Global biodiversity is in decline, driven in part by changes in the spatial characteristics of landscapes through land use change (Jaureguiberry et al., 2022; Maxwell et al., 2016). Land use change has accelerated from the late 19^th century onwards as the intensity of terrestrial direct exploitation and its indirect effects increased (Ellis et al., 2021). This fast transition has a wide geographical scope as human activities have directly affected more than two thirds of the earth’s land surface (Luyssaert et al., 2014). Therefore, investigating the pathways through which land use changes are altering patterns of biodiversity is key to any attempt to mitigate or reverse the trend.

Environmental heterogeneity describes the spatial variability present in an ecosystem in its biotic and abiotic elements and has been identified as a key spatial driver of biodiversity (Johnson and Simberloff, 1974; MacArthur and MacArthur, 1961; Stein et al., 2014; Tonetti et al., 2023; Udy et al., 2021). Changes to terrestrial land use that alter environmental heterogeneity are therefore important for understanding changes in biodiversity patterns. Environmental heterogeneity takes two key forms: compositional, which describes the variety of landscape elements present, and configurational, which describes their spatial arrangement (Li and Reynolds, 1995; Fahrig et al., 2011). Heterogeneity is thought to determine biodiversity through three mechanisms (Stein et al., 2014): increasing the available niche space, promoting species coexistence (Potts et al., 2004; Regolin et al., 2020); providing refugia from environmental extremes, enabling species persistence (Fjeldså et al., 2012); and, over the long term, promoting speciation through niche space differentiation (Antonelli and Sanmartín, 2011). The implication of this relationship is that the temporal process of the temporal process of landscape simplification through the loss of heterogeneity will result in a loss of biodiversity.

However, increasing environmental heterogeneity may not always promote higher biodiversity. Temporal increases in environmental heterogeneity within a given area will eventually reach a point where the reduction in the amount of suitable space for individual species increases the likelihood of stochastic extinctions, referred to as an “area-heterogeneity trade-off” (Allouche et al., 2012). Therefore, at very high levels of heterogeneity, suitable habitat patches may be so fragmented that they cannot support viable populations viable populations resulting in overall fewer overall fewer species present. Hence, biodiversity peaks at intermediate levels of environmental heterogeneity (Fahrig et al., 2011).

While spatial aspects have been frequently explored (e.g. Moreno-Rueda and Pizarro, 2009; Chiron et al., 2010; Stein et al., 2014; Tonetti et al., 2023), the temporal dynamics have not received the same level of focus. We define the temporal dynamics of spatial environmental heterogeneity as changes through time in the variability of environmental variables across space. This is a separate concept to temporal heterogeneity which refers to variability through time that is experienced homogeneously across a spatial extent (Menge and Sutherland, 1976; Stein et al., 2014). Crucially, understanding these dynamics is fundamental to ecological theory as well as being important practical context for ecological interventions aiming to bend the curve of biodiversity loss. Among 100 key research questions in ecology, the influence of spatial and temporal environmental heterogeneity on diversity across scales remains unanswered (Sutherland et al., 2013). Despite this importance there are currently no systematic reviews addressing temporal dynamics of heterogeneity – biodiversity relationship based on longitudinal temporal studies.

Changes to spatial environmental heterogeneity through time present important shifts in niche space availability (Macarthur and Levins, 1967; Stein et al., 2014). Species-specific responses to these changes are mediated by traits including trophic level, dispersal ability, and life history (Hutchinson, 1957). The cumulative effect of these trait-mediated responses produces the community level biodiversity response to changes in environmental heterogeneity (McGill et al., 2006). Over multiyear time scales, meta-community processes such as dispersal, colonization, and local extinction will play a strong role in biodiversity responses alongside changes in niche availability (Leibold et al., 2004; Jarzyna et al., 2025). Conversely, short-term dynamics may be characterized by responses to sudden changes in heterogeneity, refugia from climatic extremes that promote species persistence, and seasonality, giving rise to interlinked spatial and temporal partitioning through storage effects (Ryo et al., 2019; Doxa et al., 2022; Chesson, 2000).

This gap in our current understanding of the effects of temporal environmental heterogeneity on biodiversity is an obstacle to enhancing ecological theory and hinders efforts to address the biodiversity crisis. In this study, we set out to determine the current state of knowledge regarding how temporal dynamics shape the spatial heterogeneity-biodiversity relationship and propose a conceptual framework to guide future research. We propose three priorities for future research: long-term trends, short-term dynamics, and interactions with spatial grain.

