Forecasts of fog events in northern India dramatically improve when weather prediction models include irrigation effects

The study addresses why wintertime fog frequency and severity have increased over northern India, particularly the Indo-Gangetic Plains (IGP) and Delhi, and whether increased winter irrigation is a decisive driver. Fog in the IGP is common in winter and has major societal and economic impacts. Observations show a 71% increase in winter fog days in Delhi over three decades (from 38 in the 1980s to 65 recently), alongside deteriorating air quality. While aerosols can modify fog microphysics and radiative properties, air pollution alone does not explain the observed fog trends. A parallel rapid increase in low-level humidity has been observed (e.g., Safdarjung airport December–January mean dew point increased from 7 °C in 1980 to 10.5 °C in 2019), suggesting an alternative driver. The authors hypothesize that expanded winter irrigation for Rabi cropping has increased soil moisture and near-surface humidity, promoting fog formation. Fog forecasting with NWP is challenging due to multiscale processes and parameterization uncertainties, with known near-surface dry biases over the IGP in several models. The authors propose that missing irrigation in model land-surface representation is a key source of this dry bias, and test whether explicitly representing winter irrigation improves fog simulation.

Prior studies have used statistical, diagnostic, and NWP approaches for fog prediction over the IGP, yet reliable operational forecasts remain elusive. Fog simulations are highly sensitive to land-surface and boundary-layer parameterizations for onset, and microphysics for development and dissipation. Multiple NWP systems (e.g., WRF, MetUM) show systematic wintertime near-surface dry biases over the IGP, leading to missed fog events or timing errors. While soil moisture data assimilation using satellite products can reduce some biases, model climatologies lacking irrigation remain too dry, limiting correction effectiveness. The Green Revolution increased crop yields without large area expansion by adopting double-cropping, adding a dry-season (Rabi) crop requiring expanded irrigation; winter irrigation water use has trended upward by ≥10 mm/month/decade since the 1970s in NW IGP. Land-atmosphere impacts of irrigation in India have been widely studied for summer/monsoon, showing increased latent heat flux, reduced sensible heat, temperature impacts, and precipitation changes. However, prior work has not assessed winter irrigation impacts on fog formation; only limited studies discussed winter soil moisture anomalies’ broader circulation impacts. This study fills that gap by directly testing winter irrigation effects on fog using a high-resolution NWP framework and satellite/surface observations.

Study period and domain: Three consecutive dense fog events over 21–24 January 2016 across the Indo-Gangetic Plains (IGP), including Delhi, were simulated with a regional Met Office Unified Model (MetUM) configuration named the Delhi Research Unified Model (DRUM). Model configuration: DRUM is a limited-area MetUM using the Regional Atmosphere Tropical configuration version 2 (RA2T), nested within a 17 km global MetUM (GA6.1) analysis providing initial and boundary conditions. The regional horizontal resolution is 1.5 km. Each forecast was initialized at 00Z and integrated for 36 h; three runs were performed starting 00Z on 21, 22, and 23 January 2016 to cover evolving fog events. Irrigation representation: The irrigated area mask for winter was derived by intersecting ESA CCI Land Cover 2016 pixels classified as irrigated cropland (300 m) with grid cells having January irrigation water use >5 mm from the global monthly irrigation water use dataset (0.5°) of Huang et al. (2018). Within this irrigated mask, the initial soil moisture was set to the saturated soil moisture content for sand (38.3 kg m−2), following Fletcher et al., adding approximately 20–25 mm of water (about 25% of the typical January monthly irrigation total). After initialization, soil moisture evolved prognostically with the land-surface scheme. The non-irrigated control used unmodified initial soil moisture from the global analysis. This simple perturbation approach aims to insert a realistic amount of moisture representative of flood irrigation practice during Rabi season. Observational datasets for evaluation: 1) INSAT-3D Fog/Low-Cloud product (MOSDAC/ISRO) at 4 km, 30 min resolution: night-time fog detected via multispectral brightness temperature difference (mid-wave IR and thermal IR), sensitive primarily to optically thick (deep) fog; daytime detection uses visible reflectance plus thermal IR, allowing better sensitivity to thin fog. 2) WiFEX in situ observations at IGI Airport, Delhi (2015–2020), including automatic weather station data and radiation fluxes; used here for visibility, relative humidity (RH), and downwelling longwave radiation (LWD). 3) SYNOP station RH and visibility from six airports: Delhi Safdarjung (DEL), Bareilly (BRL), Dehradun (DHD), Patiala (PTL), Gwalior (GWL), and Hisar (HSR). 4) IMD daily gridded rainfall (0.25°) for climatological context and January trends. 5) Global irrigation water use (Huang et al., 0.5°) for 1972–2010 to analyze trends. Model diagnostics and fog classification: Model outputs of Very Low Cloud (cloud base <111 m) and Low Cloud (111–1949 m) were used to diagnose fog categories: shallow fog (Very Low Cloud fraction >0 alone), deep fog (both Very Low and Low Cloud fractions >0), and low cloud (Low Cloud fraction >0 without Very Low Cloud). Visibility is a derived diagnostic dependent on assumed aerosol size/mass and humidity; known uncertainties exist. Near-surface (5 m) specific humidity and 10 m winds were analyzed to assess moisture advection. Trend analysis: January spatial trends (1972–2010) in irrigation water use and rainfall were computed; significance tested at 95% using Student’s t-test. Maps illustrate strong positive irrigation trends over NW IGP (Punjab, northern Rajasthan, western Haryana) with relatively small rainfall trends in this dry season. Evaluation strategy: Spatial evaluation against INSAT-3D at development (18Z 21 Jan), mature (00Z 22 Jan), and dissipation (06Z 22 Jan) stages; temporal evaluation at WiFEX (IGI Airport) over 22–24 Jan for visibility, RH, and LWD; multi-site RH and visibility comparisons at six SYNOP stations. Moisture and wind fields compared between irrigated and non-irrigated runs to diagnose processes (evaporation, advection, flux changes).

