Desert locust breeding habitat risk mapping
Desert locust outbreaks begin in remote semi-arid zones where moist soils and green flush create brief breeding windows. Satellite-derived soil moisture, vegetation anomalies, and rainfall estimates can flag those windows days to weeks before ground teams can reach them.
Sensors
- SMAP L-band radiometer (NASA): Measures surface soil moisture at 36 km passive radiometer resolution (9 km enhanced product with radar-radiometer fusion). Two-to-three day global revisit. L-band (1.4 GHz) penetrates light vegetation canopy and dry surface crust to sense the top 5 cm of soil, the layer critical for locust oviposition. Latency roughly 12–24 hours after overpass.
- Sentinel-2 MSI (ESA/Copernicus): 10 m resolution in visible and near-infrared bands; 20 m in red-edge and SWIR. Five-day revisit at the equator with both satellites. Provides NDVI, EVI, and moisture-sensitive SWIR indices to detect green vegetation flush in recession zones. Cloud cover is a real constraint in the Inter-Tropical Convergence Zone during active monsoon periods.
- MODIS Terra and Aqua (NASA): 250 m NDVI at daily revisit, with 16-day composites that suppress cloud contamination. Lower spatial detail than Sentinel-2 but far better temporal density, which matters when a vegetation flush can emerge and senesce within two to three weeks. MODIS NDVI anomaly products relative to multi-year baselines are operationally used by FAO's RAMSES system.
- CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data): Quasi-global rainfall estimates at 0.05-degree (~5 km) resolution, blending TRMM/GPM infrared estimates with weather-station records. Five-day and monthly products available with roughly two-day latency. Useful for identifying rainfall events in data-sparse regions of the Sahel, Horn of Africa, and Arabian Peninsula where gauge networks are thin.
- Meteosat SEVIRI (EUMETSAT): Geostationary sensor covering Africa, the Middle East, and parts of South Asia at 15-minute repeat. At 3 km visible and 1–3 km infrared resolution it is too coarse for habitat mapping but invaluable for tracking convective rainfall cells that trigger soil moisture pulses, providing near-real-time context that polar-orbiting sensors cannot match.
Why a locust swarm is already late news
By the time a swarm is visible from the ground, the breeding that produced it happened weeks earlier, often in terrain that field teams cannot reach quickly. Schistocerca gregaria requires moist sandy soils for oviposition and green vegetation for the fledgling hoppers to feed on. Both conditions are transient: a rainfall event in a semi-arid recession zone can create a viable breeding window that opens and closes within three to four weeks.
Satellite data cannot show you a locust. What it can show you is the environmental signature that makes breeding probable, and it can do so within days of a triggering rainfall event, across the full extent of the recession zone spanning the Sahel, Horn of Africa, Red Sea coast, and Arabian Peninsula. That is the practical value: not confirmation, but early warning of where to look.
The ecological logic that makes remote sensing useful
FAO's locust-ecology documentation establishes three necessary conditions for successful breeding: soil moisture sufficient for egg survival (roughly 10–15 percent volumetric water content in the top 5 cm), green vegetation providing food for hatchlings, and loose sandy or sandy-loam substrate that females can probe with their ovipositor. The first two are directly observable from orbit. The third, soil texture, is largely static and can be pre-mapped from existing soil surveys and Sentinel-1 backscatter, so it functions as a fixed mask rather than a dynamic variable.
SMAP's L-band radiometer is physically well-suited to the first condition. L-band microwave emission from soil is sensitive to dielectric constant, which changes sharply with liquid water content, and the wavelength is long enough to partially see through sparse vegetation canopy and dry surface crusts. The 36 km native resolution is coarse, but the enhanced 9 km product is adequate to identify which sub-catchments within a recession zone have received meaningful moisture. NDVI anomalies from MODIS or Sentinel-2, expressed as departures from the multi-year mean for the same calendar period, then identify where green flush has followed that moisture. The intersection of both anomalies, constrained to sandy-soil areas, defines the habitat-risk footprint.
What the data cannot do, and why that matters operationally
Habitat suitability is not locust presence. A risk map showing favourable breeding conditions in southern Mauritania tells a national plant-protection directorate where to prioritise ground surveys, not that locusts are definitely there. This distinction is not a technicality; acting on suitability alone without ground confirmation wastes scarce pesticide and aircraft resources and erodes trust in the early-warning system.
Cloud cover is a genuine constraint on the optical layers. During active monsoon periods across the Sahel, Sentinel-2 and MODIS can be cloud-contaminated for days at a time, precisely when rainfall-triggered vegetation flush is occurring. Microwave-based SMAP data is unaffected by cloud, so a practical workflow uses SMAP moisture as the primary trigger layer and fills the optical gap with CHIRPS rainfall accumulation as a proxy for likely greenness, accepting that the vegetation confirmation will arrive with a lag once skies clear.
Revisit also matters. A 16-day Sentinel-2 composite may miss a flush that peaks and browns off in ten days. The operational answer is to use MODIS daily data at lower resolution as the primary change-detection layer and reserve Sentinel-2 for targeted high-resolution characterisation once a candidate area is flagged.
Wind trajectories close the gap between breeding and arrival
Adult locusts are strong fliers but largely travel with the wind. Integrating ERA5 reanalysis wind fields or NOAA HYSPLIT trajectory outputs with the breeding-habitat layer allows analysts to estimate where a cohort of adults, if present in a flagged breeding zone, would likely travel over the following three to seven days. This is how FAO's Desert Locust Information Service constructs its watch bulletins: known or suspected breeding areas combined with prevailing wind patterns to project invasion risk into agricultural zones.
