Ecosystems and social systems are hardly ever static: they change over time, sometimes slowly and sometimes abruptly, and in response to both internal dynamics and external drivers. These changes play out over the surface of the Earth in the form of observable ‘patterns’, at a range of scales, from that of individual organisms, to patches showing similar attributes, to whole ecosystems, and to the entire biosphere. There is a broad tendency (which is by no means a fixed rule) that slow processes are associated with large spatial scales, and fast processes with small scales (Vance & Doel 2010). For useful information to be gained, it is necessary to match the scales of observation, in both time and space, with the scales of what is being observed (Englund & Cooper 2003). Furthermore, it is plausible that interventions in ecosystem processes are more effective if they are matched to the scale of the phenomenon being addressed.

It is therefore an area of concern that the majority of the data collected relating to the state and functioning of the natural and human world is on small scales and over short periods of time (Carpenter 1996; Englund & Cooper 2003; Levin 1992). This is partly because information at large scales and over long periods was not part of our human experience until relatively recently: we observed locally, and usually for no longer than a human lifespan (and often much less). Written records in archives, travel, telecommunications and space exploration have changed all that. At the same time, human dominance of the world has led to a range of phenomena on regional and global scale – such as climate change, biodiversity loss and ecosystem degradation – which require investigation and intervention on scales far beyond the local and immediate.

There has been a trend in science towards finer and finer resolution, more detail, increased specialisation and narrower disciplines. The scientific approach of reductionism – looking for the causes of things by isolating them and delving down into their underlying elements – has been enormously successful in many fields; however, it has been unhelpful in solving issues which are systemic in nature. System problems emerge from interactions and context, and require a wider rather than a narrower view; a view which is often associated with larger scales in space and time. Fortunately, theoretical and technological advances the past few decades have made it increasingly feasible to observe, experiment with and manage systems on the scales at which they actually operate (Schindler 1998).

There is a growing realisation that the phenomena relevant to environmental management operate on a range of scales, rather than on a single scale. Even within a class of issues (‘desertification’, for instance) several nested scales are often relevant. Thus there is seldom a single ‘best’ scale for all purposes. The trend is towards multi-scale observation and analysis, and in particular towards trying to understand how phenomena propagate across scales and interact between scales (Scholes et al. 2013).

In classical science, still reflected in science education and attitudes towards data ownership, the researcher made his or her own observations. This is not technically feasible on large scales and over long times; we depend on the accumulation of data from many individuals, the custodianship of institutions and complex and expensive equipment, such as satellites, which are designed and operated by specialists. As a result, the activity of making accurate, repeated observations for use by other scientists is a legitimate aspect of science in its own right, which must be funded and recognised as such.

The act which governs South African protected areas (National Environment Management: Protected Areas Act [Act no. 57 of 2003]) states that the purpose of national parks, amongst other things, is to ‘provide spiritual, scientific, educational, recreational and tourism opportunities which are environmentally compatible’. Therefore, conducting scientific activities within parks should not be at the whim of individual park managers: it is part of their mandate, provided it is not in conflict with the broad objective of environmental protection. Nor should such research be selected solely on the basis of its immediate usefulness for park management. The scale, relative absence of human impact, institutional stability and infrastructure of national parks make them uniquely appropriate for certain types of research. Research activities, properly designed and communicated, seldom conflict materially with biodiversity management and tourism activities: generally, visitors are interested in the research and find that it adds value to their experience. Visits by scientists for purposes of research are a significant source of income to the park system.

National parks have accumulated data of value to both themselves and the wider community. Kruger National Park (KNP) is exemplary here, because of its history and size. In particular, the long-term records pertaining to weather, vegetation composition and state, the populations of mammals and birds and areas affected by fire are near-unique.

The philosophical shift from keeping such data for internal use only to making it openly and freely available which occurred in KNP over the past two decades has resulted in a massive increase in scientific publications relating to the park. It is recognised that collecting and servicing such datasets involves a significant cost to the park system. Experience in many organisations has shown that attempting to recover this cost directly from users is counter-productive: data becomes unaffordable to the users and the expense of invoicing exceeds the income stream. A more sensible approach economically (i.e. considering the needs of society as a whole, rather than the finances of the accounting unit alone) is to treat the cost of data collection as an operating expense to be centrally borne (perhaps supported by specific external grants) and then make the information freely available in order to maximise its use. Use of such data by third parties carries obligations: the data source must be correctly acknowledged; publications arising from its use must be lodged with the data collectors for reporting purposes; and data collected in or on the park must in turn become publically and freely available.

There are many domains in which research in large protected areas in South Africa has contributed to global scientific understanding; the following are a few examples amongst many.

Much of the day-to-day management in protected areas relates to symptomatic issues, immediate concerns and small scales. This bias affects the information needs as perceived by conservation managers; for example, they routinely prioritise issues relating to population excesses of ‘problem species’, population viability in rare species, localised degradation and specific instances of conflict with neighbours. The slower, larger-scale processes, which are more likely to threaten the ability to satisfy the mandate of protected area management, and which may be the root cause of many of the crises which dominate management attention, often go unrecorded and unrecognised until too late – land use changes in adjacent areas, climate and atmosphere changes, and economic and political shifts at regional to global scale are some examples. The KNP has been a leader in the practical implementation of adaptive management, the style suggested as most appropriate for ecosystem management in the presence of limited knowledge (Van Wilgen & Biggs 2011).