Aquatic ecosystem responses to fire and flood size in the Okavango Delta: observations from the seasonal floodplains
The frequency of fires in the Okavango Delta seasonal floodplains peaked at an intermediate frequency of flooding. Floodplains are commonly burnt every 3–5 years. This study showed fundamental changes in ecosystem properties due to burning. A burnt seasonal floodplain in the aquatic phase had oxygen levels well above saturation, 100–200%, while the levels in the un-burnt control site were below saturation and, at night, could decline to 10–40% saturation. The total phosphorous and total nitrogen concentrations were similar on both floodplains but considerably enriched relative to inflowing water, due to nutrient release from the flooded soil-sediment and animal droppings. Zooplankton biomass was very high in both systems although the abundance of fish fry was ten times higher on the un-burnt floodplain. In a low flood year the un-burnt floodplain water had high nutrient levels, primary production, methane emission, and subsequent uptake of methane in biota, as well as a high zooplankton biomass. The very high flood the following year showed the opposite with much lower production at all levels owing primarily to greater dilution of nutrients. The abundance of fish, however, was much higher during the high flood year. Macrophytes and litter provide direct shelter for fish fry but also promote low oxygen levels when decaying. Large flooded areas result in high fish production by removing obstacles related to congestion. This interplay between hydroperiod and fire may be crucial for the maintenance of high biological productivity both in the aquatic and terrestrial phases in a very nutrient poor wetland landscape. Understanding these interactions is crucial for optimal management.
KeywordsSeasonal floodplain Fire Flood pulse Macrophyte Okavango Delta Nutrients Zooplankton biomass Fish productivity
The consequences for aquatic biota of these flooding-fire sequences have previously not been analyzed in the Okavango Delta. Here, we consider the major effects on biota by comparing the effects of two major “natural experiments” on two seasonal floodplains. The first assessment involves the removal of macrophytes by fire in which we compare basic ecosystem properties in burned and un-burned habitats, while the second assessment is a flooding experiment where conditions during extremely high and very low flooding seasons are compared.
These kinds of natural experiments are of course not comparable with controlled experiments. Although only two changes are made—(i) removal of macrophytes and detritus, and (ii) a switch from very low water levels to very high water levels (with consequent higher duration and greater extent of flooding), these drivers can lead to a number of ecosystem responses that may have cascading effects. Nevertheless, the nature of observed responses can provide some valuable insights to the basic ecological functions of the flooding-fire coupling and the importance of variability in the magnitude of flooding. In particular, large tropical and sub-tropical wetland systems often have considerable socio-economic importance. A better understanding of the interplay between the two major drivers of these ecosystem processes, namely flooding and fire, is a prerequisite for improved management of these important systems.
The study area
The study area along the Boro River (Fig. 1) has typical vegetation gradients caused by small differences in elevation which determine the duration of flooding (Bonyongo 1999). The permanent swamps along the river are dominated by Phragmites australis (Cav.) Steud. When the flood arrives in this area (usually in the period mid-May to mid-June), water extends out over the river banks and moves through a system of channels and small pools onto the floodplains. The deepest parts of the floodplains are characterized by floating leafy vegetation such as Ludwigia stolonifera (Guill & Perr.) P.H. Raven and Nymphaea nouchali Burm, and are kept open by habitual movements of hippopotamus. In slightly less deep water, almost mono-specific dense 1–1.5 m high stands of the sedge Cyperus articulatus L. exist. These two deepest zones are often flooded for 4–8 months. In the shallower areas, perennial grasslands with typical species like Panicum repens L. and Paspalidium obtusifolium (Delile) N. D. Simpson occur. These slightly elevated zones are usually flooded for 1–3 months and are often heavily grazed by large mammals. Between this raised area and the riparian woodland are stands of tall perennial grasses such as Imperata cylindrica (L.) Rueusch, which are inundated only during years with exceptionally high flooding. This typical zonation can be several hundred meters wide, yet occupy a total vertical depth gradient of only about one meter (Bonyongo 1999).
Since 1996, a number of studies within the study area have been undertaken, particularly on the two seasonal floodplains, Lechwe Floodplain and Phelo Floodplain. Both floodplains receive their water from the Boro River and their water chemistry is therefore similar. They have the same vegetation zonation and are on a similar substrate, Kalahari sand, which comprises the whole Delta, as well as the greater river basin, and is very homogenous. It is also a very nutrient poor substrate and the biota in rivers and the lagoons of the Delta can be described as oligotrophic.
