More stories

  • in

    Shifts in vegetation activity of terrestrial ecosystems attributable to climate trends

    Plant growth model without environmental forcingThe model without environmental forcing closely follows the original description of the Thornley transport resistance (TTR) model29. A summary of the model parameters is provided in Supplementary Table 2. The shoot and root mass pools (MS and MR, in kg structural dry matter) change as a function of growth and loss (equations (1) and (2)). The litter (kL) and maintenance respiration (r) loss rates (in kg kg−1 d−1) are treated as constants. In the original model description29 r = 0. The parameter KM (units kg) describes how loss varies with mass (MS or MR). Growth (Gs and Gr, in kg d−1) varies as a function of the carbon and nitrogen concentrations (equations (3) and (4)). CS, CR, NS and NR are the amounts (kg) of carbon and nitrogen in the roots and shoots. These assumptions yield the following equations for shoot and root dry matter,$${mathrm{MS}}[t+1]={mathrm{MS}}[t]+{G}_{{mathrm{S}}}[t]-frac{({k}_{{mathrm{L}}}+r){mathrm{MS}}[t]}{1+frac{{K}_{{{M}}}}{{mathrm{MS}}[t]}},$$
    (1)
    $${mathrm{MR}}[t+1]={mathrm{MR}}[t]+{G}_{{mathrm{R}}}[t]-frac{({k}_{{mathrm{L}}}+r){mathrm{MR}}[t]}{1+frac{{K}_{{{M}}}}{{mathrm{MR}}[t]}},$$
    (2)
    where GS and GR are$${G}_{{mathrm{S}}}=gfrac{{mathrm{CS}}times {mathrm{NS}}}{{mathrm{MS}}},$$
    (3)
    $${G}_{{mathrm{R}}}=gfrac{{mathrm{CR}}times {mathrm{NR}}}{{mathrm{MR}}},$$
    (4)
    and g is the growth coefficient (in kg kg−1 d−1).Carbon uptake UC is determined by the net photosynthetic rate (a, in kg kg−1 d−1) and the shoot mass (equation (5)). Similarly, nitrogen uptake (UN) is determined by the nitrogen uptake rate (b, in kg kg−1 d−1) and the root mass. The parameter KA (units kg) forces both photosynthesis and nitrogen uptake to be asymptotic with mass. The second terms in the denominators of equations (5) and (6) model product inhibitions of carbon and nitrogen uptake, respectively; that is, the parameters JC and JN (in kg kg−1) mimic the inhibition of source activity when substrate concentrations are high,$${U}_{{mathrm{C}}}=frac{a{mathrm{MS}}}{left(1+frac{{mathrm{MS}}}{{K}_{{mathrm{A}}}}right)left(1+frac{{mathrm{CS}}}{{mathrm{MS}}times {J}_{{mathrm{C}}}}right)},$$
    (5)
    $${U}_{{mathrm{N}}}=frac{b{mathrm{MR}}}{left(1+frac{{mathrm{MR}}}{{K}_{{mathrm{A}}}}right)left(1+frac{{mathrm{NR}}}{{mathrm{MR}}times {J}_{{mathrm{N}}}}right)}.$$
    (6)
    The substrate transport fluxes of C and N (τC and τN, in kg d−1) between roots and shoots are determined by the concentration gradients between root and shoot and by the resistances. In the original model description29, these resistances are defined flexibly, but we simplify and assume that they scale linearly with plant mass,$${tau }_{{mathrm{C}}}=frac{{mathrm{MS}}times {mathrm{MR}}}{{mathrm{MS}}+{mathrm{MR}}}left(frac{{mathrm{CS}}}{{mathrm{MS}}}-frac{{mathrm{CR}}}{{mathrm{MR}}}right)$$
    (7)
    $${tau }_{{mathrm{N}}}=frac{{mathrm{MS}}times {mathrm{MR}}}{{mathrm{MS}}+{mathrm{MR}}}left(frac{{mathrm{NR}}}{{mathrm{MR}}}-frac{{mathrm{NS}}}{{mathrm{MS}}}right)$$
    (8)
    The changes in mass of carbon and nitrogen in the roots and shoots are then$${mathrm{CS}}[t+1]={mathrm{CS}}[t]+{U}_{{mathrm{C}}}[t]-{f}_{{mathrm{C}}}{G}_{{mathrm{s}}}[t]-{tau }_{{mathrm{C}}}[t]$$
    (9)
    $${mathrm{CR}}[t+1]={mathrm{CR}}[t]+{tau }_{{mathrm{C}}}[t]-{f}_{{mathrm{C}}}{G}_{{mathrm{r}}}[t]$$
    (10)
    $${mathrm{NS}}[t+1]={mathrm{NS}}[t]+{tau }_{{mathrm{N}}}[t]-{f}_{{mathrm{N}}}{G}_{{mathrm{s}}}[t]$$
    (11)
    $${mathrm{NR}}[t+1]={mathrm{NR}}[t]+{U}_{{mathrm{N}}}[t]-{f}_{{mathrm{N}}}{G}_{{mathrm{r}}}[t]-{tau }_{{mathrm{N}}}[t]$$
    (12)
    where fC and fN (in kg kg−1) are the fractions of structural carbon and nitrogen in dry matter.Adding environmental forcing to the plant growth modelIn this section, we describe how the net photosynthetic rate (a), the nitrogen uptake rate (b), the growth rate (g) and the respiration rate (r) are influenced by environmental-forcing factors. These environmental-forcing effects are described in equations (13)–(17) and summarized graphically in Extended Data Fig. 1. All other model parameters are treated as constants. Previous work that implemented the TTR model as a species distribution model30 is used as a starting point for adding environmental forcing. As in this previous work30, we assume that parameters a, b and g are co-limited by environmental factors in a manner analogous to Liebig’s law of the minimum, which is a crude but pragmatic abstraction. The implementation here differs in some details.Unlike previous work30, we use the Farquhar model of photosynthesis47,48 to represent how solar radiation, atmospheric CO2 concentration and air temperature co-limit photosynthesis35. We assume that the Farquhar model parameters are universal and that all vegetation in our study uses the C3 photosynthetic pathway. The Farquhar model photosynthetic rates are rescaled to [0,amax] to yield afqr. The effects of soil moisture (Msoil) on photosynthesis are represented as an increasing step function ({{{{S}}}}(M_{mathrm{soil}},{beta }_{1},{beta }_{2})=max left{min left(frac{M_{mathrm{soil}}-{beta }_{1}}{{beta }_{2}-{beta }_{1}},1right),0right}). This allows us to redefine a as,$$a={a}_{{mathrm{fqr}}} {{{{S}}}}(M_{mathrm{soil}},{beta }_{1},{beta }_{2})$$
    (13)
    The processes influencing nitrogen availability are complex, and global data products on plant available nitrogen are uncertain. We therefore assume that nitrogen uptake will vary with soil temperature and soil moisture. That is, the nitrogen uptake rate b is assumed to have a maximum rate (bmax) that is co-limited by soil temperature Tsoil and soil moisture Msoil,$$b={b}_{{mathrm{max}}} {{{{S}}}}({T}_{soil},{beta }_{3},{beta }_{4}) {{{{Z}}}}(M_{mathrm{soil}},{beta }_{5},{beta }_{6},{beta }_{7},{beta }_{8}).$$
    (14)
    In equation (14), we have assumed that the nitrogen uptake rate is a simple increasing and saturating function of temperature. We have also assumed that the nitrogen uptake rate is a trapezoidal function of soil moisture with low uptake rates in dry soils, higher uptake rates at intermediate moisture levels and lower rates once soils are so moist as to be waterlogged. The trapezoidal function is ({{{{Z}}}}(M_{mathrm{soil}},{beta }_{5},{beta }_{6},{beta }_{7},{beta }_{8})=max left{min left(frac{M_{mathrm{soil}}-{{{{{beta }}}}}_{5}}{{{{{{beta }}}}}_{6}-{{{{{beta }}}}}_{5}},1,frac{{{{{{beta }}}}}_{8}-M_{mathrm{soil}}}{{beta }_{8}-{beta }_{7}}right),0right}).The previous sections describe how the assimilation of carbon and nitrogen by a plant are influenced by environmental factors. The TTR model describes how these assimilate concentrations influence growth (equations (3) and (4)). In our implementation, we additionally allow the growth rate to be co-limited by temperature (soil temperature, Tsoil) and soil moisture (Msoil),$$g={g}_{{mathrm{max}}} {{{{Z}}}}({T}_{{mathrm{soil}}},{beta }_{9},{beta }_{10},{beta }_{11},{beta }_{12}) {{{{S}}}}(M_{mathrm{soil}},{beta }_{13},{beta }_{14}).$$
    (15)
    We use Tsoil since we assume that growth is more closely linked to soil temperature, which varies slower than air temperature. The respiration rate (r, equations (1) and (2)) increases as a function of air temperature (Tair) to a maximum rmax,$$r={r}_{{mathrm{max}}}{{{{S}}}}({T}_{{mathrm{air}}},{beta }_{15},{beta }_{16}).$$
    (16)
    The parameter r is best interpreted as a maintenance respiration. Growth respiration is not explicitly considered; it is implicitly incorporated in the growth rate parameter (g, equation (15)), and any temperature dependence in growth respiration is therefore assumed to be accommodated by equation (15).Fire can reduce the structural shoot mass MS as follows,$${mathrm{MS}}[t+1]={mathrm{MS}}[t](1-{{{{S}}}}(F,{beta }_{17},{beta }_{18})).$$
    (17)
    where F is an indicator of fire severity at a point in time (for example, burnt area) and the function S(F, β17, β18) allows MS to decrease when the fire severity indicator F is high. If F = 0, this process plays no role. This fire impact equation was used in preliminary analyses, but the data on fire activity did not provide sufficient information to estimate β17 and β18; we therefore excluded this process from the final analyses.We further estimate two additional β parameters (βa and βb) so that each site can have unique maximum carbon and nitrogen uptake rates. Specifically, we redefine a as ({a}^{{prime} }={beta }_{a} a) and b as ({b}^{{prime} }={beta }_{b} b).Data sources and preparationTo describe vegetation activity, we use the GIMMS 3g v.1 NDVI data26,27 and the MODIS EVI28 data. The GIMMS data product has been derived from the AVHRR satellite programme and controls for atmospheric contamination, calibration loss, orbital drift and volcanic eruptions26,27. The data provide 24 NDVI raster grids for each year, starting in July 1981 and ending in December 2015. The spatial resolution is 1/12° (~9 × 9 km). The EVI data used are from the MODIS programme’s Terra satellite; it is a 1 km data product provided at a 16-day interval. We use data from the start of the record (February 2000) to December 2019. The MODIS data product (MOD13A2) uses a temporal compositing algorithm to produce an estimate every 16 days that filters out atmospheric contamination. The EVI is designed to reduce the effects of atmospheric, bare-ground and surface water on the vegetation index28.For environmental forcing, we use the ERA5-Land data31,32 (European Centre for Medium-Range Weather Forecasts Reanalysis v. 5; hereafter, ERA5). The ERA5 products are global reanalysis products based on hourly estimates of atmospheric variables and extend from present back to 1979. The data products are supplied at a variety of spatial and temporal resolutions. We used the monthly averages from 1981 to 2019 at a 0.1° spatial resolution (~11 km). The ERA5 data provide air temperature (2 m surface air temperature), soil temperature (0–7 cm soil depth), surface solar radiation and volumetric soil water (0–7 cm soil depth). Fire data were taken from the European Space Agency Fire Disturbance Climate Change Initiative’s AVHRR Long-Term Data Record Grid v.1.0 product49. This product provides gridded (0.25° resolution) data of monthly global (from 1982 to 2017) burned area derived from the AVHRR satellite programme. As mentioned, the fire data did not enrich our analysis, and the analyses we present here therefore exclude further consideration of the fire data.All data were resampled to the GIMMS grid. The mean pixel EVI was then calculated for each GIMMS cell for each time point in the MODIS EVI data. We used linear interpolation on the NDVI, EVI and ERA5 environmental-forcing data to estimate each variable on a weekly time step. This served to set the time step of the TTR difference equations to one week and to synchronize the different time series.Site selectionThe GIMMS grid cells define the spatial resolution of our sample points. GIMMS grid cells are large (1/12°, ~9 km), meaning that most grid cells contain multiple land-cover types. We focused on wilderness landscapes, and our aim was to find multiple grid cells for the major ecosystems of the world. We used the following classification of ecosystem types to guide the stratification of our grid-cell selection: tropical evergreen forest (RF), boreal forest (BF), temperate evergreen and temperate deciduous forest (TF), savannah (SA), shrubland (SH), grassland (GR), tundra (TU) and Mediterranean-type ecosystems (MT).We used the following criteria to select grid cells. (1) Selected grid cells should contain relatively homogeneous vegetation. Small-scale heterogeneity was allowed (for example, catenas, drainage lines, peatlands) as long as many of these elements are repeated in the scene (for example, rolling hills were accepted, but elevation gradients were rejected). (2) Sites should have no signs of transformative human activity (for example, tree harvesting, crop cultivation, paved surfaces). We used the Time Tool in Google Earth Pro, which provides annual satellite imagery of the Earth from 1984 onwards, to ensure that no such activity occurred during the observation period (note that the GIMMS record starts in July 1981; however, it is likely that evidence of transformative activity between July 1981 and 1984 would be visible in 1984). Grid cells with extensive livestock holding on non-improved pasture were included. In some cases, a small agricultural field or pasture was present, and such grid cells were used as long as the field or pasture was small and remained constant in size. (3) Grid cells should not include large water bodies, but small drainage lines or ponds were accepted as long as they did not violate the first criterion. (4) Grid cells should be independent (neighbouring grid cells were not selected) and cover the major ecosystems of the world. Using these criteria, we were able to include 100 sites in the study (Extended Data Figs. 2 and 3 and Supplementary Table 4).State-space modelWe used a Bayesian state-space approach. Conceptually, the analysis takes the form,$$M[t]=f(M[t-1],{boldsymbol{beta}},{boldsymbol{theta}}_{t-1},{epsilon }_{t-1})$$
    (18)
    $${mathrm{VI}}[t]=m M[t]+eta .$$
    (19)
    Here M[t] is the plant growth model’s prediction of biomass (M = MS + MR) at time t, and ϵt−1 is the process error associated with the state variables. In the model, each underlying state variable (MS, MR, CS, CR, NS and NR) has an associated process error term. The function f(M[t − 1], β, θt−1, ϵt−1) summarizes that the development of M is influenced by the state variables MS, MR, CS, CR, NS and NR, the environmental-forcing data θt−1 and the β parameters. The observation equation (equation (19)) uses the parameter m to link the VI (vegetation index, either NDVI or EVI) observations to the modelled state M. The parameter η is the observation error. Equation (19) assumes that there is a linear relationship between modelled biomass (M) and VI, which is a simplification of reality50,51,52. The observation error η is structured by our simplification of the data products quality scores (coded Q = 0, 1, 2, with 0 being good and 2 being poor; Supplementary Table 3) to allow the error to increase with each level of the quality score. Specifically, we define η = e0 + e1 × Q.The model was formulated using the R package LaplacesDemon53. All β parameters are given vague uniform priors. The parameter m is given a vague normal prior (truncated to be >0). The process error terms are modelled using normal distributions, and the variances of the error terms are given vague half-Cauchy priors. The ex parameters are given vague normal priors. The model also requires the parameterization of M[0], the initial vegetation biomass; M[0] is given a vague uniform prior. We used the twalk Markov chain Monte Carlo (MCMC) algorithm as implemented in LaplacesDemon53 and its default control parameters to estimate the posterior distributions of the model parameters. We further fitted the model using DEoptim54,55, which is a robust genetic algorithm that is known to perform stably on high-dimensional and multi-modal problems56, to verify that the MCMC algorithm had not missed important regions of the parameter space. The models estimated with MCMC had significantly lower log root-mean-square error than models estimated with DEoptim (paired t-test NDVI analysis: t = –2.9806, degrees of freedom (d.f.) = 99, P = 0.00362; EVI analysis: t = –4.6229, d.f. = 99, P = 1.144 × 10–5), suggesting that the MCMC algorithm performed well compared with the genetic algorithm.Anomaly extraction and trend estimationWe use the ‘seasonal and trend decomposition using Loess’ (STL57) as implemented in the R58 base function stl. STL extracts the seasonal component s of a time series using Loess smoothing. What remains after seasonal extraction (the anomaly or remainder, r) is the sum of any long-term trend and stochastic variation. We estimate the trend in two ways. First, we estimate the trend by fitting a quadratic polynomial (r = a + bx + cx2) to the remainder (r is the remainder, x is time and a, b and c are regression coefficients). The use of polynomials allows the data to specify whether a trend exists, whether the trend is linear, cup or hat shaped and whether the overall trend is increasing or decreasing. As a second method, we estimate the trend by fitting a bent-cable regression to the remainder. Bent-cable regression is a type of piecewise linear regression for estimating the point of transition between two linear phases in a time series59,60. The model takes the form r = b0 + b1x + b2 q(x, τ, γ)60. Here r is the remainder, x is time, b0 is the initial intercept, b1 is the slope in phase 1, the slope in phase 2 is b2 − b1 and q is a function that defines the change point: (q(x,tau ,gamma )=frac{{(x-tau +gamma )}^{2}}{4gamma }I(tau -gamma < tau +gamma )+(x-tau )I(x > tau +gamma )); τ represents the location of the change point and γ the span of the bent cable that joins the two linear phases; I(A) is an indicator function that returns 1 if A is true and 0 if A is false. The bent-cable model allows the data to specify whether a trend exists and whether there has been a switch in the trend, thereby allowing the identification of whether the trend is linear, cup or hat shaped and whether the overall trend is increasing or decreasing. Both the polynomial and bent-cable models were estimated using LaplacesDemon’s53 Adaptive Metropolis MCMC algorithm and vague priors, although for the bent-cable model we constrained τ to be in the middle 70% of the time series and γ to be at most two years.The STL extraction of the seasonal components in the air temperature, soil temperature, soil moisture and solar radiation data (there is no stochasticity or seasonal trend in the CO2 data we used) allows us to simulate detrended time series d of these forcing variables as (d=bar{y}+s+{{{{N}}}}(mu ,sigma )) where N(μ, σ) is a normally distributed random variable with mean and standard deviation estimated from the remainder r (we verified that r was well described by the normal distribution), (bar{y}) is the mean of the data over the time series and s is the seasonal component extracted by STL. For CO2, the detrended time series is simply the average CO2 over the time series. More