Following the PRISMA guidelines (Page et al., 2021), we conducted a comprehensive systematic literature search of Web of Science using a string capturing three key concepts: heterogeneity, biodiversity, and temporality (S1). We screened these results according to inclusion criteria that the study must be observational, original, and peer reviewed. It must also consider terrestrial ecosystems and not microorganisms. Either environmental heterogeneity and/or biodiversity must have been quantitatively sampled longitudinally, meaning at least two time points in the same place. In longitudinal studies, time is considered as an independent variable, allowing a deeper understanding of phenomena that are not observable with a cross-sectional approach (Lindenmayer et al., 2012). Longitudinal designs avoid the issues linked to space for time substitution studies that may lead to erroneous conclusions (Damgaard, 2019). The initial results of the search returned 2180 papers published between 1990 and 2024. After screening by title and abstract, 149 articles remained for full text screening, and a final group of 23 papers met the full screening criteria (Supplementary Figure S2).

Subsequently, we extracted the following key information from the selected papers: the spatial grain, the landscape type, the biodiversity metric used, along with the taxa studied and the timespan of biodiversity and heterogeneity sampling. Landscape types were categorized as agricultural, natural and semi-natural. Heterogeneity measures were categorized according to the typology outlined by Stein and Kreft (2015): vegetation, land cover, soil, topography, and climate. Spatial grains of heterogeneity analysis for each paper were harmonized to km. Where grain was reported as an area, the side length of the square plot or the diameter of the circle plot was recorded. The dates of biodiversity and heterogeneity samples were recorded where available.

Extracted relationships were grouped into five categories based on their primary significant direction, grounded in established ecological theory as shown in Figure 1c. “Positive” relationships were identified where at least one spatial or temporal scale showed a positive association, a result generally expected under niche theory (Stein et al., 2014). Conversely, “negative” relationships, defined by a negative association at one or more scales, were expected due to the area-heterogeneity trade-off where extreme fragmentation increases stochastic extinction (Allouche et al., 2012; Fahrig et al., 2011). Results were labeled non-significant when no significant associations emerged across any scale, potentially indicating a spatiotemporal mismatch or response lag (Ryo et al., 2019). Finally, studies were categorized as “both positive and negative” where directions varied across scales, or “significant change in composition” where richness remained stable.

We found just eight papers that measured both biodiversity and heterogeneity through time, 12 measured just biodiversity through time, and 3 only sampled heterogeneity temporally. Among the 23 that passed screening, there was a deficit of studies considering both longer biodiversity and heterogeneity datasets. However, this seems to be driven by a broader pattern of a lack of long-term biodiversity data rather than historical heterogeneity data (Figure 1a).

Across the 20 of 23 papers that measured biodiversity through time, the independent variables fell into 3 categories: time since disturbance/intervention (Barton et al., 2014; Harvey and Holzman, 2014; Hodecek et al., 2016; Owens et al., 2014; Webster and Halpern, 2010; Eckerter et al., 2021; Doyon et al., 2005), seasonality (Apellaniz et al., 2012; Atkinson et al., 2002; Corcos et al., 2017; Duflot et al., 2016; Gonzalez-Megias et al., 2011; Jimenez-Navarro et al., 2023; Martinez-Nunez et al., 2022), and long-term temporal intervals utilizing sampling years or intervals (Duan et al., 2019; Wretenberg et al., 2010; Papanikolaou et al., 2017; Rusch et al., 2016; Herrault et al., 2016; Vanbergen et al., 2005). Within these studies, the specific environmental heterogeneity metrics used as independent variables included land cover composition (e.g. Shannon-Weaver index of land use), vegetation structure (e.g. coefficient of variation of stem density), and landscape configuration (e.g. ecological connectivity index) as shown in Figure 1c. We clarify that only the subset of eight studies re-measured heterogeneity metrics at each time point, they are identified with an asterisk in Figure 1.

Among all 23 papers, 5 heterogeneity datasets looked at time periods longer than 20 years compared to just 1 biodiversity dataset of that time scale. Conversely, 12 of the biodiversity datasets used were less than one year in duration, compared to only 1 heterogeneity dataset of that length. Figure 1a shows the frequency of biodiversity and heterogeneity sampling for all 23 papers. For studies that measured both variables longitudinally, apart from (Webster and Halpern, 2010), most had sparse sampling between the start and end of the study. For some studies, particularly (Duan et al., 2019) and (Barton et al., 2014), the sampling of each variable category did not take place simultaneously.