The satellite data contribution to this step is primarily positional: defining where the source population plausibly exists. Wind modelling is then a meteorological computation, not a remote-sensing one. The two layers are complementary rather than competing.
Building a working risk layer: the data stack in practice
A practical habitat-risk product combines four inputs. First, a static soil-texture mask derived from FAO's Harmonized World Soil Database or equivalent, identifying sandy and sandy-loam areas within the known recession zone. Second, CHIRPS five-day rainfall accumulation to flag grid cells that have received above-threshold precipitation (the FAO guidance suggests roughly 25 mm over ten days as a useful trigger, though this varies by region and season). Third, SMAP surface moisture to confirm that rainfall has actually infiltrated rather than run off. Fourth, MODIS NDVI anomaly to detect green flush.
Each layer is individually imperfect. Together, requiring two or more to coincide with the soil mask, they produce a risk footprint with meaningfully fewer false positives than any single indicator. The output is a probability-of-suitability surface, not a binary map, and it should be delivered with a clear statement of which inputs were available and which were gap-filled. Satellize applies this kind of multi-layer fusion across open constellations for clients in food-security contexts, including its crop-estimation work with the Kingdom of Tonga, and can configure the same pipeline for locust-monitoring mandates.
Latency from satellite overpass to delivered risk layer is typically 24–48 hours for the microwave and rainfall inputs. Optical confirmation may lag by several days depending on cloud conditions. That timeline is still far faster than the weeks required to organise and complete ground surveys across the Sahel or Arabian Peninsula.
What a good risk product looks like on delivery
A habitat-risk layer is most useful when it arrives as a GIS-ready raster with an accompanying confidence flag, not as a static PDF map. National plant-protection directorates and FAO field offices need to overlay the layer on their own survey-team locations and road-network data to plan the most efficient ground-verification routes. A five-class suitability surface (no risk through very high risk), updated on a five-day cycle aligned to CHIRPS release, with SMAP and MODIS inputs refreshed daily, gives operational planners something they can actually act on.
Archive depth matters for calibration. MODIS provides a continuous record from 2000, SMAP from 2015. Comparing current anomaly signals against historical outbreak years documented in FAO records allows analysts to assess whether the current signature resembles conditions that preceded past upsurges. That comparison is not prediction; it is context, and context is what separates a useful early-warning product from an automated alert that nobody trusts.
Typical figures
| Soil moisture spatial resolution | 36 km (SMAP native radiometer); 9 km enhanced product |
| Vegetation index resolution | 250 m (MODIS NDVI); 10–20 m (Sentinel-2 NDVI/EVI) |
| Rainfall estimate resolution | ~5 km / 0.05 degree (CHIRPS five-day product) |
| Revisit cadence | SMAP: 2–3 days global; MODIS: daily; Sentinel-2: 5 days at equator; CHIRPS: 5-day accumulation |
| Soil moisture sensing depth | Top 5 cm (SMAP L-band, 1.4 GHz) |
| Typical product latency | 24–48 hours for SMAP and CHIRPS layers; optical confirmation may lag 2–7 days in cloudy periods |
| Geographic coverage | Quasi-global; recession zone focus: Sahel, Horn of Africa, Red Sea littoral, Arabian Peninsula, northwest India/Pakistan |
| Archive depth | MODIS from 2000; SMAP from April 2015; CHIRPS from 1981 |
| Minimum detectable moisture anomaly | SMAP volumetric water content uncertainty ~0.04 m³/m³ under vegetation-free conditions |
| Delivery format | GeoTIFF suitability raster, GeoJSON alert polygons, five-class confidence flag layer |
Analytics Satellize can run
| Five-day habitat suitability surface | Multi-criteria index combining SMAP moisture, CHIRPS rainfall accumulation, MODIS NDVI anomaly, and static soil-texture mask; threshold-based fusion following FAO ecological criteria | GeoTIFF raster, five suitability classes, updated on CHIRPS release cycle |
| Rainfall-triggered alert polygons | CHIRPS threshold exceedance (configurable, default ~25 mm per 10 days) within sandy-soil mask; confirmed by SMAP moisture signal where available | GeoJSON polygon feed with timestamp, rainfall total, and SMAP confirmation flag |
| NDVI anomaly map for recession zones | MODIS MOD13 16-day NDVI versus multi-year mean for same calendar period; z-score departure surface clipped to recession-zone boundary | GeoTIFF anomaly layer with historical percentile ranking per pixel |
| High-resolution vegetation confirmation | Sentinel-2 MSI NDVI and SWIR moisture index for candidate areas flagged by MODIS; cloud-gap filling via temporal interpolation where consecutive clear scenes exist | 10 m GeoTIFF for priority sub-regions, delivered within 48 hours of clear overpass |
| Wind-trajectory risk overlay | ERA5 reanalysis wind fields or NOAA HYSPLIT forward trajectories originating from high-suitability breeding polygons; 3- and 7-day projection | Trajectory polygon layer showing potential adult dispersal corridors, overlaid on national crop-area boundaries |
| Historical analogue comparison report | Current anomaly signature compared against MODIS and CHIRPS archive for years with documented upsurges in FAO records; pattern-similarity scoring | PDF briefing note with ranked historical analogues and confidence statement, issued when suitability index exceeds defined threshold |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.