  • in

    Enhancing the ecological value of oil palm agriculture through set-asides

    Phalan, B. et al. Crop expansion and conservation priorities in tropical countries. PLoS ONE 8, e51759 (2013).Article 
    CAS 

    Google Scholar 
    Poore, J. & Nemecek, T. Reducing food’s environmental impacts through producers and consumers. Science 360, 987–992 (2018).Article 
    CAS 

    Google Scholar 
    Springmann, M. et al. Options for keeping the food system within environmental limits. Nature 562, 519–525 (2018).Tilman, D., Balzer, C., Hill, J. & Befort, B. L. Global food demand and the sustainable intensification of agriculture. Proc. Natl Acad. Sci. USA 108, 20260–20264 (2011).Article 
    CAS 

    Google Scholar 
    Searchinger, T. D., Wirsenius, S., Beringer, T. & Dumas, P. Assessing the efficiency of changes in land use for mitigating climate change. Nature 564, 249–253 (2018).Edwards, D. P. et al. Conservation of tropical forests in the Anthropocene. Curr. Biol. 29, R1008–R1020 (2019).Article 
    CAS 

    Google Scholar 
    Newbold, T. et al. Global patterns of terrestrial assemblage turnover within and among land uses. Ecography 39, 1151–1163 (2016).Article 

    Google Scholar 
    Newbold, T. et al. Global effects of land use on local terrestrial biodiversity. Nature 520, 45–50 (2015).Article 
    CAS 

    Google Scholar 
    Gibbs, H. K. et al. Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proc. Natl Acad. Sci. USA 107, 16732–16737 (2010).Hansen, M. C. et al. High-resolution global maps of 21st-century forest cover change. Science 342, 850–853 (2013).Article 
    CAS 

    Google Scholar 
    Newbold, T. et al. A global model of the response of tropical and sub-tropical forest biodiversity to anthropogenic pressures. Proc. R. Soc. B 281, 20141371 (2014).Article 

    Google Scholar 
    Clough, Y. et al. Combining high biodiversity with high yields in tropical agroforests. Proc. Natl Acad. Sci. USA 108, 8311–8316 (2011).Article 
    CAS 

    Google Scholar 
    Giam, X. Global biodiversity loss from tropical deforestation. Proc. Natl Acad. Sci. USA 114, 5775–5777 (2017).Article 
    CAS 

    Google Scholar 
    van der Werf, G. R. et al. CO2 emissions from forest loss. Nat. Geosci. 2, 737–738 (2009).Article 

    Google Scholar 
    Harvey, C. A. et al. Climate‐smart landscapes: opportunities and challenges for integrating adaptation and mitigation in tropical agriculture. Conserv. Lett. 7, 77–90 (2014).Article 