The spatial grain used to quantify heterogeneity also varied significantly, studies using plot-based designs varied in size from 0.01km to 10km (Figure 1b). However, it should be noted that the 10km grain seems to be an outlier, and the two papers that consider this scale both looked at vertebrate species. Across the 23 papers, only four papers looked at more than one spatial grain of environmental heterogeneity. Five papers based their analysis grain on the dispersal ability of the response taxa, two conformed to the resolution of the biodiversity data. However, this means that 12 out of the 23 papers did not give a specific justification for their chosen spatial grain for measuring heterogeneity.

Across all 23 papers, including the multi-taxa study by Tello et al. (2020), 8 papers analyzed birds, making them the most frequently studied taxon. Notably, this meant that of the 8 studies that considered both heterogeneity and biodiversity temporally, 4 used birds as their study taxa. The majority (14) of the studies were situated in agricultural landscapes, 4 semi-natural landscapes, 5 were based in natural landscapes.

Despite claims of its importance, our understanding of the impacts of temporal dynamics of environmental heterogeneity on biodiversity is scattered and limited. The limited extent of literature meeting our criteria echoes a wider neglect of temporality in ecology in comparison to work that focusses on spatial phenomena (Ryo et al., 2019). The consequences of such gaps are captured by the term the “invisible present” which describes the shortsightedness that emerges without long-term research into processes that occur on timescales imperceptible to anecdotal observations (Magnuson, 1990). He suggested that “in the absence of the temporal context provided by long-term research, serious misjudgments can occur not only in our attempts to understand and predict change in the world around us, but also in our attempts to manage our environment”.

The gap in temporal data limits our ability to test how environmental heterogeneity–biodiversity relationships unfold through time. For example, an intervention predicted to increase biodiversity on a spoil heap in Czechia through adding topsoil and planting trees resulted in lower functional diversity than naturally colonized areas when monitored over a 14-year period (Hodecek et al., 2016). This outcome underscores how essential temporal monitoring is to validate ecological predictions. Our framework allows evaluating long-term dynamics in the relationships, that cannot be captured by static predictions.

The main limitation preventing the inclusion of papers was inadequate quantification of heterogeneity. Of the 125 studies excluded during the full text screening, 72 were rejected on this basis. Frequently, the relationship between heterogeneity and biodiversity was only implied by extrapolation in the discussion of the paper. For example, studies investigating landscape-scale patterns often focus on different environmental factors, such as fire history or grazing intensity. However, these factors are frequently studied in a chrono-sequence of patches as a proxy for succession through time e.g (Hovick et al., 2015). Furthermore, many forestry and succession studies used treatment prescriptions as proxies for heterogeneity rather than taking quantitative measurements after intervention.

Despite limited literature, we found that between the same two variables the relationship between environmental heterogeneity and biodiversity varied significantly across different spatial and temporal scales (Figure 1c). While it is too early to draw firm conclusions, this reinforces the importance of investigating our three suggested domains of temporal dynamics as a structured way to address the knowledge gap.

What could a well-designed investigation into the temporal dynamics of the environmental heterogeneity biodiversity relationship reveal? With appropriate sampling and analysis, we would observe important differences in temporal behavior between study systems. There is a spatial precedent for context-dependent patterns. Despite a prediction of maximum biodiversity at intermediate levels of heterogeneity, specific systems are expected to display a variety of patterns depending on their taxonomic focus and scales of analysis (Allouche et al., 2012). The same ecological processes that give rise to spatial variation also govern how these systems behave through time. It is therefore plausible that they may also shape temporal dynamics.

Informed by our preliminary body of literature we identify three key domains, based on the ecological processes identified in the introduction, within which differences are expected (see Figure 2): (a) long-term trends, (b) short-term trends, and (c) differences across spatial grains. Where long-term trends refer to a cumulative trajectory defined by statistical properties such as trend stationarity, short-term dynamics are events that occur within the period of a trajectory (Ryo et al., 2019). Distinguishing between systems based on these temporal variables would allow us to better understand and predict the effects of past changes and forecast the impacts of future changes and interventions.

We found a paucity of studies that sampled biodiversity and heterogeneity longitudinally through time, reflecting a wider neglect of temporality in ecology. Nevertheless, where research addressed these issues, we observed complex temporal dynamics that distinguish ecological systems from each other. Our typology suggests that short-term dynamics, long-term trajectories, and the temporal interaction with spatial grain are fruitful avenues for future investigation. Exploring these complexities further will not only provide additional insight into ecological theory but also inform practical interventions in conservation and management, ensuring we can adequately understand the changes that have already occurred and predict those to come.