Satellite forecasting of crop harvest can trigger a cross-hemispheric production response and improve global food security

The COVID-19 pandemic and the Russia–Ukraine war exposed fragilities in the global food system, compounding anticipated production shocks from climate change. Shortfalls in regional production can transmit globally through trade, elevating prices and harming consumption in distant, vulnerable import-dependent regions. Satellite Earth observations provide synoptic, timely monitoring of crop conditions and have recently achieved national-scale pre-harvest yield forecasts with superior accuracy to many alternatives, 1–2 months before harvest. Given the complementarity of crop calendars between hemispheres, reliable early information about harvest outcomes in one hemisphere could induce supply adjustments in the opposite hemisphere, mitigating global price volatility. While producers in Brazil and the United States reportedly monitor each other’s growing seasons, systematic evidence linking satellite forecasts to cross-hemispheric production responses and market stabilization has been lacking. This study tests the hypothesis that timely satellite-based harvest forecasts in major exporting regions can trigger opposite-hemisphere production responses that stabilize global markets and improve food security.

The paper builds on decades of research using remote sensing for pre-harvest crop yield estimation, including biophysical crop models (e.g., CERES, WOFOST, CROPSYST, STICS) and statistical approaches linking vegetation indices (notably NDVI) and agrometeorological data to yields. Foundational studies demonstrated strong correlations between time-integrated or peak NDVI and cereal yields and progressed to generalized empirical models for national/state-level forecasting 1–2 months pre-harvest (e.g., Becker-Reshef et al.; Franch et al.; Skakun et al.). Reviews by Pearlman et al., Häggquist and Söderholm, and Leslie et al. highlighted a gap between remote sensing data usage and demonstrated societal outcomes or decision improvements. The study addresses this gap by quantifying how satellite forecasts can influence producer behavior across hemispheres and stabilize markets. Additional literature underpins the CGE modeling framework (Walras, Arrow–Debreu; GTAP-based applications) and market mechanisms (Armington, CET land allocation), as well as evidence on demand elasticities, land-use modeling, and the value of public information in storable commodity markets.

Case selection: Three cases were chosen where (1) a GTAP global SAM exists for year t−1, (2) reliable pre-harvest satellite forecasts exist for year t, and (3) harvest deviations are sufficiently large to affect global markets: (i) better-than-normal wheat harvest in Russia–Ukraine in 2008 vs 2007; (ii) poor wheat harvest in Russia–Ukraine in 2012 vs 2011; (iii) poor soybean harvest in southern Brazil (Paraná, Santa Catarina, Rio Grande do Sul) in 2012 vs 2011. Remote sensing forecasts: For wheat, the study applied a generalized regression-based model (Becker-Reshef et al.; subsequent enhancements) using MODIS-derived seasonal peak NDVI and administrative-unit wheat area fractions, with Growing Degree Day information to improve timeliness. Forecast snapshots for Russia were on 10 May and 9 June and for Ukraine on 5 May and 4 June, in 2008 and 2012. For soybean, a machine-learning classification model (Song et al.) using Landsat and MODIS mapped soybean cover at 30 m; mapped soybean area serves as a surrogate for production, as failed crops are typically not mapped. The soybean anomaly information was available by 30 April 2012. Forecast performance and shocks: Relative to prior-year production, satellite forecasts for Russia–Ukraine wheat indicated +48.0% (May 2008) and +45.0% (June 2008); −15.3% (May 2012) and −19.9% (June 2012). FAO records showed actual changes of +61.2% (2008) and −27.2% (2012). For Brazil soybean, official statistics indicated a national −12% driven by −35% in the southern states; the remote-sensing forecast indicated −44% in southern Brazil, corresponding to −18% nationally relative to 2011. Forecast accuracies for wheat vs FAO were 78% and 74% (May/June 2008) and 56% and 73% (May/June 2012), outperforming USDA WASDE for the same periods. CGE framework: Three global computable general equilibrium (CGE) models were constructed using GTAP v10 SAMs for 2007 and 2011, aggregated to focus on major wheat/soybean trading regions and food-related sectors. The model features Leontief technology for intermediate inputs; CES aggregation of primary factors (labor, capital, farmland); Armington differentiation of imports with halved short-run elasticities for food-related sectors; CET-based land allocation across livestock vs crops, crop groups, and within cereals/oilseeds (wheat, coarse grains, oilseeds/soybean) to capture short-run planting responses. Household consumption uses CES substitution calibrated to food demand elasticities. Elasticities follow GTAP and literature (including Haile et al. for short-run acreage responses). Short-run Armington elasticities were halved per Bajzik et al. to reflect short-run conditions. Scenarios and estimation procedure: Base years were 2007 (wheat case 2008) and 2011 (wheat case 2012; soybean case 2012). Real-shock scenarios (W-Real08, W-Real12, S-Real12) applied the realized yield/production shocks to Russia–Ukraine (wheat) or Brazil (soybean) with factors immobile elsewhere, representing information arriving too late for opposite-hemisphere adjustments. Response scenarios (W-May08, W-June08, W-May12, W-June12; S-April12) applied forecast shocks ahead of opposite-hemisphere planting windows and allowed factor mobility and land reallocation within agronomic calendar constraints to form an intermediate equilibrium. Then the realized shocks were introduced to produce a final equilibrium with factors fixed, representing actual harvest outcomes following proactive adjustments. Outcomes compared include prices (local and CIF), household consumption, and production. Robustness: Sensitivity analyses varied key elasticities (±30% for Armington in grain/oilseeds, value-added substitution in food sectors, household substitution, land transformation) and halved production factor substitution elasticities in food-related sectors. Qualitative results were robust, with moderate quantitative variation across tests. Data: MODIS Surface Reflectance (MOD09Q1), Brazil soybean maps (GLAD), GTAP v10 SAMs, FAOSTAT production and GIEWS prices, and CONAB subnational soybean statistics.