    Google Scholar 
    Harris, N. L. et al. Baseline map of carbon emissions from deforestation in tropical regions. Science 336, 1573–1576 (2012).Article 
    CAS 

    Google Scholar 
    Song, X.-P. et al. Global land change from 1982 to 2016. Nature 560, 639–643 (2018).Quezada, J. C., Etter, A., Ghazoul, J., Buttler, A. & Guillaume, T. Carbon neutral expansion of oil palm plantations in the Neotropics. Sci. Adv. 5, eaaw4418 (2019).Article 
    CAS 

    Google Scholar 
    Oil Palm and Biodiversity: a Situation Analysis by the IUCN Oil Palm Task Force (International Union for Conservation of Nature, 2018). https://doi.org/10.2305/IUCN.CH.2018.11.enMeijaard, E. & Sheil, D. The moral minefield of ethical oil palm and sustainable development. Front. For. Glob. Change 2, 22 (2019).The Future of Food and Agriculture – Alternative Pathways to 2050 (FAO, 2018).Henders, S., Persson, U. M. & Kastner, T. Trading forests: land-use change and carbon emissions embodied in production and exports of forest-risk commodities. Environ. Res. Lett. 10, 125012 (2015).Article 

    Google Scholar 
    Donofrio, S., Rothrock, P. & Leonard, J. Supply Change: Tracking Corporate Commitments to Deforestation-Free Supply Chains (Forest Trends, 2017).Terrenoire, E., Hauglustaine, D. A., Gasser, T. & Penanhoat, O. The contribution of carbon dioxide emissions from the aviation sector to future climate change. Environ. Res. Lett. 14, 084019 (2019).Article 
    CAS 

    Google Scholar 
    Parsons, S., Raikova, S. & Chuck, C. J. The viability and desirability of replacing palm oil. Nat. Sustain. 3, 412–418 (2020).Article 

    Google Scholar 
    Taheripour, F., Hertel, T. W. & Ramankutty, N. Market-mediated responses confound policies to limit deforestation from oil palm expansion in Malaysia and Indonesia. Proc. Natl Acad. Sci. USA 116, 19193–19199 (2019).Article 
    CAS 

    Google Scholar 
    Laurance, W. F. et al. Improving the performance of the roundtable on sustainable palm oil for nature conservation. Conserv. Biol. 24, 377–381 (2010).Article 

    Google Scholar 
    Meijaard, E., Abrams, J. F., Juffe-Bignoli, D., Voigt, M. & Sheil, D. Coconut oil, conservation and the conscientious consumer. Curr. Biol. 30, R757–R758 (2020).Article 
    CAS 

    Google Scholar 
    Driving Change With Sustainable Palm Oil (Roundtable on Sustainable Palm Oil, accessed August 2022). https://rspo.org/aboutGarrett, R. D., Carlson, K. M., Rueda, X. & Noojipady, P. Assessing the potential additionality of certification by the round table on responsible soybeans and the roundtable on sustainable palm oil. Environ. Res. Lett. 11, 045003 (2016).Article 

    Google Scholar 
    Mittermeier, R. A., Myers, N., Mittermeier, C. G. & Robles, G. Hotspots: Earth’s Biologically Richest and Most Endangered Terrestrial Ecoregions (Conservation International, 1999).Gaveau, D. L. et al. Rapid conversions and avoided deforestation: examining four decades of industrial plantation expansion in Borneo. Sci. Rep. 6, 32017 (2016).Article 
    CAS 

    Google Scholar 
    Luke, S. H. et al. Riparian buffers in tropical agriculture: scientific support, effectiveness and directions for policy. J. Appl. Ecol. 56, 85–92 (2019).Article 

    Google Scholar 
    Mitchell, S. L. et al. Riparian reserves help protect forest bird communities in oil palm dominated landscapes. J. Appl. Ecol. 55, 2744–2755 (2018).Article 

    Google Scholar 
    Scriven, S. A. et al. Testing the benefits of conservation set-asides for improved habitat connectivity in tropical agricultural landscapes. J. Appl. Ecol. 56, 2274–2285 (2019).Article 

    Google Scholar 
    Deere, N. J. et al. Riparian buffers can help mitigate biodiversity declines in oil palm agriculture. Front. Ecol. Environ. 20, 459–466 (2021).Woodham, C. R. et al. Effects of replanting and retention of mature oil palm riparian buffers on ecosystem functioning in oil palm plantations. Front. Glob. Change 2, 29 (2019).Article 

    Google Scholar 
    Carlson, K. M. et al. Influence of watershed‐climate interactions on stream temperature, sediment yield, and metabolism along a land use intensity gradient in Indonesian Borneo. J. Geophys. Res. Biogeosci. 119, 1110–1128 (2014).Article 

    Google Scholar 
    Carlson, K. M. et al. Effect of oil palm sustainability certification on deforestation and fire in Indonesia. Proc. Natl Acad. Sci. USA 115, 121–126 (2018).Article 
    CAS 

    Google Scholar 
    Fleiss, S. et al. Conservation set-asides improve carbon storage and support associated plant diversity in certified sustainable oil palm plantations. Biol. Conserv. 248, 108631 (2020).Article 

    Google Scholar 
    Wunder, S., Angelsen, A. & Belcher, B. Forests, livelihoods, and conservation: broadening the empirical base. World Dev. 64, S1–S11 (2014).Struebig, M. J. et al. Quantifying the biodiversity value of repeatedly logged rainforests: gradient and comparative approaches from Borneo. Adv. Ecol. Res. 48, 183–224 (2013).Article 

    Google Scholar 
    Shevade, V. S. & Loboda, T. V. Oil palm plantations in Peninsular Malaysia: determinants and constraints on expansion. PLoS ONE 14, e0210628 (2019).Article 
    CAS 

    Google Scholar 
    Pirker, J., Mosnier, A., Kraxner, F., Havlík, P. & Obersteiner, M. What are the limits to oil palm expansion? Glob. Environ. Change 40, 73–81 (2016).Article 

    Google Scholar 
    Launching the RSPO Jurisdictional Approach (JA) Piloting Framework (Roundtable on Sustainable Palm Oil, accessed August 2022).Abram, N. K. et al. Synergies for improving oil palm production and forest conservation in floodplain landscapes. PLoS ONE 9, e95388 (2014).Article 

    Google Scholar 
    Othman, N. et al. Shift of paradigm needed towards improving human–elephant coexistence in monoculture landscapes in Sabah. Int. Zoo Yearb. 53, 161–173 (2019).Article 

    Google Scholar 
    Horton, A. J. et al. Can riparian forest buffers increase yields from oil palm plantations? Earths Future 6, 1082–1096 (2018).Article 

    Google Scholar 
    Ewers, R. M. et al. A large-scale forest fragmentation experiment: the Stability of Altered Forest Ecosystems Project. Phil. Trans. R. Soc. Lond. B 366, 3292–3302 (2011).Article 

    Google Scholar 
    Pfeifer, M. et al. Creation of forest edges has a global impact on forest vertebrates. Nature 551, 187–191 (2017).Ewers, R. M., Thorpe, S. & Didham, R. K. Synergistic interactions between edge and area effects in a heavily fragmented landscape. Ecology 88, 96–106 (2007).Article 

    Google Scholar 
    Deere, N. J. et al. High carbon stock forests provide co-benefits for tropical biodiversity. J. Appl. Ecol. 55, 997–1008 (2018).Article 
    CAS 

    Google Scholar 
    Hemprich-Bennett, D. R. et al. Altered structure of bat–prey interaction networks in logged tropical forests revealed by metabarcoding. Mol. Ecol. 30, 5844–5857 (2021).Article 

    Google Scholar 
    Williamson, J. et al. Riparian buffers act as microclimatic refugia in oil palm landscapes. J. Appl. Ecol. 58, 431–442 (2021).Article 

    Google Scholar 
    Slade, E. M., Mann, D. J. & Lewis, O. T. Biodiversity and ecosystem function of tropical forest dung beetles under contrasting logging regimes. Biol. Conserv. 144, 166–174 (2011).Article 

    Google Scholar 
    Gray, R. E. J. et al. Movement of forest-dependent dung beetles through riparian buffers in Bornean oil palm plantations. J. Appl. Ecol. 59, 238–250 (2022).Woodman, S. M. et al. esdm: a tool for creating and exploring ensembles of predictions from species distribution and abundance models. Methods Ecol. Evol. 10, 1923–1933 (2019).Article 

    Google Scholar 
    Liu, C., Berry, P. M., Dawson, T. P. & Pearson, R. G. Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28, 385–393 (2005).Article 

    Google Scholar 
    Piccini, I. et al. Greenhouse gas emissions from dung pats vary with dung beetle species and with assemblage composition. PloS ONE 12, e0178077 (2017).Article 

    Google Scholar 
    Raine, E. H. & Slade, E. M. Dung beetle–mammal associations: methods, research trends and future directions. Proc. R. Soc. B 286, 20182002 (2019).Article 

    Google Scholar 
    Nichols, E., Gardner, T., Peres, C., Spector, S. & Network, S. R. Co‐declining mammals and dung beetles: an impending ecological cascade. Oikos 118, 481–487 (2009).Article 

    Google Scholar 
    Asner, G. P. et al. Mapped aboveground carbon stocks to advance forest conservation and recovery in Malaysian Borneo. Biol. Conserv. 217, 289–310 (2018).Article 

    Google Scholar 
    Jucker, T. et al. Estimating aboveground carbon density and its uncertainty in Borneo’s structurally complex tropical forests using airborne laser scanning. Biogeosciences 15, 3811–3830 (2018).Article 

    Google Scholar 
    Philipson, C. D. et al. Active restoration accelerates the carbon recovery of human-modified tropical forests. Science 369, 838–841 (2020).Article 
    CAS 

    Google Scholar 
    Nunes, M. H. et al. Recovery of logged forest fragments in a human-modified tropical landscape during the 2015–16 El Niño. Nat. Commun. 12, 1526 (2021).Article 
    CAS 

    Google Scholar 
    Woittiez, L. S., van Wijk, M. T., Slingerland, M., van Noordwijk, M. & Giller, K. E. Yield gaps in oil palm: a quantitative review of contributing factors. Eur. J. Agron. 83, 57–77 (2017).Article 

