Two Kabylie fires, two years apart, sharing only 334 hectares

The published record for this terrain is entirely about extent and severity in the weeks after the flames. Three studies exist for 2021: a severity map of Tizi Ouzou, a Béjaïa severity preprint, and a two wilaya extent assessment. For 2023 there is not one. For either year there is no published recovery time series, no reburn analysis, and no comparison between the two fires at all.

International mapping was real, and it was inverted between the two events. August 2021 drew an activation from the European emergency mapping service, EMSR533, opened on 11 August and the only Algerian activation in the history of that service, but no International Charter call. July 2023 drew the opposite: Charter activation 831, requested by the Algerian civil protection with the space agency as project manager, and no European activation. The UN satellite centre worked both, but in 2021 published hotspot density rather than burned area, and only in 2023 delineated what had actually burned.

Almost none of what burned was insured, and the industry's own accounts show it. Algeria's agricultural risks branch wrote 2,157 million dinars of premium in 2021, 1.6 per cent of all damage insurance, and paid 923 million in claims. That is 16.1 per cent less than the 1,100 million it paid in 2020. In the worst fire year in over a decade, the national agricultural insurer's claims went down.

The gap is take-up rather than product. The mutual that holds 65.1 per cent of the branch already sells fire cover for orchards, fruit trees and beehives, but the director of its Tizi Ouzou arm said in 2017 that orchard cover there does not exceed 3 per cent, and orchards are precisely what burned. Algerian law does not treat wildfire as a natural catastrophe, so the mechanism built for those does not respond to it.

The state paid instead, largely in kind: a million dinars to the family of each victim, 700,000 for a destroyed home, and for farmers no cash indemnity at all. Compensation came as livestock and saplings. A year on, 18,256 farmers had been compensated across the two worst hit wilayas, 4,160 of them in Béjaïa, for about 2.6 billion dinars, and Béjaïa had been planted with 146,934 fruit trees of which 113,291 were olives.

Loss assessment here was already a remote sensing problem and Algeria already knew it. The wilaya commissions that valued the damage used satellite imagery alongside farmers' declarations, working with the national space agency. What none of that answered is the question that follows a payout: how long the ground takes to come back, and whether the next fire will burn the same place.

Both fires exceed their control by more than an order of magnitude. Two agencies mapping the 2023 fire independently land beside these figures rather than away from them: the UN satellite centre found about 18,000 hectares of burned vegetation over a 540 square kilometre window on 25 July, and the European fire information system recorded the biggest fire of its year at over 14,000 hectares in the Béjaïa region against our largest single connected scar of 15,603. The windows and boundaries differ, so neither is a like-for-like score, and neither was consulted before the run.

Two fires of almost identical size, two years apart, in the same 1,244 square kilometres, and they scarcely touched. Whatever drove the 2023 fire, it was not fuel left standing by the 2021 one.

Drawn together, the two burns abut along a north to south line, which reads immediately as a two year old scar acting as a fuel break. Measured against the ground the 2023 fire could have burned instead, the 75,151 hectares of vegetated, twice unburned land in the same box, that reading does not hold. The 2023 scar sits at a median 2,478 metres from the 2021 scar against the pool's 2,192, and it is less likely than chance to fall within 100 metres of it, 2.2 per cent against 4.4. At the scale this box resolves, the old scar neither drew the new fire nor blocked it. The appearance is two large patches in a small area.

The 2021 scar closed its gap to unburned forest in four summers. At four summers the 95 per cent interval on that gap runs from −0.003 to +0.009 and contains zero: the canopy signal over that ground is no longer separable from forest that never burned. Published Mediterranean recovery times run from about two years in northwest Spain to roughly four in Portuguese seeder dominated shrubland and nearly six in Catalonia, and the closest Algerian analogue, the Dahra forest in the northwest, recovered in two years after its 2012 fire and nine after its 2000 one.

The 2023 fire burned deeper but recovers at the same rate. Its opening gap is 0.723 against 0.524, 38 per cent more severe on identical settings, and measured 30 days after ignition against the 2021 fire's 13, so the difference is if anything understated. After one summer it had closed 59 per cent of that gap against the older fire's 57, and after two summers 79 against 82. The 2023 scar is not recovering slowly. It started further down.

Tracking the identical scars on the vegetation index instead of the burn ratio gives 47, 79, 91 and 98 per cent for 2021 and 55 and 77 for 2023, with the 2021 interval containing zero there too. The published comparison of these two index families finds burn ratio measures report systematically slower recovery than vegetation index ones over the same pixels. Ours agree to within two points, in the direction that literature predicts.

One more thing the series shows. In August 2021 the fire is a single connected patch of 10,009 hectares. One year later the largest connected patch inside the same footprint is 462 hectares, across 719 fragments. The window in which a fire of this kind reads as one coherent object in imagery is about one growing season wide.