconnect where official data contradicts research findings.

Developing Patagonia-specific remote sensing indices — calibrated to the spectral and structural characteristics of Patagonian vegetation and soils — would enable automated, large-scale degradation monitoring. CONICET's IADIZA lab has identified the key spectral targets: biological soil crust health (which has distinctive absorption features in shortwave infrared), shrub-to-grass ratio (detectable with hyperspectral data), and surface roughness changes from wind erosion (detectable with radar). Combining these with IADIZA's validated field indicators to create a calibrated, scalable monitoring framework is technically feasible but requires the kind of dedicated, Patagonia-specific remote sensing campaign that global earth observation programs have not prioritized.

A remote sensing team could use freely available Sentinel-2 data to characterize the spectral signatures of known-degraded versus known-intact Patagonian steppe sites (using IADIZA's field data as ground truth) to evaluate which spectral indices discriminate degradation in this ecosystem. A data science team could build a degradation classification model trained on IADIZA's field measurements and satellite imagery, testing whether machine learning can overcome the limitations of single-index approaches like NDVI. An environmental science team could compare desertification assessment frameworks across analog environments (Patagonia, Icelandic rangeland, Central Asian steppe) to identify which monitoring approaches transfer and which are ecosystem-specific.