    Google Scholar  More

  • in

    This baby turtle surprised scientists by swimming against the current

    In 2008, I had just begun volunteering at Equilibrio Azul — a non-profit marine-research and -conservation organization based in Quito — when colleagues discovered a hawksbill sea turtle (Eretmochelys imbricata) nesting at La Playita beach in Ecuador. The eastern Pacific population of hawksbill sea turtles is one of the most endangered in the world and was considered functionally extinct in the region before this turtle and others were observed.That discovery was a tipping point for hawksbill research in Ecuador and throughout the Pacific Ocean. Since 2008, we’ve found about 20 nests each year at La Playita, and one season, we documented 50.We have tagged 11 adult females with satellite transmitters. Previously, most of our understanding of these turtles had been based on observations in the Caribbean, where the reptiles are strictly coral-reef dwellers. But Ecuador’s reefs are mostly rocky, with patches of coral, and we were surprised to see females migrate south to mangroves, mainly for food.
    Women in science
    In this image, we have just attached a transmitter to a baby turtle — a first for hawksbill turtles this young and in the eastern Pacific region. We did not know much about hawksbills at this young age. It is tricky working with baby turtles, because they grow very fast, and the transmitters, which give us location data, can easily fall off. We’ve used cement to glue the devices to the shells of six newborns so far. The longest the transmitters have lasted is three months and the shortest period was only six weeks — but the devices provided our first insights into the ‘lost years’ of sea-turtle biology.Our findings have overturned assumptions that neonates were just carried along by currents. Instead, we found that one-day-old turtles can swim against the current. They aim for a specific direction — north by northwest — as they learn to dive and swim. We tracked one-year-old hawksbills to Costa Rican waters, a journey of roughly 2,000 kilometres, before we lost their signal.Cristina Miranda is a scientific coordinator at Equilibrio Azul in Quito, Ecuador. Interview by Virginia Gewin. More

  • in

    Rapid diversification underlying the global dominance of a cosmopolitan phytoplankton

    Genetic and morphological delineation between G. huxleyi strainsWe first assessed genetic variability through analysis of genomic polymorphism to determine whether distinct genetic lineages exist in G. huxleyi and to test whether these relate to morphotypes. We used 2,086,643 high-quality biallelic single nucleotide polymorphisms (SNPs) retrieved from the 47 clonal culture strains with the best genome sequence coverage ( >20×). A principal component analysis (PCA) and a discriminant analysis in principal component (DAPC) both delineate three well-defined genetic groups, with the distribution of strains being unequal and with no overlap on the principal components (Fig. 1a; Supplementary Fig. S3a,b). With regards to population structure, the DAPC analysis suggested that 3 clusters (K = 3) can be used to depict a genotype membership matrix for each strain (Fig. 1b; Supplementary Fig. S4). As such, it confirmed the three-lineage delineation proposed by the PCA, while illustrating no admixture between lineages.Fig. 1: Relationship between genetic structure and morphotypes in G. huxleyi.a Principal component analysis (PCA) based on 2,086,643 SNPs recovered from 47 G. huxleyi genomes; b Relationship between coalescent species phylogeny (ASTRAL tree based on 1000 supergenes) and DAPC clustering; c Correspondence between morphotypes and lineages within G. huxleyi, and sub-lineages within A1 (scale bar = 4 μm). Variable elements in relation to genotypes are highlighted in the schematics under the SEM pictures; d Distribution of coccolith length for 5 randomly chosen strains representing each clade and sub-clade, with a jittered box-plot on the left and a half-violin plot on the right for each group; e Matrix plot of Bonferroni corrected p-value corresponding to the Dunn-test for the comparison of coccolith length measurements between groups.Full size imagePhylogenetic inference based on alignments with higher mapping coverage only (47 strains) or including sequences with lower mapping coverage (59 strains) all supported segregation of strains into three main lineages, which we term clades A1, A2 and B, with A1 and A2 being more closely related to each other than to B (Fig. 1b; Supplementary Fig. S5a, b). This delineation is congruent with previous studies on the phylogeny of the Gephyrocapsa genus [17, 46, 65]. These clades also correspond to differences in morphotypes (Fig. 1b, c). All strains in clade A1 produce unambiguous A-group coccolith morphotypes (type A and type R). Similarly, all strains in clade B produce unambiguous B-group coccolith morphotypes (type B and type O). Clade A2 is less distinctive, with strains producing lightly calcified type A coccoliths. Some of these strains could be classified as type B/C [66] or C (both regarded as B-group morphotypes), but distinctive by the lower elevation of distal shield elements and by greater degree of calcification of the central area grid (which is reduced and sometimes absent in morphotypes B/C and C). At a finer level, clade A1 is composed of four sub-clades, which we term A1a, A1b, A1c, and A1d. Strains in sub-clades A1a, A1c and A1d all produce coccoliths with type A morphologies and distinctive degrees of calcification: strains in the sub-clade A1a form relatively lightly calcified coccoliths with regular elements, while strains in sub-clades A1c and A1d produce similar moderately calcified coccoliths, sometimes with conspicuous irregularities (inner tube elements overlapping into the central area). Strains in clade A1b produce distinct coccoliths exhibiting A-group morphology but with heavy calcification, including forms with heavily calcified shields which have been termed type R and also forms with heavily calcified central areas which have been referred to as “type A overcalcified”. Some clade A2 strains produce coccoliths with a similar morphology to strains in A1a, indicative of partially cryptic lineages (Supplementary Fig. S2; Supplementary Table S4).The congruence between morphotypes and clades is also supported by significant differences in the length of coccoliths measured between some of the clades (Fig. 1d, e). The morphogroups A and B differ significantly, and insignificant comparison relates to the comparison of sub-clades against the clade A2, which reinforces the closest relationship between A1 and A2. We denote also that the case of A1a and A2 demonstrating no significant difference in coccolith length concurs with the cryptic delineation mentioned above.Based on the clustering analyses and the phylogenetic reconstructions, we tested whether different groupings are distinct species with regards to the null hypothesis “G. huxleyi is a single species”, which correspond to the current state of taxonomy. Species delimitation based on comparison of Marginal Likelihood Estimators (MLE) with Bayes Factors (BF) supported the hypothesis that the three lineages depicted by ordination and phylogenetic reconstructions are distinct species as the best model (Table 1).Table 1 Species delimitation based on Bayes Factor Delimitation (BDF).Full size tableD-statistics calculated to estimate gene flow reveal a non-significant excess of alleles shared between the three lineages (Fig. 2a; Supplementary Table S5). Fbranch statistics, (fb) revealed significant signatures of gene-flow between sub-lineages within A1 associated with correlated estimates in relation to A1a, A2 and B (Fig. 2a) [60]. Signatures on the basal branch of diversification in A1 may correspond to genetic exchanges between A1 and B, with gene-flow signatures attributed to A2 corresponding to correlated estimates due to common ancestry. Recent signatures of gene-flow throughout the evolution of A1 are thus likely associated to the common ancestry between A1a, A2 and B during gene-flow events between the sub-lineages, as supported by the non-significant D statistics between the three lineages. Moreover, the phylogenetic network revealed similar convolutions between A1 sub-lineages but clear separation of the main lineages and longer branches in the A2 lineage (Fig. 2b).Fig. 2: Excess of allele sharing and differentiation in G. huxleyi.a f-branch (fb) statistics between lineages and sub-lineages. The gradient represents the fb score, grey blocks represents tests not consistent with the species tree (for each branch on the topology of the y axis, having itself or a sister taxon as donor on the topology of the x axis); asterisks denote block jack-knifing significance at p  More

  • in

    Publisher Correction: Future temperature extremes threaten land vertebrates

    Authors and AffiliationsJacob Blaustein Center for Scientific Cooperation, The Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Midreshet Ben-Gurion, IsraelGopal MuraliMitrani Department of Desert Ecology, The Swiss Institute for Dryland Environments and Energy Research, The Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Midreshet Ben-Gurion, IsraelGopal Murali & Uri RollDepartment F.-A. Forel for Aquatic and Environmental Sciences, Faculty of Science, University of Geneva, Geneva, SwitzerlandTakuya IwamuraDepartment of Forest Ecosystems and Society, College of Forestry, Oregon State University, Corvallis, OR, USATakuya IwamuraSchool of Zoology, Tel Aviv University, Tel Aviv, IsraelShai MeiriThe Steinhardt Museum of Natural History, Tel Aviv University, Tel Aviv, IsraelShai MeiriAuthorsGopal MuraliTakuya IwamuraShai MeiriUri RollCorresponding authorCorrespondence to
    Gopal Murali. More

  • in

    Ocean warming and acidification affect the transitional C:N:P ratio and macromolecular accumulation in the harmful raphidophyte Heterosigma akashiwo

    Pachauri, R. K. et al. Climate change 2014: synthesis report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2014).Fu, F. X., Warner, M. E., Zhang, Y., Feng, Y. & Hutchins, D. A. Effects of Increased temperature and CO2 on photosynthesis, growth, and elemental ratios in marine Synechococcus and Prochlorococcus (cyanobacteria) 1. J. Phycol. 43, 485–496 (2007).
    Google Scholar 
    Schippers, P., Lürling, M. & Scheffer, M. Increase of atmospheric CO2 promotes phytoplankton productivity. Ecol. Lett. 7, 446–451 (2004).
    Google Scholar 
    Raven, J. A., Gobler, C. J. & Hansen, P. J. Dynamic CO2 and pH levels in coastal, estuarine, and inland waters: Theoretical and observed effects on harmful algal blooms. Harmful Algae 91, 101594 (2020).CAS 

    Google Scholar 
    Gobler, C. J. et al. Ocean warming since 1982 has expanded the niche of toxic algal blooms in the North Atlantic and North Pacific oceans. Proc. Natl Acad. Sci. 114, 4975–4980 (2017).CAS 

    Google Scholar 
    Rost, B., Richter, K. U., Riebesell, U. & Hansen, P. J. Inorganic carbon acquisition in red tide dinoflagellates. Plant, Cell Environ. I 29, 810–822 (2006).CAS 

    Google Scholar 
    Honjo, T. Harmful Red Tides of Heterosigma akashiwo. NOAA Technical Report NMFS. 111, 27–32 (1992).Rensel, J. J. & Haigh, N. Fraser river sockeye salmon marine survival decline and harmful blooms of Heterosigma akashiwo. Harmful Algae 10, 98–115 (2010).
    Google Scholar 
    Herndon, J. & Cochlan, W. P. Nitrogen utilization by the raphidophyte Heterosigma akashiwo: growth and uptake kinetics in laboratory cultures. Harmful Algae 6, 260–270 (2007).
    Google Scholar 
    Haley, S. T., Alexander, H., Juhl, A. R. & Dyhrman, S. T. Transcriptional response of the harmful raphidophyte Heterosigma akashiwo to nitrate and phosphate stress. Harmful Algae 68, 258–270 (2017).CAS 