In April 1998 a large fire swept through the southern part of the study area including the Lechwe, but not the Phelo, floodplain (Fig. 1) (Meyer 1999). Due to several preceding dry years, the fire was hot and consumed not only all the dry debris but also the green above-ground stands of Cyperus articulatus. As a result the whole floodplain was covered with a 1–4 cm ash layer and very few emergent macrophytes were present in the Lechwe Floodplain at the time of its next flooding. This next flood was relatively small (similar to the one experienced in 2003, ~0.1 km2) and was confined to the deeper Ludwigia and Cyperus zones. The following year the typical vegetation zonation described previously was re-established in the Lechwe Floodplain, making it structurally similar to the un-burnt Phelo Floodplain and providing an opportunity to study the effect of fire on aquatic biota. It must be acknowledged, however, that differences in nutrient loading between the two floodplains exist, owing to differences in the input of fecal matter from grazing mammals that had favoured the Phelo Floodplain in previous years, as well as differences in dilution.
In 2003 a small flood event in the Phelo Floodplain extended only into the Cyperus zone and lasted for about six months. In this situation, all aquatic biota were confined to the small area of open water and the Cyperus habitat. The flooding the following year was much greater; it extended into the rarely flooded Imperata zone and lasted for ten months. A large, predominantly terrestrial area was flooded and became accessible to aquatic biota. These two years thus offered the opportunity to assess for effects of flooding size. A series of limnological studies were conducted on the Phelo Floodplain in 2003 and 2004, including studies on nutrient dynamics, primary and secondary production, food web interactions, and methane dynamics (Lindholm 2006).
Sampling and analytical methods
Sampling was undertaken during three periods, (1) during rising water level at the end of May–June; (2) at peak water levels in July–August, and (3) during the receding flood in September–October. Water samples for analysis of total nitrogen, total phosphorus, chlorophyll a and zooplankton biomass were taken at five locations in the deepest open water pool-channel system from the inlet to the innermost part of the floodplains. Temperature and oxygen were measured in situ at these same locations to assess diurnal variations. Pelagic primary production was measured during 2003–2004 by the standard 14C method. During the1998 field work, fish fry were sampled at ten randomly selected sites in the Panicum zone, using a 1.33-m-diameter ring-net, with a floating top ring and a heavy metal ring at the bottom that was thrown out and emptied by a hand net. Fishing during the 2003–2004 floods was done with a standardized set of gillnets and the fish abundance calculated as catch per unit effort (CPUE). Standard analytical procedures were used throughout and are detailed in Lindholm (2006) and Lindholm et al. (2007). Macrophyte data on species composition and biomass were taken from Bonyongo (1999) and from Murray-Hudson (unpublished), while water chemistry data for the Boro River, usually sampled fortnightly, were taken from environmental monitoring databases held within the Harry Oppenheimer Okavango Research Centre (HOORC) at the University of Botswana.
Macrophyte and litter biomass
Mean elevation above bottom of Boro River (= 0 m), total phosphorous in flooded sediment and biomass (gram DW m−2) for vegetation zones on Phelo Floodplain 1997 with the data from Bonyongo (1999), Bonyongo et al. (2000) and for Cyperus biomass and litter from various sources as indicated below. Biomass data from March represents likely maximum values at the end of the rainy-growth season, for june just before flooding and for November just after flooding
Total P (mg kg−1)
March gram DW m−2
June gram DW m−2
November gram DW m−2
Effects of fire
Fires on floodplains in the Okavango Delta remove around 78% of the litter (Rutz, 2004) with a large variation dependent primarily on wetness and fuel load. The fire on Lechwe Floodplain occurred after a long drought of several years and was of high severity (DeBano et al. 1998). It burnt not only the litter but also the green stands of Cyperus articulatus and Schoenoplectus corymbosus, leaving only a 1–4 cm thick ash layer.
