Mowing or Planting Trees? Japan’s Abandoned Rice Paddies Face a Climate and Biodiversity Trade-Off

Across rural Japan, a quiet transformation is underway. As populations age and shrink, farmers are walking away from terraced rice paddies and small mountain plots that their families have tended for generations. A new study from the University of Tokyo, published in Regional Environmental Change, suggests that what happens to this abandoned land in the coming decades could determine whether depopulating countryside becomes a climate asset or a biodiversity casualty. The research offers one of the first quantitative assessments of management strategies that sit between full-scale farming and complete abandonment, and its findings carry weight far beyond Japan’s rice fields.

The study focused on Fukui Prefecture, a mountainous region on the Sea of Japan coast where forests cover 74 percent of the land and rice paddies dominate the agricultural landscape. The prefecture’s population peaked at roughly 829,000 in 2000, fell to about 767,000 by 2020, and is projected to drop to 573,000 by 2050. That demographic decline has already driven rapid farmland abandonment, eroding the traditional Satoyama landscape, a mosaic of paddies, forests, irrigation canals, ponds, and villages that has sustained both people and wildlife for centuries. Because rice paddies function as surrogate wetlands, their loss hits aquatic and semi-aquatic species particularly hard. A meta-analysis of Japanese studies found that species richness declined by an average of 72 percent after paddy abandonment, and plant diversity often failed to recover even 10 to 15 years later.

To explore what could be done, researchers Taira Ishiguro and Shizuka Hashimoto built three spatially explicit scenarios for the year 2050. The reference scenario assumed that farmland would either be cultivated conventionally or abandoned outright, with total farmland shrinking from about 50,317 hectares in 2016 to roughly 31,303 hectares, a loss of nearly 37 percent. The second scenario, called extensive management, assumed that 10 percent of farmland would be maintained through periodic mowing, preserving paddy levees, plow soles, and irrigation canals without any agricultural production. The third scenario, forestation, assumed that 10 percent of farmland would be deliberately afforested and converted to managed forestry, a strategy already encouraged by Japan’s carbon credit system.

The methodological machinery behind these projections was considerable. The team analyzed land use change between 2006 and 2016 using the Land Change Modeler in TerrSet, employing a multi-layer perceptron neural network to capture the complex, nonlinear relationships driving abandonment. Seventeen explanatory variables were tested, including slope, elevation, and the density and proximity of different land use types within a 500-meter radius. The final abandonment potential model, built on just five variables, achieved an accuracy of 80.51 percent and a skill measure of 0.7401, indicating performance well above random chance. The model revealed a striking spatial pattern: small, scattered farmlands along mountain valley lines showed markedly higher abandonment potential, while large, clustered fields in the lowland plains were far more resilient.

With future land use maps in hand, the researchers evaluated three ecosystem services using the InVEST modeling suite: carbon sequestration, water purification, and water yield. Biodiversity was assessed through landscape heterogeneity, measured as the total edge length between farmland and forest within 500-meter grid cells, a proxy that reflects the structural complexity on which Satoyama biodiversity depends. Vegetation succession was explicitly modeled, with abandoned farmland remaining grassland for roughly a decade before transitioning linearly to forest over the following 25 years, mirroring observed dynamics in Japanese landscapes.

The results revealed a fundamental trade-off. All three scenarios showed the same directional trends between 2016 and 2050: carbon sequestration rose, nitrogen export to rivers fell, water yield declined, and landscape heterogeneity eroded. But the magnitude of change differed sharply. The forestation scenario delivered the largest carbon gains, adding 1,502 kilotons of carbon compared with 1,277 kilotons under the reference scenario, and achieved the biggest reduction in nitrogen export at 338 tonnes per year. The extensive management scenario performed nearly as well on both counts. Yet both alternative scenarios also produced larger declines in water yield, a consequence of increased evapotranspiration as vegetation developed, echoing a well-documented tension between regulating services and water provisioning in post-agricultural landscapes.

The biodiversity story was more nuanced. Landscape heterogeneity declined in every scenario, but the extensive management scenario preserved it best, losing 1,038 kilometers of farmland-forest edge compared with 1,359 kilometers under the reference scenario and 1,371 kilometers under forestation. The mechanism is elegant: periodic mowing keeps underused farmland as open grassland adjacent to remaining forest, sustaining the spatial complexity that open-habitat and wetland species require. Afforestation, by contrast, offered essentially no biodiversity advantage over simple abandonment, because it converts farmland into the same homogeneous forest cover that succession would eventually produce anyway.

Perhaps the most policy-relevant finding concerns geography. In remote mountainous regions, the fate of farmland was largely sealed regardless of scenario, converging toward forest cover as abandonment proceeded along valley lines. The real divergence occurred in peripheral lowland areas, the marginal fields surrounding the plains. There, extensive management maintained grassland habitats and kept ecosystem service changes moderate, while forestation paradoxically accelerated abandonment in these same lowland margins. Because the researchers held total agricultural labor constant across all scenarios, every hectare allocated to mowing or tree planting meant labor diverted from conventional farming, pushing marginal lowland fields over the abandonment threshold. This spatial dichotomy means that the consequences of choosing between mowing and planting will be felt most intensely not in the remote mountains, but in the transitional zones where farmland and forest still intermingle.

The implications reach well beyond Fukui. Global fertility projections suggest that rural depopulation will spread to an increasing number of countries over the coming decades, making land underuse a growing worldwide phenomenon. In Europe, debates over rewilding versus extensive re-farming have generated rich literature, but East Asian paddy landscapes have remained understudied despite their outsized biodiversity role as surrogate wetlands. This study demonstrates that the European framing of abandonment as a rewilding opportunity does not transfer cleanly to rice-farming regions, where abandonment consistently erodes biodiversity. Instead, the findings argue for spatially explicit policy: deploying periodic mowing to conserve traditional landscape structure where it matters most, while directing afforestation toward remote areas where abandonment is inevitable and carbon gains can be maximized.

The authors acknowledge limitations, including the absence of technological change in their labor estimates, the use of proxy parameters for abandoned and extensively managed land, and the reliance on landscape heterogeneity rather than direct species data. Field-based verification remains scarce, particularly in Japan. Still, the core message stands: the binary choice between farming and abandoning land conceals a spectrum of intermediate options with measurably different ecological outcomes. As governments grapple simultaneously with climate mitigation, biodiversity loss, and rural decline, this research suggests that agricultural policy, environmental policy, and spatial planning can no longer be designed in isolation. The humble act of mowing an unused paddy, it turns out, may be one of the cheapest biodiversity conservation tools a depopulating nation possesses.

Subject of Research: Impacts of alternative land management options on biodiversity and ecosystem services in depopulating rural rice-farming landscapes of Japan

Article Title: Alternative land management in depopulating rural landscapes: Scenario analysis of impacts on biodiversity and ecosystem services

Article References: Ishiguro, T., & Hashimoto, S. (2026). Alternative land management in depopulating rural landscapes: Scenario analysis of impacts on biodiversity and ecosystem services. Regional Environmental Change, 26(3), Article 182. https://doi.org/10.1007/s10113-026-02659-y

Image Credits: AI Generated

DOI: 10.1007/s10113-026-02659-y

Keywords: farmland abandonment, rice paddies, Satoyama, biodiversity, ecosystem services, carbon sequestration, water purification, afforestation, extensive management, rural depopulation, scenario analysis, landscape heterogeneity