    Google Scholar 
    Wang, Z.-h, Liang, Y. & Kang, W. Utilization of dissolved organic phosphorus by different groups of phytoplankton taxa. Harmful Algae 12, 113–118 (2011).CAS 

    Google Scholar 
    Ji, N. et al. Metatranscriptome analysis reveals environmental and diel regulation of a Heterosigma akashiwo (raphidophyceae) bloom. Environ. Microbiol. 20, 1078–1094 (2018).CAS 

    Google Scholar 
    Zhang, H. et al. Functional differences in the blooming phytoplankton Heterosigma akashiwo and Prorocentrum donghaiense revealed by comparative metaproteomics. Appl. Environ. Microbiol. 85, e01425–01419 (2019).CAS 

    Google Scholar 
    Redfield, A. C. The biological control of chemical factors in the environment. Am. Scientist 46, 230A–221 (1958).
    Google Scholar 
    Liefer, J. D. et al. The macromolecular basis of phytoplankton C: N: P under nitrogen starvation. Front. Microbiol. 10, 763 (2019).
    Google Scholar 
    Matsumoto, K., Tanioka, T. & Rickaby, R. Linkages between dynamic phytoplankton C: N: P and the ocean carbon cycle under climate change. Oceanography 33, 44–52 (2020).
    Google Scholar 
    Thrane, J. E., Hessen, D. O. & Andersen, T. Plasticity in algal stoichiometry: Experimental evidence of a temperature‐induced shift in optimal supply N: P ratio. Limnol. Oceanogr. 62, 1346–1354 (2017).CAS 

    Google Scholar 
    Toseland, A. et al. The impact of temperature on marine phytoplankton resource allocation and metabolism. Nat. Clim. Change 3, 979–984 (2013).CAS 

    Google Scholar 
    Mittler, R., Finka, A. & Goloubinoff, P. How do plants feel the heat? Trends Biochem. Sci. 37, 118–125 (2012).CAS 

    Google Scholar 
    Dingman, J. E. & Lawrence, J. E. Heat-stress-induced programmed cell death in Heterosigma akashiwo (Raphidophyceae). Harmful Algae 16, 108–116 (2012).
    Google Scholar 
    Whitten, S. T., García-Moreno E, B. & Hilser, V. J. Local conformational fluctuations can modulate the coupling between proton binding and global structural transitions in proteins. Proc. Natl Acad. Sci. 102, 4282–4287 (2005).CAS 

    Google Scholar 
    Casey, J. R., Grinstein, S. & Orlowski, J. Sensors and regulators of intracellular pH. Nat. Rev. Mol. Cell Biol. 11, 50–61 (2010).CAS 

    Google Scholar 
    Kim, H., Spivack, A. J. & Menden-Deuer, S. pH alters the swimming behaviors of the raphidophyte Heterosigma akashiwo: implications for bloom formation in an acidified ocean. Harmful Algae 26, 1–11 (2013).CAS 

    Google Scholar 
    Hennon, G. M., Williamson, O. M., Limón, M. D. H., Haley, S. T. & Dyhrman, S. T. Non-linear physiology and gene expression responses of harmful alga Heterosigma akashiwo to rising CO2. Protist 170, 38–51 (2019).CAS 

    Google Scholar 
    Xu, H., Jaynes, J. & Ding, X. Combining two-level and three-level orthogonal arrays for factor screening and response surface exploration. Statistica Sin. 24, 269–289 (2014).
    Google Scholar 
    Boyd, P. W. & Ellwood, M. J. The biogeochemical cycle of iron in the ocean. Nat. Geosci. 3, 675–682 (2010).CAS 

    Google Scholar 
    Sterner, R. W. & Elser, J. J. in Ecological Stoichiometry (Princeton university press, 2002).Liu, H. C., Liao, H. T. & Charng, Y. Y. The role of class A1 heat shock factors (HSFA1s) in response to heat and other stresses in Arabidopsis. Plant Cell Environ. 34, 738–751 (2011).CAS 

    Google Scholar 
    Geider, R. J. & La Roche, J. J. Redfield revisited: variability of C [ratio] N [ratio] P in marine microalgae and its biochemical basis. Eur. J. Phycol. 37, 1–17 (2002).
    Google Scholar 
    Loladze, I. & Elser, J. J. The origins of the Redfield nitrogen‐to‐phosphorus ratio are in a homoeostatic protein‐to‐rRNA ratio. Ecol. Lett. 14, 244–250 (2011).
    Google Scholar 
    Hennige, S. J., Coyne, K. J., MacIntyre, H., Liefer, J. & Warner, M. E. The photobiology of Heterosigma akashiwo. Photoacclimation, diurnal periodicity, and its ability to rapidly exploit exposure to high light. J. Phycol. 49, 349–360 (2013).CAS 

    Google Scholar 
    Collier, J. L. & Grossman, A. A small polypeptide triggers complete degradation of light‐harvesting phycobiliproteins in nutrient‐deprived cyanobacteria. EMBO J. 13, 1039–1047 (1994).CAS 

    Google Scholar 
    Gordillo, F. J., Jimenez, C., Figueroa, F. L. & Niell, F. X. Influence of elevated CO2 and nitrogen supply on the carbon assimilation performance and cell composition of the unicellular alga Dunaliella viridis. Physiologia Plant. 119, 513–518 (2003).CAS 

    Google Scholar 
    Satoh, E., Watanabe, M. M. & Fujii, T. Photoperiodic regulation of cell division and chloroplast replication in Heterosigma akashiwo. Plant Cell Physiol. 28, 1093–1099 (1987).
    Google Scholar 
    Ashworth, J. et al. Genome-wide diel growth state transitions in the diatom Thalassiosira pseudonana. Proc. Natl Acad. Sci. 110, 7518–7523 (2013).CAS 

    Google Scholar 
    Thangaraj, S. & Sun, J. J. E. M. Transcriptomic reprogramming of the oceanic diatom Skeletonema dohrnii under warming ocean and acidification. Environ. Microbiol. 23, 980–995 (2021).CAS 

    Google Scholar 
    Nakajima, K., Tanaka, A. & Matsuda, Y. SLC4 family transporters in a marine diatom directly pump bicarbonate from seawater. Proc. Natl Acad. Sci. 110, 1767–1772 (2013).CAS 

    Google Scholar 
    Kranz, S. A. et al. Low temperature reduces the energetic requirement for the CO2 concentrating mechanism in diatoms. N. Phytologist 205, 192–201 (2015).CAS 

    Google Scholar 
    Ralston, A. & Shaw, K. Gene expression regulates cell differentiation. Nat. Educ. 1, 127–131 (2008).
    Google Scholar 
    Lobo, I. Environmental influences on gene expression. Nat. Educ. 1, 39 (2008).
    Google Scholar 
    Suzuki, N. et al. Respiratory burst oxidases: the engines of ROS signaling. Curr. Opin. Plant Biol. 14, 691–699 (2011).CAS 

    Google Scholar 
    Saidi, Y., Finka, A. & Goloubinoff, P. Heat perception and signalling in plants: a tortuous path to thermotolerance. N. Phytologist 190, 556–565 (2011).CAS 

    Google Scholar 
    Saidi, Y. et al. The heat shock response in moss plants is regulated by specific calcium-permeable channels in the plasma membrane. Plant Cell 21, 2829–2843 (2009).CAS 

    Google Scholar 
    Zhang, W. et al. Molecular and genetic evidence for the key role of AtCaM3 in heat-shock signal transduction in Arabidopsis. Plant Physiol. 149, 1773–1784 (2009).CAS 

    Google Scholar 
    Li, S. et al. Functional characterization of Arabidopsis thaliana WRKY39 in heat stress. Mol. Cells 29, 475–483 (2010).CAS 

    Google Scholar 
    Sangwan, V., Örvar, B. L., Beyerly, J., Hirt, H. & Dhindsa, R. S. Opposite changes in membrane fluidity mimic cold and heat stress activation of distinct plant MAP kinase pathways. Plant J. 31, 629–638 (2002).CAS 

    Google Scholar 
    Reddy, A. S., Ali, G. S., Celesnik, H. & Day, I. S. Coping with stresses: roles of calcium-and calcium/calmodulin-regulated gene expression. Plant Cell 23, 2010–2032 (2011).CAS 

    Google Scholar 
    Meiri, D. & Breiman, A. J. Arabidopsis ROF1 (FKBP62) modulates thermotolerance by interacting with HSP90. 1 and affecting the accumulation of HsfA2‐regulated sHSPs. Plant J. 59, 387–399 (2009).CAS 

    Google Scholar 
    Mishkind, M., Vermeer, J. E., Darwish, E. & Munnik, T. J. Heat stress activates phospholipase D and triggers PIP2 accumulation at the plasma membrane and nucleus. Plant J. 60, 10–21 (2009).CAS 

    Google Scholar 
    Zheng, S. Z. et al. Phosphoinositide‐specific phospholipase C9 is involved in the thermotolerance of Arabidopsis. Plant J. 69, 689–700 (2012).CAS 

    Google Scholar 
    Pincus, D. et al. BiP binding to the ER-stress sensor Ire1 tunes the homeostatic behavior of the unfolded protein response. PLoS Biol. 8, e1000415 (2010).
    Google Scholar 
    Sugio, A., Dreos, R., Aparicio, F. & Maule, A. J. The cytosolic protein response as a subcomponent of the wider heat shock response in Arabidopsis. Plant Cell 21, 642–654 (2009).CAS 

    Google Scholar 
    Vasseur, F., Pantin, F. & Vile, D. J. P. Cell & Environment. Changes in light intensity reveal a major role for carbon balance in Arabidopsis responses to high temperature. Plant Cell Environ. 34, 1563–1576 (2011).CAS 

    Google Scholar 
    Paroutis, P., Touret, N. & Grinstein, S. The pH of the secretory pathway: measurement, determinants, and regulation. J. Physiol. 19, 207–215 (2004).CAS 

    Google Scholar 
    Forgac, M. Vacuolar ATPases: rotary proton pumps in physiology and pathophysiology. Nat. Rev. Mol. Cell Biol. 8, 917–929 (2007).CAS 

    Google Scholar 
    Cipriano, D. J. et al. Structure and regulation of the vacuolar ATPases. Biochim. et. Biophys. Acta -Bioenerg. 1777, 599–604 (2008).CAS 

    Google Scholar 
    Abad, M. F. C., Di Benedetto, G., Magalhães, P. J., Filippin, L. & Pozzan, T. Mitochondrial pH monitored by a new engineered green fluorescent protein mutant. J. Biol. Chem. 279, 11521–11529 (2004).CAS 

    Google Scholar 
    McCORMACK, J. G., Halestrap, A. P. & Denton, R. M. Role of calcium ions in regulation of mammalian intramitochondrial metabolism. Physiol. Rev. 70, 391–425 (1990).CAS 