- Satellite forecasts accurately signaled major harvest deviations: For Russia–Ukraine wheat, forecasts captured increases of +48%/+45% (May/June 2008) vs actual +61.2%, and declines of −15.3%/−19.9% (May/June 2012) vs actual −27.2%. For Brazil soybean in 2012, a −44% loss in southern Brazil (≈−18% nationally) was forecast vs official −35% in the south (−12% national). - Opposite-hemisphere production responses: In 2008 (good RU–UA wheat), Southern Hemisphere wheat output would decrease by about 2.83 and 2.70 million tons in W-May08 and W-June08 (≈2.3% and 2.2% of 2007 global wheat exports). In 2012 (poor RU–UA wheat), Southern Hemisphere wheat output would increase by about 2.61 and 3.47 million tons in W-May12 and W-June12 (≈1.8% and 2.3% of 2011 global wheat exports). For soybean 2012 (poor southern Brazil), US and Canada producers would expand soybean output by 5% and 7%, totaling 4.4 million tons (≈4.8% of 2011 global soybean exports). - Price stabilization in importing countries (ceteris paribus): For wheat importers in 2008, local price drops under W-Real08 (−27% to −35%) were moderated by 1.1–12.5 percentage points under response scenarios (e.g., Japan −28.5% to −20.6% with W-May08). In 2012, price spikes under W-Real12 (+12% to +31%) were reduced by 5.0–12.5 percentage points (e.g., Korea +24.1% to +11.6% with W-June12). Household wheat consumption fluctuations were reduced by 0.3–2.5 pp (2008) and 0.8–2.0 pp (2012). - Soybean price impacts (2012): Under S-Real12, local soybean prices rose by +12.2% to +15.9% across major importers. Under S-April12, increases were mitigated to between −6.4% and +8.3%, with some importers seeing price declines (Mexico −6.4%, Japan −4.5%, Indonesia −3.2%). Household soybean consumption variation narrowed by 1.3–5.0 percentage points (e.g., China from −2.9% to +2.1%). - CIF import prices: Wheat CIF price volatility was smoothed by 1.2–9.3 pp in 2008 and 5.5–12.6 pp in 2012 relative to real-shock scenarios, reinforcing stabilization benefits at the border. - Overall, timely satellite forecasting coupled with cross-hemispheric planting responses can materially dampen global price volatility and consumption shocks in importing countries.

The findings confirm that timely, reliable satellite-based pre-harvest forecasts can trigger economically meaningful supply responses in the opposite hemisphere due to seasonal lags in crop calendars, thereby stabilizing international and domestic markets. By enabling producers to anticipate and partially offset shocks (surpluses or deficits) in major exporting regions, the mechanism reduces price volatility and moderates impacts on household consumption in importing countries. Market stabilization manifests both locally and at the border (CIF prices), especially benefiting import-dependent nations. The inter-hemispheric mechanism delivers three broad economic benefits: (1) risk mitigation for risk-averse producers and consumers via reduced price volatility (lower risk premia); (2) improved allocative efficiency through better information, reducing deadweight losses and enhancing factor allocation (labor, capital, land); and (3) potential dampening of speculative inflows when credible pre-harvest information reduces expected excess returns from commodity speculation. Policy relevance is significant: the mechanism relies on efficient price transmission between international and domestic markets. Trade restrictions (export bans/quotas or import barriers) impede transmission and diminish stabilization effects. Conversely, lowering import barriers in food-deficit countries can enhance benefits. Supporting R&D to improve the timeliness and accuracy of satellite forecasts (ideally exceeding 90% accuracy two months pre-harvest) would increase producer confidence and response magnitude, further strengthening global food security.

This study demonstrates a concrete pathway by which satellite Earth observations, through timely and accurate pre-harvest forecasts, can activate cross-hemispheric production adjustments that stabilize global agricultural markets. Applying this insight to wheat (Russia–Ukraine) and soybean (southern Brazil) cases, the analysis shows substantial reductions in price volatility and moderated consumption impacts in major importing countries. The approach leverages predictable seasonal offsets in crop calendars and integrates remote-sensing intelligence into forward-looking planting decisions captured within a global CGE framework. Future work should: (1) improve early-season forecast accuracy and dissemination to broaden producer uptake; (2) extend the modeling framework to recursive semi-annual dynamics to assess intra-annual price volatility; (3) evaluate second-stage market impacts involving speculative behavior; and (4) incorporate more responsive regions/sectors and operational constraints to refine welfare and stabilization estimates.

- Producer adjustment constraints: The model assumes producers can reallocate land and adjust crop mixes in response to forecasts, potentially overestimating responses due to real-world constraints (e.g., crop rotation, agronomic prerequisites, farm-level commitments). - Speculative activity not fully captured: The simulations isolate harvest-driven price effects; in reality, speculation could amplify initial reactions and, if anticipated by farmers, could increase cross-hemispheric adjustments. Thus, benefits may be underestimated. - Limited set of responding countries: Only a subset of opposite-hemisphere producers (five in the Southern Hemisphere for wheat; two in the Northern Hemisphere for soybean) are modeled as responsive. Wider participation would likely yield larger stabilization effects, implying potential underestimation. - Annual time step: The annual CGE setup treats production in one hemisphere as exogenous within the year and only allows opposite-hemisphere adjustments, focusing on a particular trajectory. A recursive semi-annual model could better capture intra-annual dynamics and price volatility responses to pre-harvest information.