That a hyperarid desert could become a carbon sink was long considered scientifically improbable. A study published in the Proceedings of the National Academy of Sciences (PNAS) in January 2026, based on 25 years of NASA satellite data, proves it has happened in the Taklamakan: the planted margins of one of Earth's driest deserts now absorb more carbon dioxide than they release. The basis is a Chinese state program spanning 46 years: on November 28, 2024, workers pressed the final seedlings into sand and closed a 3,046-kilometer protective belt. The result is scientifically significant, yet the fuller picture is more complex.
Four decades for 337,600 square kilometers
The Taklamakan in the Xinjiang Basin covers 337,600 square kilometers, making it the world's second-largest shifting sand desert. Annual rainfall: less than 100 millimeters. Annual evaporation: more than 2,500 millimeters. In 1978, the Chinese government launched the Three-North Shelterbelt Program, designed to span multiple decades and ending in 2050. Its goal: stop sand encroachment and protect farmland and settlements.
The Taklamakan Ring is the program's centerpiece. On 1.6 million hectares along the desert's margins, workers planted mainly poplars, tamarisks, and native shrubs. China's forestry administration counts over 66 billion trees planted across northern China. The national forest cover rose from roughly 10 percent in the 1950s to 25 percent today. Satellite measurements show the desert area shrank by some 5,300 square kilometers between 2003 and 2022, particularly at the western margin.
What NASA satellites measured
The research team led by Yuk Yung from the California Institute of Technology (Caltech) used NASA's Orbiting Carbon Observatory-2 (OCO-2) and solar-induced fluorescence measurements, which detect photosynthetic activity directly from space. The team analyzed 25 years of data.
The finding: during the growing season, CO2 concentration over the planted desert margins drops from 416 to 413 ppm. King-Fai Li of the University of California Riverside, a coauthor, describes the vegetation matter-of-factly: it is "not tropical rainforest ecology like the Amazon or Congo; some reforested zones contain only shrub vegetation." But the vegetation is photosynthetically active. The theoretical maximum if the entire desert were planted would be 58.7 million tons of CO2 per year. For perspective: globally, roughly 40 billion tons are emitted annually, making that 0.15 percent. The study does not explicitly quantify the current marginal plantings' actual contribution; it lies well below the theoretical maximum.
Senior author Yung states the scientific core: "We have shown for the first time that human intervention can measurably boost carbon sequestration even in extreme dryland environments."
What critics point out
The program has documented scientific weaknesses. Shixiong Cao from Beijing Forestry University found that in drylands, up to 85 percent of plantings fail to survive long-term. Poplar monocultures, long favored for rapid growth, die in parts of the belt at rates of 120,000 hectares per year from pest damage. In 2000, a single fungal disease destroyed one billion poplars in Ningxia Province.
The structural problem lies in water balance. A 2025 study in the journal "Earth's Future" from Tianjin University and Utrecht University found that while China's landscape changes increased evaporation, net available water declined by 0.46 millimeters per year. Jiang Gaoming from the Chinese Academy of Sciences argues the program in drylands without adapted native plant species is structurally flawed. An alternative model within China is the Loess Plateau project: on 2.5 million hectares, native grasses and shrubs adapted to local water conditions were planted. Ecologists view it as more durable than poplar-dominated portions of the Three-North Belt.
In comparison: Other green walls worldwide
Africa's Great Green Wall aims to restore 100 million hectares in the Sahel by 2030 and sequester 250 million tons of CO2. By 2024, roughly 30 million hectares were restored, 30 percent of the goal. Financing remains the central barrier: $33 billion was pledged but substantially less delivered.
Israel's Yatir Forest at the Negev Desert's edge has been the world's most-studied dryland project since 1964, with four million trees on 30 square kilometers. Its carbon uptake is scientifically proven. Critics fault the use of non-native pine species and land conflicts with Bedouin communities. Both projects show the same pattern as the Taklamakan: planting in drylands is feasible and measurably effective, but efficiency rises substantially with native mixed vegetation.
What other desert projects can learn from the Taklamakan Ring
The PNAS study provides a NASA-satellite-reproducible measurement framework for dryland projects globally. Future projects in other hyperarid regions can use the same method for independent success verification without relying on operator self-reports.
The central lesson from China's mistakes is clear: poplar monocultures rarely survive long-term in drylands, native mixed vegetation does. Whether China switches to this approach in its ongoing program expansion to 4.45 million hectares will determine how many of the 66 billion planted trees still stand in a generation.