    Google Scholar 
    Garlid, K. D., Sun, X., Paucek, P. & Woldegiorgis, G. in Methods in enzymology Vol. 260 331–348 (Elsevier, 1995).Yamada, E. W. & Huzel, N. J. J. B. Calcium-binding ATPase inhibitor protein of bovine heart mitochondria. Role in ATP synthesis and effect of calcium. Biochemistry 28, 9714–9718 (1989).CAS 

    Google Scholar 
    Moreno-Sánchez, R. Inhibition of oxidative phosphorylation by a Ca2+-induced diminution of the adenine nucleotide translocator. Biochim. et. Biophys. Acta -Bioenerg. 724, 278–285 (1983).
    Google Scholar 
    Matsuyama, S., Llopis, J., Deveraux, Q. L., Tsien, R. Y. & Reed, J. Changes in intramitochondrial and cytosolic pH: early events that modulate caspase activation during apoptosis. Nat. Cell Biol. 2, 318–325 (2000).CAS 

    Google Scholar 
    Sunda, W. G., Price, N. M. & Morel, F. M. Trace metal ion buffers and their use in culture studies. Algal Cultur. Tech. 4, 35–63 (2005).
    Google Scholar 
    Sun, J. et al. Effects of changing pCO2 and phosphate availability on domoic acid production and physiology of the marine harmful bloom diatom Pseudo‐nitzschia multiseries. Limnol. Oceanogr. 56, 829–840 (2011).CAS 

    Google Scholar 
    Pierrot, D., Lewis, E. & Wallace, D. J. MS Excel Program Developed for CO2 System Calculations ORNL/CDIAC‐105 (US Dept. of Energy, Oak Ridge, TN, 2006).Wilbur, K. M. & Anderson, N. G. Electrometric and colorimetric determination of carbonic anhydrase. J. Biol. Chem. 176, 147–154 (1948).CAS 

    Google Scholar 
    Solórzano, L. & Sharp, J. H. Determination of total dissolved phosphorus and particulate phosphorus in natural waters 1. Limnol. Oceanogr. 25, 754–758 (1980).
    Google Scholar 
    Myklestad, S. M., Skånøy, E. & Hestmann, S. J. Sensitive and rapid method for analysis of dissolved mono-and polysaccharides in seawater. Mar. Chem. 56, 279–286 (1997).CAS 

    Google Scholar 
    Pakulski, J. D. & Benner, R. J. An improved method for the hydrolysis and MBTH analysis of dissolved and particulate carbohydrates in seawater. Mar. Chem. 40, 143–160 (1992).CAS 

    Google Scholar 
    Folch, J. & Lees, M. & Sloane Stanley, G. H. A simple method for the isolation and purification of total lipids from animal tissues. J. Biol. Chem. 226, 497–509 (1957).CAS 

    Google Scholar 
    Pande, S., Khan, R. P. & Venkitasubramanian, T. Microdetermination of lipids and serum total fatty acids. Anal. Biochem. 6, 415–423 (1963).CAS 

    Google Scholar 
    Lowry, O., Rosebrough, N., Farr, A. L. & Randall, R. J. Protein measurement with the Folin phenol reagent. J. Biol. Chem. 193, 265–275 (1951).CAS 

    Google Scholar 
    Berdalet, E., Roldán, C., Olivar, M. P. & Lysnes, K. Quantifying RNA and DNA in planktonic organisms with SYBR Green II and nucleases. Part A. Optimisation of the assay. Sci. Mar. 69, 1–16 (2005).CAS 

    Google Scholar 
    Chomoczynski, P. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloro-form extraction. Anal. Biochem. 162, 156–159 (1987).
    Google Scholar 
    Sañudo-Wilhelmy, S. A. et al. The impact of surface-adsorbed phosphorus on phytoplankton Redfield stoichiometry. Phycol. Res. 432, 897–901 (2004).
    Google Scholar 
    Dyhrman, S. T. Nutrients and their acquisition: phosphorus physiology in microalgae. Physiol. Microalgae 155–183 (2016).Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).CAS 

    Google Scholar 
    Waterhouse, R. M. et al. BUSCO applications from quality assessments to gene prediction and phylogenomics. Mol. Biol. Evol. 35, 543–548 (2018).CAS 

    Google Scholar 
    Pruitt, K. D., Tatusova, T. & Maglott, D. R. J. N. A. R. NCBI reference sequences (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins. Nucleic Acids Res. 35, D61–D65 (2007).CAS 

    Google Scholar 
    Kanehisa, M. & Goto, S. J. N. A. R. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27–30 (2000).CAS 

    Google Scholar 
    Mitchell, A. L. et al. InterPro in 2019: improving coverage, classification and access to protein sequence annotations. Nucleic Acids Res. 47, D351–D360 (2019).CAS 

    Google Scholar 
    Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 1–21 (2014).
    Google Scholar 
    Wang, L., Feng, Z., Wang, X., Wang, X. & Zhang, X. J. B. DEGseq: an R package for identifying differentially expressed genes from RNA-seq data. Bioinformatics 26, 136–138 (2010).
    Google Scholar  More

  • in

    Nocardiopsis changdeensis sp. nov., an endophytic actinomycete isolated from the roots of Eucommia ulmoides Oliv

    Rainey FA, WardRainey N, Kroppenstedt RM, Stackebrandt E. The genus Nocardiopsis represents a phylogenetically coherent taxon and a distinct actinomycete lineage: proposal of Nocardiopsaceae fam. nov. Int J Syst Evol Microbiol. 1996;46:1088–92.CAS 

    Google Scholar 
    Goodfellow M, Order XV Streptosporangiales ord. nov. In: Goodfellow M, Kämpfer P, Busse HJ, Trujillo ME, Suzuki K, Ludwig W, Whitman WB (eds), Bergey’s Manual of Systematic Bacteriology vol. 5, 2nd edn., Springer, New York, 2012, p. 1805.Meyer J. Nocardiopsis, a new genus of the order Actinomycetales. Int J Sys Bacteriol. 1976;26:487–93.Article 

    Google Scholar 
    Chen YG, Cui XL, Kroppenstedt RM, Stackebrandt E, Wen ML, et al. Nocardiopsis quinghaiensis sp. nov. isolated from saline soil in China. Int J Syst Evol Microbiol. 2008;58:699–705.Article 
    CAS 

    Google Scholar 
    Chen YG, Zhang YQ, Tang SK, Liu ZX, Xu LH, et al. Nocardiopsis terrae sp. nov., a halophilic actinomycete isolated from saline soil. Antonie van Leeuwenhoek. 2010;98:31–8.Article 

    Google Scholar 
    Pan HQ, Zhang DF, Li L, Jiang Z, Li WJ. Nocardiopsis oceani sp. nov. and nocardiopsis nanhaiensis sp. nov. actinomycetes isolated from marine sediment of the south china sea. Int J Syst Evol Microbiol. 2015;65:3384–91.Article 
    CAS 

    Google Scholar 
    Akhwale JK, Göker M, Rohde M, Schumann P, Boga HI, et al. Nocardiopsis mwathae sp. nov., isolated from the haloalkaline Lake Elmenteita in the African Rift Valley. Antonie van Leeuwenhoek. 2016;109:421–30.Article 
    CAS 

    Google Scholar 
    Schippers A. Nocardiopsis metallicus sp. nov. a metal-leaching actinomycete isolated from an alkaline slag dump. Int J Syst Evol Microbiol. 2002;52:2291–5.CAS 

    Google Scholar 
    Devi AM, Nimaichand S, Hamidah I, Xiao-Tong Z, Bull AT, et al. Nocardiopsis deserti sp. nov. isolated from a high altitude atacama desert soil. Int J Syst Evol Microbiol. 2020;70:3210–8.Article 

    Google Scholar 
    Hamedi J, Mohammadipanah F, Von JM, Potter G, Schumann P, et al. Nocardiopsis sinuspersici sp. nov. isolated from sandy rhizospheric soil. Int J Syst Evol Microbiol. 2010;60:2346–52.Article 
    CAS 

    Google Scholar 
    Zhang YG, Lu XH, Ding YB, Zhou XK, Wan HF, et al. Nocardiopsis rhizosphaerae sp. nov., isolated from rhizosphere soil of Halocnermum strobilaceum (Pall.) Bieb. Int J Syst Evol Microbiol. 2016;66:5129–33.Article 
    CAS 

    Google Scholar 
    Muangham S, Suksaard P, Mingma R, Matsumoto A, Takahashi Y, et al. Nocardiopsis sediminis sp. nov., isolated from mangrove sediment Free. Int J Syst Evol Microbiol. 2016;66:3835–40.Article 
    CAS 

    Google Scholar 
    Qin S, Li J, Chen HH, Zhao GZ, Zhu WY, et al. Isolation, diversity, and antimicrobial activity of rare actinobacteria from medicinal plants of tropical rain forests in Xishuangbanna, China. Appl Environ Microbiol. 2009;75:6176–86.Article 
    CAS 

    Google Scholar 
    Sindhuphak W, Macdonald E. Head actinomycetoma caused by Nocardiopsis dassonvillei. Arch. Dermatol. 1985;121:1332–4.Article 
    CAS 

    Google Scholar 
    Mordarska H, Zakrzewska-Czerwiñska J, Paściak M, Szponar B, Rowiñski S. Rare, suppurative pulmonary infection caused by Nocardiopsis dassonvillei recognized by glycolipid markers. FEMS Immunol Med Microbiol. 1998;21:47–55.Article 
    CAS 

    Google Scholar 
    Bennur T, Kumar AR, Zinjarde SS, Javdekar V. Nocardiopsis species: a potential source of bioactive compounds. J Appl Microbiol. 2016;120:1–16.Article 
    CAS 

    Google Scholar 
    Mo P, Yu YZ, Zhao JR, Gao J. Streptomyces xiangtanensis sp. nov., isolated from a manganese-contaminated soil. Antonie van Leeuwenhoek. 2017;110:297–304.Article 
    CAS 

    Google Scholar 
    Atlas RM In: Parks LC (ed) Handbook of microbiological media. CRC Press, Boca Raton, 1993;pp: 666–72.Shirling EB, Gottlieb D. Methods for characterization of Streptomyces species. Int J Syst Bacteriol. 1966;16:313–40.Article 