- Observed changes and context: Winter fog days in Delhi increased by 71% over three decades (from 38 in the 1980s to 65 recently). December–January average dew point at Safdarjung rose from 7 °C (1980) to 10.5 °C (2019). Winter irrigation water use in NW IGP increased by ≥10 mm/month/decade since the 1970s, while winter rainfall magnitudes and trends remained small. - Spatial fog simulation: For the 21–22 Jan 2016 event, the irrigated simulation closely reproduced INSAT-3D deep fog coverage and evolution at development (18Z 21 Jan) and mature (00Z 22 Jan) stages, and captured dissipation onset in the SE by 06Z 22 Jan. The non-irrigated run produced less widespread shallow fog and dissipated fog too early. - Humidity and advection: Including irrigation raised near-surface (5 m) specific humidity by typically ~1 g/kg across much of the IGP. Predominantly light (<5 m/s) northwesterly winds advected moisture from irrigated NW areas into downwind regions (e.g., Uttar Pradesh and Delhi), demonstrating a strong non-local effect. - Surface fluxes: Increased latent heat flux over irrigated areas (from added soil moisture) was balanced by decreased sensible heat flux, supporting higher near-surface humidity conducive to fog formation and persistence. - Site-level verification: At IGI Airport (WiFEX), the irrigated runs more accurately reproduced diurnal cycles and magnitudes of visibility, RH, and downwelling longwave radiation (LWD) over 22–24 Jan. Across SYNOP sites, irrigated runs produced higher RH closer to observations, with largest improvements near/within irrigated areas (BRL, HSR, PTL, DEL) and smaller but sometimes notable impacts at downwind sites (e.g., DHD on 22 Jan). These RH improvements translated into better visibility simulations. - Remaining biases: Despite improvements, both configurations failed to capture the rapid post-sunset visibility drop, likely linked to a simple urban land-surface parameterization causing ~1 K nighttime warm bias and limitations of the visibility diagnostic reliant on assumed aerosol properties. - Overall: Incorporating winter irrigation into NWP initial land states effectively corrects prevalent dry biases over the IGP, substantially enhancing the spatial distribution, timing, and optical depth of fog and resultant visibility forecasts.

The findings directly support the hypothesis that increased winter irrigation over the IGP enhances near-surface humidity and promotes dense fog formation, explaining part of the observed upward trend in fog frequency. By adding realistic amounts of soil moisture to irrigated cropland areas, the NWP model’s near-surface dry bias is reduced, leading to better agreement with satellite-observed deep fog extent and surface observations of RH, visibility, and LWD. Moisture advection from irrigated NW regions induces non-local increases in humidity and fog over urban areas like Delhi, underscoring the regional-scale influence of irrigation on fog. Although the case study spans three days, the fog events were representative in density, duration, and spatial extent, and the model dry bias is typical for wintertime over the IGP, suggesting broader applicability of the conclusions. Nonetheless, accurate fog forecasting still depends on a suite of processes (urban land-surface heterogeneity, aerosol emissions/chemistry, cloud–aerosol microphysics), and further improvements are needed to fully resolve remaining biases and timing errors. The demonstrated benefits of including irrigation indicate a path toward substantial operational forecast skill gains for winter fog in northern India.

This work provides the first direct modeling evidence that wintertime irrigation in the IGP plays a decisive role in widespread dense fog formation and persistence, and that explicitly representing irrigation in NWP initial land states markedly improves fog forecasts. The study links increased winter irrigation—with associated soil moisture and latent heat flux changes—to higher near-surface humidity and non-local advection effects that aid fog development over Delhi and surrounding regions. These results offer a plausible explanation for the observed increase in fog frequency in recent decades. For operational forecasting, a step-change improvement is likely achievable by incorporating irrigation effects via enhanced soil moisture analysis, smart irrigation parameterizations with realistic spatial/temporal variability, or hybrid approaches. Future work should refine urban land-surface representations, aerosol sources and chemistry, and aerosol–fog microphysics, and develop robust, operationally feasible irrigation schemes and data assimilation strategies.

- Case study scope: Main results are based on three consecutive fog events (21–24 January 2016); generalization is supported by representativeness but not exhaustively tested across seasons/years. - Simplified irrigation representation: Irrigation was represented by saturating initial soil moisture in irrigated cropland areas (uniform addition of ~20–25 mm), without explicit temporal application, irrigation method variability, or farmer practice dynamics. - Urban and aerosol processes: Simplified urban land-surface parameterization led to a ~1 K nighttime warm bias and contributed to failure to simulate rapid post-sunset visibility drops; aerosol sources, chemistry, and microphysics were not fully addressed, affecting fog–visibility diagnostics. - Visibility diagnostic: Visibility is a derived variable depending on assumed aerosol properties and humidity; diagnostic limitations may contribute to residual discrepancies. - Satellite product sensitivity: Night-time INSAT-3D fog product is less sensitive to optically thin fog, constraining evaluation primarily to deep fog and potentially underestimating thin fog occurrence.