Total-N and Total-P (mg l−1) and P:N quotients by weight on Lechwe burnt and Phelo un-burnt floodplain, and high and low flood, as mean values for the entire flooding season
Phelo low flood
Phelo high flood
Zooplankton biomass and fish fry abundance on burnt Lechwe- and un-burnt Phelo floodplains as mean values during flooding year 1998
Lechwe burnt (mg DW l−1)
Phelo un-burnt (mg DW l−1)
Zooplankton biomass (high water)
Zooplankton biomass (receding water)
Fish fry abundance (high water)
Effects of flooding and flood size
When dry litter is flooded it decays very fast. On Phelo’s Floodplain, when the 2001 flood arrived on 17th May, there was a mean DW of 1025 gm−2 of litter that, after 3.5 months, had been reduced by 80% to 205 gm−2 (Table 1). Through microbial processes it was converted to dissolved organic carbon (DOC) of which 51% was further oxidized to CO2, 26% infiltrated to groundwater, and 22% was lost through outflow which (unusually) occurred during this high flood year (Mladenov et al. 2007). In spite of these processes the concentration of DOC remained fairly similar during the flooding season with 10–15 mg C l−1. A contributing reason for this, however, was that 65% of the DOC was introduced with incoming flood water from upstream sources.
The total-N concentration in Phelo Floodplain during the low flood year of 2003 and the high flood year of 2004 was similar (1.3 mg l−1 and 1.0 mg l−1, respectively) for the whole season and appeared to not be significantly higher than the source Boro River water (see Table 2). The total-P concentrations on the other hand seemed very different with 0.4 mg l−1 in the first year, 7× higher than the total P concentration in the Boro River. However, during the high flood year, the total-P concentration was as low as 0.009 mg l−1 on the floodplain and was consistently below 0.01 mg l−1 in the Boro River. This has a pronounced effect on the P:N quotients that were 1:3.2 and 1:110, respectively, for the two years.
The fire on Lechwe Floodplain in 1998 seemed not to have had any effect on the concentrations of total-N and total-P in floodplain waters. The nitrogen in the litter and within Cyperus stands probably volatilized at the high temperatures that prevailed as indicated by the formation of a thick ash layer (DeBano et al. 1998). This did not result in lower nitrogen levels in the water and supported the idea that nitrogen fixation alone is usually sufficient to keep the P:N ratio around the optimal of 1:7.
The removal of macrophytes and litter by fire caused a significant increase in oxygen levels (Fig. 4). The litter accumulation, by lowering oxygen levels, may have a number of effects on water quality by affecting both the solubility of phosphorus and the de-nitrification process. For higher trophic levels, however, the elevated oxygen levels might be a disadvantage. For example, some species of fish fry avoid predation by tolerating low oxygen levels (Kolding 1993). The saturated oxygen concentrations might have made the floodplain more accessible for ambush predators allowing them to remain during the night and feed effectively the whole day. The lack of protective macrophyte refugia on the burnt floodplain may make the fish fry even more exposed as discussed below.
All water supplied to the floodplains was provided by the Boro River which typically has a low concentration of nutrients of around 1.0 mg l−1 for total-N and 0.07–0.01 mg l−1 for total-P (Table 2). These data were similar to those from Cronberg et al. (1996) for the years 1991–1992 for the Boro river, with a mean concentration of the inflow being 0.36 mg total-N l−1 and 0.023 mg total-P l−1 and the outflow being 1.12 mg total-N l−1 and 0.055 mg total-P l−1. In 1999, when the HOORC started analyzing total-P in the Boro River, the mean value was 0.039 mg l−1 and was thus very similar to the earlier values. The total-N concentrations for 2003 and 2004 were close to those in the Boro River. On the other hand, the total-P values, with exception of Phelo Floodplain in 2004, were 20–50 times higher than those in the river. As the turnover time for water on Phelo Floodplain was only 5–6 days and about 80–90% of the inflow infiltrated to the groundwater (Ramberg et al. 2006), the high values cannot be caused by evaporative concentration alone. There must be other processes on the floodplains that cause an enrichment of phosphorous, as commonly seen in many wetlands (Mitsch and Gosselink 2007). This may be caused by dissolution from the wetted sandy soils. Noe and Hupp (2007) found large releases of N and P on flooding in a short-hydroperiod floodplain. If this is the case, then these soils must have a considerable storage capacity, as suggested by Mubyana et al. (2003) who showed that dry-land top-soils in the Delta had available P concentrations around 0.1%, while floodplains had considerably higher concentrations of 0.6–1.0%. After the flood in March, these decreased to about 0.1% (op. cit.). It may be calculated that this loss from the sediment, if all dissolved at once, could cause an increase of total-P in the water up to 1.8 g l−1 and could thus readily explain the observed initial high total-P values in floodplain waters. The release of phosphorous when the water arrives might be triggered by development of anoxic conditions in accordance with the Mortimer process (Pearsall and Mortimer 1939). This could also explain the low total-P concentrations on Phelo Floodplain (around 0.01 mg l−1) during the whole high flood year of 2004. During that time, the high water levels and extensive open-water areas might have caused better mixing of the water column and prevented anoxia from developing. In addition, the low concentrations might simply be the result of both dilution by the Boro River, which had consistently low total-P levels (below 0.01 mg l−1), and the fact that soils on higher ground have low concentrations of available phosphorous (Mubyana et al. 2003) which, when flooded, contribute little to the concentration in water.
