Thresholds of fire response to moisture and fuel load differ between tropical savannas 1 and grasslands across continents 2 3

8 Aim: An emerging framework for tropical ecosystems states that fire activity is either ‘fuel 9 build-up limited’ or ‘fuel moisture limited’ i.e. as you move up along rainfall gradients, the 10 major control on fire occurrence switches from being the amount of fuel, to the moisture 11 content of the fuel. Here we used remotely sensed datasets to assess whether interannual 12 variability of burned area is better explained by annual rainfall totals driving fuel build-up, or 13 by dry season rainfall driving fuel moisture. 14 Location: Pantropical savannas and grasslands 15 Time period: 2002-2016 16 Methods: We explored the response of annual burned area to interannual variability in 17 rainfall. We compared several linear models to understand how fuel moisture and fuel build18 up effect (accumulated rainfall during 6 and 24 months prior to the end of the burning season 19 respectively) determine the interannual variability of burned area and explore if tree cover, 20 dry season duration and human activity modified these relationships. 21 Results: Fuel and moisture controls on fire occurrence in tropical savannas varied across 22 continents. Only 24% of South American savannas were fuel build-up limited against 61% of 23 Australian savannas and 47% of African savannas. On average, South America switched from 24 fuel limited to moisture limited at 500 mm yr, Africa at 800 mm yr and Australia at 1000 25 mm yr of mean annual rainfall. 26 Main conclusions: In 42% of tropical savannas (accounting for 41% of current area burned) 27 increased drought and higher temperatures will not increase fire, but there are savannas, 28 particularly in South America, that are likely to become more flammable with increasing 29 temperatures. These findings highlight that we cannot transfer knowledge of fire responses to 30 global change across ecosystems/regions – local solutions to local fire management issues are 31 required, and different tropical savanna regions may show contrasting responses to the same 32 drivers of global change. 33 This is the peer reviewed version of the following article: Alvarado, ST, Andela, N, Silva, TSF, Archibald, S. Thresholds of fire response to moisture and fuel load differ between tropical savannas and grasslands across continents. Global Ecology and Biogeography 2020; 29: 331– 344, which has been published in final form at https://doi.org/10.1111/geb.13034. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for self-archiving.


36
Understanding global controls on fire activity has become increasingly important in the 37 context of ecosystem drying and climatic change (Jolly et al., 2015). In some ecosystems 38 drought events and rising temperatures may exacerbate fire risk (Bowman et al., 2011;Price 39 et al., 2015), and increase the incidence of large wildfires and fire-associated CO2 emissions 40 (Voulgarakis & Field, 2015;Hantson et al., 2017). However, not all ecosystems burn more 41 when exposed to drought and high temperatures. Pausas and Ribeiro (2013) showed that fire 42 in lower-productivity systems was unresponsive to temperature, and paleo-records highlight  build-up effects on interannual variability in burned area. We defined the fuel moisture effect 137 as the accumulated rainfall during the six months prior to the end of the burning season. We 138 assumed that rainfall occurring during, or just before the burning season determines the 139 probability of ignition and fire spread. The fuel build-up effect was defined as the accumulated 140 rainfall during 24 months prior to the end of the burning season, as previous rainfall is an 141 important control on the amount of biomass produced. We selected the 6-and 24-months cut-142 off as, on average, the strongest negative response in fuel moisture limited landscapes was 143 found around 6-7 months of antecedent rainfall (Fig. A1), while across fuel build-up limited 144 landscapes accumulated rainfall over two wet seasons (24-months) had a slightly higher 145 explanatory power than over a single wet season (12-months) (Fig. A1). 146 We considered three explanatory variables for our initial analysis of the drivers of with the long lead times, but in these cases, effects were generally not significant at the same 177 time (p<0.05 in 5% of total grid cells). For simplicity, we therefore selected the strongest 178 absolute correlation for each grid cell.

179
Based on the biome wide characterization of burned area response to antecedent rainfall,  Drivers of burned area response. We used two different approaches to explore the drivers 188 of spatial differences in the relationship between annual burned area and antecedent rainfall.     Fig 3a). Interestingly, when dry season length exceeded 9 months, annual burned area 259 typically declined again, likely because very short growing seasons may limit ecosystem 260 productivity and thus fuel availability. In addition to burned area, we also analyzed its response in savannas with MAR below 900 mm yr -1 , independently of the HII (Fig. 4c).

287
Vegetation structure also influenced biome wide patterns of burned area and the strength 288 and sign of correlation coefficients between antecedent rainfall and burned area (Fig. 5). As 289 expected, we observed that higher tree cover was often associated with reduced burned area, 290 particularly in the humid tropics (Fig. 5a). In productive savannas (MAR ranging from 900 to 291 2000 mm yr-1), the fuel moisture effect tended to strengthen with increasing tree cover, 292 although relationships were often weak (Fig. 5c). In fuel limited ecosystems of Australia,   (Table A1). Surprisingly, BA did not vary as  (Table 1 and A1, Fig. 2).

319
The models with the highest explanatory power (lowest AIC) explained 0.4% of the 320 variance of the interannual variability of BA (Table 1) and 29% of spatial occurrence ( we included all three variables in the model, we detected a slight decrease in the slope of fuel 331 build-up effect for African (from 0.00041 to 0.00035% mm -1 ) and South America savannas 332 (from 0.00058 to 0.00046% mm -1 ) and a larger increase for Australian savannas (from 333 0.00088 to 0.0019% mm -1 , Table 1). The inclusion of these three factors also modified the increased from 0.00058 to 0.010% mm -1. When we analyzed the absolute BA (Table A1) Table A1). in Australia (Fig. 2). Together, these two factors resulted in continental scale differences in  frequent fires to occur in productive and humid savannas, but we only detected a weak relationship between annual burned area and increasing dry season lengths longer than six 396 months. A possible explanation for this weak relationship could be that dry season duration 397 longer than six months may limit herbaceous productivity by shortening the growing season 398 in spite of MAR. In addition, our results suggest that observed differences in rainfall 399 seasonality may also modify the response of burned area to antecedent rainfall across different 400 regions ( Fig. 3b and c). Although the relatively long and pronounced African dry season is To confirm these findings, we used a range of multiple linear regression models to explore 432 if the continental scale differences could be explained by differences in DS, HII, and TC.

433
Allowing the burned area to respond differently to antecedent rainfall across continents 434 caused a considerable model improvement both when modeling absolute burned area (Table   435 A1) or it's variability (Table 1). While the introduction of DS, HII and TC as additional