    Google Scholar 
    Ridgway R Color standards and color nomenclature. Ridgway, Washington, DC, 1912;pp: 1–43.Ruan JS, Huang Y Rapid identification and systematics of Actinobacteria. Science Press, Beijing, China, 2011;pp: 72–7.Xu LH, Li WJ, Liu ZH, Jiang CL Actinomycetes systematics: principles, methods and practices. Science Press, Beijing, China. 2007;pp: 40–53.MIDI. Sherlock Microbial Identification System Operating Manual, Version 6.0. Newark DE: MIDI Inc. 2005;pp: 1–7.Hasegawa T, Takizawa M, Tanida S. A rapid analysis for chemical grouping of aerobic actinomycetes. J Gen Appl Microbiol. 1983;29:319–22.Article 
    CAS 

    Google Scholar 
    Lechevalier MP, Lechevalier H. Chemical composition as a criterion in the classification of aerobic actinomycetes. Int J Syst Bacteriol. 1970;20:435–43.Article 
    CAS 

    Google Scholar 
    Collins MD, Pirouz T, Goodfellow M, Minnikin DE. Distribution of menaquinones in actinomycetes and corynebacteria. J Gen Microbiol. 1977;100:221–30.Article 
    CAS 

    Google Scholar 
    Kroppenstedt RM Fatty acid and menaquinone analysis of actinomycetes and related organisms. In: Goodfellow M, Minnikin DE (eds) Chemical methods in bacterial systematics. Academic Press, London, England, pp, 1985: 173–99.Kates M Techniques of Lipidology, 2nd ed. Amsterdam: Elsevier, 1986.Komagata K, Suzuki KI. 4 lipid and cell-wall analysis in bacterial systematics. Method Microbiol. 1988;19:161–207.Article 

    Google Scholar 
    Lane, DJ 16S/23S rRNA sequencing. In: nucleic acid techniques in bacterial systematics. Stackebrandt E, Goodfellow M, eds., John Wiley and Sons, New York, NY, pp, 1991: 115–75.Yoon SH, Ha SM, Kwon S, Lim J, Kim Y, et al. Introducing EzBioCloud: a taxonomically united database of 16S rRNA gene sequences and whole-genome assemblies. Int J Syst Evol Microbiol. 2017;67:1613–7.Article 
    CAS 

    Google Scholar 
    Saitou N, Nei M. The Neighbor-joining Method: A New Method for Reconstructing Phylogenetic Trees. Mol Biol Evol. 1987;4:406–25.CAS 

    Google Scholar 
    Felsenstein J. Evolutionary trees from DNA sequences: a maximum likelihood approach. J Mol Evol. 1981;17:368–76.Article 
    CAS 

    Google Scholar 
    Fitch WM. Toward defining the course of evolution: minimum change for a specific tree topology. Syst Biol. 1971;20:406–16.Article 

    Google Scholar 
    Kumar S, Stecher G, Tamura K. MEGA 7: molecular evolutionary genetics analysis version 7.0 for bigger datasets. Mol Biol Evol. 2016;33:1870–4.Article 
    CAS 

    Google Scholar 
    Felsenstein J. Confidence limits on phylogenies: an approach using the bootstrap. Evolution. 1985;39:783–91.Article 

    Google Scholar 
    Brettin T, Davis JJ, Disz T, Edwards RA, Gerdes S, et al. RASTtk: a modular and extensible implementation of the RAST algorithm for building custom annotation pipelines and annotating batches of genomes. Sci Rep. 2015;5:8365.Article 

    Google Scholar 
    Overbeek R, Olson R, Pusch GD, Olsen GJ, Stevens R, et al. The SEED and the Rapid Annotation of microbial genomes using Subsystems Technology (RAST). Nucleic Acids Res. 2014;42:D206–214.Article 
    CAS 

    Google Scholar 
    Meier-Kolthoff JP, Göker M. TYGS is an automated high-throughput platform for state-of-the-art genome-based taxonomy. Nat. Commun. 2019;10:2182.Article 

    Google Scholar 
    Richter M, Rosselló-Móra R, Ckner FOG, Peplies J. JSpeciesWS: a web server for prokaryotic species circumscription based on pairwise genome comparison. Bioinformatics. 2015;32:929–31.Article 

    Google Scholar 
    Meier-Kolthoff JP, Auch AF, Klenk HP, Göker M. Genome sequence-based species delimitation with confidence intervals and improved distance functions. BMC Bioinform. 2013;14:1–14.Article 

    Google Scholar 
    Rodriguez RL, Gunturu S, Harvey WT, Rossello-Mora R, Tiedje JM, et al. The Microbial Genomes Atlas (MiGA) webserver: taxonomicand gene diversity analysis of Archaea and Bacteria at the whole genome level. Nucleic Acids Res. 2018;46:W282–W288.Article 

    Google Scholar 
    Wayne LG, Brenner DJ, Colwell RR, Grimont PAD, Kandler O. International committee on systematic bacteriology. report of the ad hoc committee on the reconciliation of approaches to bacterial systematics. Int J Syst Bacteriol. 1987;37:463–4.Article 

    Google Scholar 
    Richter M, Rossello-Mora R. Shifting the genomic gold standard for the prokaryotic species definition. Proc Nat Acad Sci USA. 2009;106:19126–31.Article 
    CAS 

    Google Scholar 
    Vincent L, Richard D, Olivier G. FastME 2.0: a comprehensive, accurate, and fast distance-based phylogeny inference program. Mol Biol Evol. 2015;32:2798–800.Article 

    Google Scholar 
    Farris JS. Estimating phylogenetictrees from distance matrices. Am Nat. 1972;106:645–68.Article 

    Google Scholar 
    Fang CY, Zhang JL, Pang HC, Li YY, Xin YH, et al. Nocardiopsis flavescens sp. nov., an actinomycete isolated from marine sediment. Int J Syst Evol Microbiol. 2011;61:2640–5.Article 
    CAS 

    Google Scholar  More

  • in

    Joint use of location and acceleration data reveals influences on transitions among habitats in wintering birds