The P:N ratio was usually above the optimal 1:7 (Table 2) indicating that in general, the floodplains were predominantly P limited with the exception of Phelo Floodplain during the low flood year of 2003, which seemed to have been nitrogen limited for the whole season. This was in contrast to Ogden et al. (2007), who found that N was limiting in a large semi-arid floodplain in Australia. In their situation, however, the catchment geology contained rocks with high to moderate concentrations of P. The siliceous Kalahari sands, which comprise the catchment and Delta of the Okavango, are very nutrient poor (Mubyana et al. 2003). An unknown factor here is the extent of denitrification. Anoxic conditions were likely, as there was a considerable production of methane within the decaying litter during 2003 but not 2004 (Lindholm et al. 2007). It is, therefore, likely that denitrification was a significant process, as it occurs at a considerably higher redox potential than methanogenesis (e.g., Wetzel 2001).
The importance of macrophytes for protection against predators, as previously alluded to, is also illustrated in an assessment of flood sizes. During the high flood year, the pelagic primary production was much reduced to only 42% of that during the low flood year, while zooplankton production was down to only 1%. Conversely, the fish abundance during the low flood year was only 35% of that during the high flood year, although zooplankton are the most important nutrient source for fish fry on the floodplain (Lindholm 2006). The high flood is likely to have caused a dilution effect on prey. During both years, emergent macrophytes occurred over the entire flooded area, but the available habitat area during the high flood was three times larger that of the low flood year (Fig. 2). During the high flood year, small fish occurred everywhere in abundance, even close up to the water’s edge among the perennial grasses. The floodplain habitat for fish was therefore much larger and fish predators could not use the advantages of prey congestion. Habitat shortage is an important factor in fish egg survival (Hilborn and Walters 1992). The positive relationship between fish yield and flood size for African floodplain fisheries (Welcomme 1979) also seems to pertain in the Okavango Delta. Most of the fish predators are also large: fishes like Clarias spp., fish-eating birds of which there are more than 40 species in the Delta, as well as otters and crocodiles. They have long life cycles in comparison with their prey, however, and cannot respond through a rapid increase in reproduction to the sudden outbursts of food availability.
Years of low flood events on un-burned floodplains imply high abundance of decaying plant litter and thus low oxygen levels. This, in turn, promotes methane production and has at least two important implications. Firstly, methane emissions are likely to make this huge delta area a significant source of atmospheric methane. Secondly, as revealed by stable isotope studies (Lindholm 2006; Lindholm et al. 2007), methane may be a significant source of carbon for the aquatic food web via methanotrophic bacteria being consumed by the dense biomass of filter-feeding cladocerans—that also serve as an important food source for fish fry.
Synthesis and conclusions
General characteristics of aquatic ecosystems on seasonal floodplains after fire, low-, and high floods
Dominant food chains
Pelagic primary production
The synergism between flooding and fire on the seasonal floodplains is expressed in several ways. High flooding frequency causes a higher fire frequency up to a certain level. Fires reduce anaerobic processes and limit heterotrophic food chains. This results in high oxygen concentrations in the water, which affects nutrient cycling, in particular of phosphorous. In addition, fires make these floodplains accessible to organisms with high oxygen demand. Fires also result in a low abundance of fish fry, to a certain extent irrespective of hydroperiod. Most fires are caused by humans (Heinl et al. 2006) and occur probably with much higher frequency than natural, and this might have long term effects on Delta biodiversity and productivity.
