Satellite detection of earlier wheat sowing in India and implications for yield trends

David B. Lobell, Iván Ortiz‐Monasterio, Adam Sibley, V. S. SohuAgricultural 2012 · DOI

TL;DR

This study uses satellite imagery to track wheat sowing dates across India, revealing a significant trend towards earlier sowing in key agricultural regions. This shift is estimated to have boosted wheat yields by over 5% in the last decade, potentially masking a stagnation or even decline in yields from other factors.

What Problem Does This Paper Address?

Wheat is a crucial staple crop in South Asia, yet its yield growth has slowed in recent years, contributing to food security concerns in a region with a large malnourished population. One proposed solution to boost yields is earlier sowing, as planting wheat outside its optimal window leads to significant yield losses due to heat stress later in the season. While practices like zero-till farming (ZRT) enable earlier sowing, there's a lack of consistent, large-scale data to track these trends and their impact on yield. This paper aims to fill that gap by using satellite data to quantify sow date trends across India's wheat-growing regions and estimate their contribution to yield changes.

Methods

The researchers utilized vegetation index (VI) products from two satellite sensors: MODIS (Moderate Resolution Imaging Spectroradiometer) and SPOT (Satellite Pour l'Observation de la Terre Vegetation). They employed the Timesat software to fit double logistic functions to the time-series data, allowing them to estimate the "green-up" date – defined as the point when a fitted curve reaches 10% of its maximum amplitude. This green-up date serves as a proxy for sowing date, with simulations using the CERES-Wheat model showing a strong correlation (R² = 0.98) and an average offset of about three weeks between green-up and actual sowing. Pixel-level estimates were aggregated to district and state levels, with national averages weighted by production. Trends were analyzed using linear regression on these estimated sow dates from 2001 to 2010.

Datasets

  • MODIS VI products (MOD13A2 and MYD13A2): Provided EVI (Enhanced Vegetation Index) data at 8-day intervals with a 1 km spatial resolution.
  • SPOT-VGT VI products: Provided NDVI (Normalized Difference Vegetation Index) data at 10-day intervals with a 1 km spatial resolution.
  • Monfreda et al. (2008) global map of wheat harvested area: Used to define the study area (regions with at least 40% wheat cultivation).
  • CERES-Wheat model simulations: Used to establish the relationship between satellite-detected green-up and actual sowing dates.
  • Punjab Agricultural University ground-based survey data: Used for validation of satellite estimates in Punjab.

Key Insights and Ideas

  • Satellite data as a powerful tool for agricultural monitoring: The study demonstrates the efficacy of readily available satellite data for tracking large-scale agricultural practices like sowing dates, offering a cost-effective and objective alternative to limited ground surveys.
  • Earlier sowing is a significant driver of recent wheat yield trends: The research suggests that the observed shift towards earlier sowing has been a primary contributor to wheat yield increases in India over the past decade, potentially masking stagnation or decline from other factors.
  • Regional variations in sowing trends: While Haryana and Uttar Pradesh show clear trends towards earlier sowing, Punjab has seen little significant change, leading to a convergence of sowing dates across these key states.
  • The impact of ZRT and other technologies: While ZRT is a likely contributor to earlier sowing, the study hints that other factors like rotavator adoption and shorter-maturing crop varieties may also play a role, especially in regions where ZRT adoption rates are lower than expected for the observed sowing shifts.
  • Diminishing returns from earlier sowing: The study highlights that as sowing dates approach the optimal window, the potential for further yield gains from earlier sowing will likely decrease.

Key Results

  • Statistically significant earlier sowing: Significant (p < 0.05) negative trends in sowing dates were observed in Haryana and Uttar Pradesh.
  • National average shift: The production-weighted average sow date across India shifted earlier by approximately one week by 2010 compared to the beginning of the decade.
  • Estimated yield gain: The earlier sowing trend is estimated to have increased wheat yields by at least 5% on average across India since 2000.
  • Counteracting yield stagnation: This 5% gain from earlier sowing is substantial, as the overall national yield increase over the same period was less than 5%. This implies that other factors may have led to a net yield effect close to zero or even negative.
  • Punjab's stable sowing dates: In Punjab, sowing date trends were positive but not statistically significant, indicating little change.
  • Convergence of sow dates: By 2010, average sow dates in Punjab and Haryana were nearly the same, with both states nearing the optimal window.

Limitations and Future Work

  • Limitations of green-up as a proxy: While well-correlated, green-up date is an indirect measure of sowing date and is influenced by factors beyond just the sowing event (e.g., emergence speed, early growth conditions).
  • Attributing causes of sowing shifts: While ZRT is a likely driver, the study acknowledges that other factors contribute to earlier sowing, and more detailed ground data is needed to fully understand these causal links, particularly in Uttar Pradesh.
  • Diminishing future gains: The study suggests that the potential for significant yield increases from further shifts to earlier sowing will likely diminish in Haryana and Punjab as they approach optimal sowing windows. Future work could focus on understanding the remaining potential in regions like UP and the impact of other factors on yield.
  • Applicability to other regions: The methodology is presented as being directly applicable to other major cropping regions globally, suggesting future work could expand this analysis.

Summary updated

BibTeX
@inproceedings{lobell2012satellite,
  author = {David B. Lobell and Iván Ortiz‐Monasterio and Adam Sibley and V. S. Sohu},
  title = {{Satellite detection of earlier wheat sowing in India and implications for yield trends}},
  booktitle = {Agricultural},
  year = {2012},
  doi = {10.1016/j.agsy.2012.09.003},
}
Figure