    Goose capture and trackingWe used rocket netting and leg snares to capture white-fronted geese in three regions in Texas (Rolling Plains, Lower Texas Coast, and South Texas Brushlands) and one region in Louisiana (Chenier Plain) from October to February 2016–2018 (Fig. 1). We determined age and sex of individuals by cloacal inversion, rectrices and other plumage characteristics27,28. We fit a solar powered GPS/ACC/Global System for Mobile communication (GSM) neckband tracking device (Cellular Tracking Technologies Versions BT3.0, BT3.5 and BT3.75; 44–54 g; Rio Grande, New Jersey, USA, and Ornitela OrniTrack-N38; 36 g; Vilnius, Lithuania), and an aluminum U.S. Geological Survey Bird Banding Laboratory metal leg band (Supplementary Fig. S1) on each bird. Geese were captured and tagged under USGS Bird Banding Permits #21314 and #23792, and Texas A&M University-Kingsville Institutional Animal Care and Use Committee #2015-09-01B. Captive geese were permitted under TAMUK IACUC #2018-01-11 and United States Fish and Wildlife Service Waterfowl Sale and Disposal permit #MB03808D-0. All applicable field methods were carried out in accordance with relevant guidelines and regulations. All animal handling protocols were approved by TAMUK IACUC committees and the USGS Bird Banding Laboratory. When multiple white-fronted geese were captured simultaneously, devices were only placed on adult females or adult males to eliminate the potential of placing devices on mated pairs, thus biasing independent data collection due to monogamous, long-term pair bonds in white-fronted geese. Location duty cycles were set to collect a GPS location every 30 min (i.e., 48/day) and location accuracy was 7.2 and 6.5 m for CTT and Ornitela devices, respectively. Data were uploaded once daily to respective online user interface websites when within areas of GSM coverage. When not in coverage areas, data were stored onboard the device until birds returned to coverage areas. All devices were equipped with a tri-axial ACC sensor which measured G-force (g; CTT devices) or millivolts (mV; Ornitela devices) at a fixed sampling scheme; CTT BT3.5 and Ornitela devices collected ACC data for a duration of 3 s every 6 min at 10 Hz, while BT3.0 devices collected data for a duration of 10 s every 6 min at 10 Hz. Generation BT3.0 devices were subsampled to match the sampling scheme of 3 s bursts before analyses. Ornitela units measured in mV were converted to G-force. We applied manufacturer- and tag-specific ACC calibration to all units, respectively, by collecting ACC data on each possible rotation for all axes when the device was stationary and applying the calibration to the raw ACC values (see Ref.29 for full calibration procedure). All devices recorded temperature in °C at each GPS fix. We censored GPS and ACC data from the time of release until individuals appeared to resume normal movement activity (i.e., roosting and foraging), as geese typically flew to the nearest wetland immediately after release where they remained without leaving while acclimating to wearing devices, which ranged from 1 to 7 days30. We defined the start of the winter period following a southward migratory movement from staging areas in Canada, without additional migratory movements southward below 40° 0′ 00″ N, or from the time of device deployment (minus device acclimation period) until geese made large northward migratory movements, or 28 February if geese remained in wintering areas.Figure 1Primary wintering regions of the Midcontinent population of greater white-fronted goose (Anser albifrons frontalis) in North America (excluding regions in Mexico). Transmitters were deployed during winters 2016–2018 in the Chenier Plain (Louisiana), Lower Texas Coast, and Rolling/High Plains regions. Geese that made winter movements outside of these defined regions were classified as ‘Other’ regions. Map created using Esri ArcMap (version 10.3.1; www.esri.com).Full size imageLand cover covariatesWe used publicly available spatial landcover data sets (30-m resolution) in combination with remote sensing to create landscape layers using programs Esri ArcMap (version 10.3.1), Erdas Imagine, and Program R (version 3.5.231). We used 2017 and 2018 National Agricultural Statistics Service Cropland Data Layer (CDL) data sets for agricultural crop types and freshwater wetlands, and the 2010 Coastal Change Analysis Program layer for saltwater and coastal wetland classifications29,32. Additionally, we used multi-spectral Landsat 8 Operational Land Imager satellite imagery, with principal component analysis on eight Landsat bands and a normalized difference vegetation index band, and unsupervised classification33,34 to accurately identify and create a spatial layer for peanut fields. We developed this layer for two regions with annual peanut agriculture (i.e., the Rolling/High Plains and South Texas Brushlands) using ground-truthed peanut fields, because the CDL layer did not identify this crop accurately based on our field observations during captures. We achieved  > 90% accuracy of peanut identification for each image independently based on annual ground-truthed observations of peanut fields. Finally, we grouped like-habitat categories to reduce the total number of categories to eight: corn, grass/winter wheat, herbaceous wetlands, other grains (i.e., soybeans, sorghum, and peanuts), rice, woody wetlands, open water/unconsolidated shore and other (Supplementary Table S1). White-fronted geese used several ecologically distinct regions in both winters of our study (Fig. 1), where the landscape composition of specific landcover types varied. To account for regional variability, we added region ID as a categorical variable to all GPS locations. Regions included the MAV, Chenier Plain, Texas Mid-coast, Lower Texas Coast, South Texas Brushlands, Texas Rolling/High Plains, and Other (i.e., locations outside of these identified wintering regions; Fig. 1). We used regional shapefiles of Gulf Coast Joint Venture Initiative Areas (Laguna Madre [Lower Texas Coast], Texas Mid-coast, and Chenier Plain35), and Level III Ecoregions (Mississippi Alluvial Valley, Texas Rolling/High Plains, and South Texas Brushlands36) as boundaries to classify data into regions. Due to insufficient and incompatible spatial layers for Mexico, we limited analyses to locations within the US.Location and acceleration data collectionRemotely determining behaviors of individuals using ACC data is most accurately addressed by developing a training dataset of known behaviors linked with ACC measurements of those behaviors18,37. To develop a training dataset, we collected video footage of two domestic white-fronted geese in Texas, US, and 18 tagged wild Greenland white-fronted geese (A. a. flavirostris) fitted with the same device types and the same data collection scheme, in Wexford, Ireland and Hvanneyri, Iceland during winters 2017–2018. We supplemented wild recordings with behavioral recordings of captive white-fronted geese as a proxy for wild individuals due to difficulty filming wild geese in inclement weather and obstructed video footage, which is common in ACC literature19,20,38,39. To replicate devices placed on wild white-fronted geese and account for potential variation in ACC measurements between device brands, among device versions and individual geese, we deployed three versions of devices used in this study on captive white-fronted geese during filming sessions38,40. We attached tracking devices to captive geese one week prior to video collection to allow geese to adjust to wearing devices. We collected ACC measurements for 3 s bursts, at 1 min intervals, at 10 Hz. We constructed a 149 m2 enclosure in an agricultural field to imitate an environment that wild geese may encounter. We created two enclosure settings allowing captive geese to forage on sprouted winter wheat (~ 2–15 cm) or on a randomly dispersed mixture of grain seeds (corn, wheat, sorghum) to account for both ‘grazing’ of vascular vegetation and ‘gleaning’ of agricultural grains to imitate foraging in wild geese. We used Sony Handycam DCR-SR45 video cameras, matched internal camera clocks with a running Universal Coordinated Time clock, and verbally re-calibrated the current time every 2 min during video footage collection. We filmed 119.5 h of video footage, and matched behavior with recorded ACC measurements by pairing video and device timestamps for each device using JWatcher41 and Program R.We characterized goose behaviors into four categories: ‘stationary’, ‘walk’, and ‘foraging’ from ground-truthed video footage, and ‘flight’ from visual inspection of the ACC data and consecutive GPS tracks during migration where device-measured speed remained  > 4.63 km/h (based on a natural break in the speed density distribution of all GPS locations). Each ACC burst was classified as only one behavior (i.e., a goose that was walking as it foraged was classified as ‘foraging’). We combined wild goose behaviors and captive goose behaviors after identifying minimal differences in ACC burst summary statistics29 for ‘stationary’ and ‘walk’ behaviors. We used ‘graze’ behaviors only from wild geese because of low sample size for captive geese and slight differences in ACC summary statistics between captive and wild geese for this behavior. ‘Glean’ foraging behavior was only classified from captive geese. We then combined ‘graze’ and ‘glean’ behaviors into an overall ‘foraging’ behavior to account for variation in foraging behavior of wild geese, and because machine learning models could not accurately distinguish between the two foraging modes40. We randomly subsampled all behaviors to 150 bursts if our dataset contained at least that many bursts to reduce the risk of artificially increasing prediction accuracy20. We determined there were insufficient differences in ACC signatures between CTT BT3.0 and BT3.5 versions by visual comparison of signatures and summary statistics, and merged all bursts into an overall CTT-specific training data set, and retained CTT- and Ornitela-specific training data sets to account for brand-specific ACC measurement schemes. The final training data sets consisted of 150 stationary, 150 walking, 118 foraging, and 150 flying bursts (CTT), and 150 stationary, 75 walking, 120 foraging, and 150 flying bursts (Ornitela).We used the training data sets to predict behaviors of tagged, wild white-fronted geese during winter with temporally aligned GPS and ACC data. We used a suite of supervised machine-learning algorithms and selected the algorithm with greatest prediction accuracy based on an 80% training, 20% testing sample approach. We tested random forest, support vector machines, K-nearest neighbors, classification and regression trees, and linear discriminant analysis, all with cross validation in Program R18,29,42. We evaluated models using three metrics defined in Ref.42: (1) overall classification accuracy as the percent of classifications in the test data set that were predicted correctly, (2) precision of assignment, the probability that an assigned behavior in the test data set was correct, and (3) model recall, the probability that a sample with a specific behavior in the test data set was correctly classified as that behavior by the model. Random forests provided the highest overall classification accuracy (95.6% for CTT and 96.0% for Ornitela), as well as high precision and recall for each behavior (CTT range 93.1–99.3, Ornitela range 88.9–100.0%), and therefore we labeled behaviors from wild goose ACC data using the random forests.Habitat transition modelOur habitat-transition model required temporally matched GPS and ACC datasets. Therefore, we subset all GPS locations to match the time-series of ACC data per individual because devices typically acquired GPS data longer than ACC data before device failure or individual mortality. For each GPS location, we extracted the landcover type and wintering region from spatial layers and retained temperature recorded from the device. To link classified ACC behaviors to GPS locations, we matched ACC timestamps between two GPS locations with the previous GPS timestamp. That is, all ACC bursts between two GPS locations were assigned backward to the previous GPS location. In this way, an individual’s first location is collected in GPS landcover type A, ACC data are collected in 5 bursts, their behaviors are classified and assigned to the first GPS location A and associated landcover type, followed by collection of GPS location B, in which the subsequent 5 ACC bursts are associated to GPS location/landcover type B. In the case of missing GPS locations, we matched ACC bursts to the previous GPS location only if the ACC timestamps were within 60 min of the GPS timestamp, and ACC bursts occurring greater than 60 min after GPS acquisition were removed until the next GPS fix. To account for temporal variation in habitat-behavior relationships, we calculated two continuous covariates representing time-of-day based on the local time associated with the timestamp of each GPS location for each individual. The variable cos(Diel) represented diurnal (negative values) and nocturnal (positive values) periods, and sin(Time) represented midnight until 11:59 a.m. (positive values) and noon until the following 11:59 p.m. (negative values), where high and low values ranged continuously between 1 and − 143. Our temporally matched time series of GPS and ACC data yielded 53,502 GPS locations linked with 300,348 ACC bursts across both winters.We used a Bayesian Markov model with Pólya-Gamma sampling following43), [cf. Refs.44,45] to determine how transitions between landcover types were influenced by behavior, temperature, time-of-day, and wintering region. The proportion of time spent foraging, walking, and stationary between each successive GPS fix was included as a covariate; flight was not included to reduce multicollinearity due to behavior proportions summing to one. Markov models account for non-independence among observations by assuming that the current state (i.e., landcover type) is dependent upon the previous state, and allow the determination of covariate influences on the probability of transitioning among states through a logistic link function. The transition probability from habitat i to habitat j at time t for individual n is modeled with multinomial logistic regression:$$begin{aligned} & logitleft( {p_{nijt} } right) = logleft( {frac{{p_{nijt} }}{{p_{niJt} }}} right) = mathop sum limits_{{r in {mathcal{R}}_{j} }} beta_{0jr} Ileft( {Region_{nt} = r} right) + beta_{1j} {text{cos}}left( {Diel_{nt} } right) \ & quad + beta_{2j} {text{sin}}left( {Time_{nt} } right) + beta_{3ij} Forage_{nt} + beta_{4ij} Walk_{nt} + beta_{5ij} Stationary_{nt} + beta_{6ij} Temperature_{nt} , \ end{aligned}$$where ({mathcal{R}}_{j}) is the set of wintering regions (r) where habitat (j) occurs, (Regio{n}_{rnt}) indicated wintering region (r), and (mathrm{cos}left({Diel}_{nt}right)) and (mathrm{sin}({Time}_{nt})) were temporal covariates (described above) for habitat j. Quantities ({Forage}_{nt}, {Walk}_{nt},mathrm{ and }{Stationary}_{nt}) were the scaled (mean = 0, standard deviation = 1) proportion of time spent in each behavior between transitions from habitat i to habitat j, and ({Temperature}_{nt}) was scaled ambient temperature (°C) for transitions from habitat i to habitat j. All coefficients for transitions to the baseline habitat (J) were set to 0 (i.e., ({beta }_{0Jr}) for all (r), ({beta }_{1J}), ({beta }_{2J}), ({beta }_{3iJ}), ({beta }_{4iJ}), ({beta }_{5iJ}),({beta }_{6iJ}), for all (i)). We set the baseline habitat (J) as open water/unconsolidated shore because this habitat is used primarily for both nocturnal roosting and diurnal loafing, included all behaviors, and transitions to all other landcover types were frequent in each region.The prior for the set of winter region intercepts for each habitat was:$${beta }_{0jr}sim N({beta }_{0j},{sigma }_{0jr}^{2}),$$for (rin {mathcal{R}}_{j}), ({beta }_{0j}) was the mean intercept, and ({sigma }_{0jr}^{2}) was set to 100. For ({beta }_{0j}), a vague prior mean 0 and σ2 = 100 was used with an assumed normal distribution.The Markov model was executed within a Bayesian framework to robustly quantify uncertainty. The Markov model assumed that data were collected at regular time intervals for both GPS (30 min) and ACC (6 min), however imperfect collection by devices created irregular data sets. Therefore, we subsampled GPS locations and constrained time series data to sequences where GPS locations missing  > 120 min intervals (i.e., 4 locations) were separated into sequences of regular time series data for each individual46. We extended43 by including a mix of both transition-specific effects (i.e., behaviors, temperature) and habitat-specific effects (i.e., wintering region, cos(Diel), and sin(Time)), where transition-specific effects were allowed to vary for a current habitat state, while habitat-specific effects were not. We included a mix of coefficients because initial model runs indicated that some effects were similar regardless of the current habitat (i.e., were habitat- and not transition-specific decisions). We also incorporated a model feature to exclude estimation of transitions that did not occur either within the dataset as a whole or within each specific wintering region because landcover types varied among them by setting those specific transition probabilities to zero. We centered and standardized all behavior and temperature covariates, sampled 50,000 iterations from the model posterior using one chain, and discarded the first 10,000 iterations as burn-in. We assessed model convergence by evaluating trace plots and setting random initial values, examined autocorrelation plots, and Geweke diagnostics using the R package ‘coda’47,48,49. More