However, the amount of seasonal flooding has a larger ecosystem impact than fires and is the primary factor in the wetland’s productivity It is not only decisive for the frequency of subsequent fires but also has a much larger effect than fire on nutrient accumulation and mobilization. This view is consistent with other studies that demonstrate, through artificial fire experiments on floodplains, that flooding is the primary factor which determines plant species composition (Ishida et al. 2008).
It seems the seasonal floodplains have a considerable ability to accumulate phosphorous, the main limiting nutrient in the Delta. Phosphorous is brought into the floodplains by the seasonal flood. It is assimilated by aquatic biota and then stored during the dry phase in the soil-sediment matrix, except for that which is found in perennial semi-aquatic grasses and that used by herbivores during this period. It is the switching between aerobic and anaerobic conditions that accumulates and releases phosphorous. Permanently dry or permanently wet conditions will not create these dynamics. Likewise, changes in flooding regimes which cause more stable hydrological conditions with more semi-permanent or permanent swamps will increase accumulation of litter, promote lower oxygen levels, increase denitrification rates, and cause nitrogen limitation in both terrestrial and aquatic phases. These two processes (phosphorus accumulation in the sediment and dissolution; organic matter accumulation and decomposition) are in essence probably the most important consequences of the flood pulse (sensu Junk et al. 1989) in the Okavango Delta.
These seasonal floodplains seem to be subject to intermediate disturbance (sensu Connell 1978) where both terrestrial and aquatic life forms can co-exist under temporal separation, and which provide for greater productivity on an annual basis and in the long run (Lindholm et al. 2009). They have a specific identity in the Okavango landscape. Thus, there is a close link between nutrient dynamics and productivity in both the terrestrial and aquatic phases of floodplains that are likely to be decisive for the total aquatic and terrestrial productivity of the Delta. These dynamically integrated parts should therefore be managed together, which is a challenge in a country like Botswana where management of water, wetlands, and dry lands is guided by fragmented legislation under the administration of different government ministries.
This study was partially funded by Grant no. 154410/720 from the Norwegian Research Council, and partially by the University of Botswana.
- Bonyongo MC (1999) Vegetation ecology of the seasonal floodplains in the Okavango Delta, Botswana. Dissertation, University of Pretoria, South AfricaGoogle Scholar
- Bonyongo MC, Bredenkamp GJ, Veenendaal E (2000) Floodplain vegetation in the Nxaraga Lagoon area, Okavango Delta, Botswana. S Afr J Bot 66(1):15–21Google Scholar
- Cronberg GA, Gieske A, Martins E, Prince-Nengu J, Stenstrom I (1996) Major ion chemistry, plankton, and bacterial assemblages of the Jao/Boro River, Okavango Delta, Botswana: the swamps and flood plains. Archiv für Hydrobiologie, Supplement b and Monographische Beiträge. ISSN 0341-2881 3:335–407Google Scholar
- DeBano LF, Neary DG, Ffolliott PF (1998) Fire’s effects on ecosystems. Wiley, New YorkGoogle Scholar
- Hilborn R, Walters CJ (1992) Quantitative fisheries stock assessment: choice, dynamics and uncertainty. Chapman and Hall, New YorkGoogle Scholar
- Junk WJ, Bayley PB, Sparks RE (1989) The flood pulse concept in river-floodplain systems. In Dodge DP (ed) Proceedings of the international Large River Symposium. J Can Fish Aquat Sci 106:110–127Google Scholar
- Lindholm M (2006) Zooplankton diversity and successions in the Okavango Delta, Botswana, diversity and ecosystem interactions in land/water ecotones. Dissertation, University of Oslo, NorwayGoogle Scholar
- Meyer T (1999) Ecological mappings in the research area of the Harry Oppenheimer Okavango Research Centre, Okavango Delta, Botswana. Dissertation, Hochschule Anhalt, Bernburg, GermanyGoogle Scholar
- Mitsch WJ, Gosselink JG (2007) Wetlands, 4th edn. Wiley, New YorkGoogle Scholar
- Rutz D (2004) The influence of fire on site productivity in southern Africa (study site: Okavango Delta, Botswana). Dissertation, Technische Universitat Munich, GermanyGoogle Scholar
- Welcomme RL (1979) The inland fisheries of Africa. Committee for inland fisheries of Africa occasional paper no. 7. FAO/UNDP, RomeGoogle Scholar
- Wetzel RG (2001) Limnology, lake and river ecosystems. Elsevier, San DiegoGoogle Scholar