betaFeedback

WorksMathematics

Mean estimation in spatial functional data

Abstract

Functional mean estimation exhibits a sparse-to-dense phase transition as the number of measurements per curve increases. For spatially indexed functional data, however, dependence between nearby curves and increasing spatial sampling density fundamentally alter this transition. We investigate this phenomenon under a unified framework allowing both increasing-domain and infill asymptotics, with irregular temporal observations. Our analysis identifies two distinct sources of estimation error: an observation-level component governed by the total number of measurements and a curve-level component governed by a spatial dependence-adjusted effective sample size. The resulting uniform convergence rates reveal how spatial dependence shifts the temporal phase-transition boundary and how spatial infill can improve estimation through additional measurements without a corresponding increase in effective curve-level information. We establish these results for three spline-based estimators: a pooled estimator of a stationary mean curve, a local-kernel estimator, and a tensor-product estimator of a spatially varying mean surface. The effects of spatial localization on the phase transition are characterized explicitly. Simulations illustrate the interplay between spatial dependence and temporal sampling density, and environmental data applications demonstrate the proposed estimation methods.

Introduction

The accuracy of functional mean estimation depends critically on the number of measurements available within each curve. With sparse temporal observations, estimation is governed by nonparametric smoothing, whereas sufficiently dense observations allow curve-level variation to become the dominant source of uncertainty. For spatially indexed functional data, this transition is further affected by dependence between curves observed at different locations. In particular, increasing the spatial sampling density may substantially increase the total number of measurements without providing a comparable increase in independent information about the underlying functional process. This raises a fundamental question: How do spatial dependence and spatial sampling density affect the sparse-to-dense transition in functional mean estimation? The problem arises naturally in environmental monitoring, where temperature, pollution, and water-quality profiles are observed at spatially dependent locations and often at irregular time points.

For independent functional data, the effect of within-curve sampling is well understood. Sparse longitudinal methods were developed by Yao et al. [41], and Hall et al. [18] studied the role of the sampling plan in functional principal component analysis. For functional mean estimation, Cai and Yuan [9] established minimax L2L^2-rates and identified the transition to the root-sample-size rate available from complete trajectories. Zhang and Wang [43] provided a unified local-smoothing theory across sparse, semi-dense, and ultra-dense designs. Related developments include simultaneous inference for dense functional means ([10]), optimal sup-norm rates under synchronous designs ([2]), non-asymptotic high-dimensional phase transitions ([17]), and simultaneous inference across sampling regimes ([8]). Together these results show how the number and placement of measurements within a curve determine whether the estimate of mean behaves as a nonparametric or parametric object.

Dependence across curves introduces a second source of information loss. Functional time-series methods describe linear and weakly dependent processes ([3, 19]), mean inference ([21, 22, 11, 29]), frequency-domain structure ([34]), high-dimensional panels ([44]), and locally stationary processes ([40, 15]). Sparse-to-dense transitions have also been studied for structural-break detection and relevant-hypothesis testing in discretely observed functional time series ([6, 7]). Spatial functional data, however, are indexed by pairing with multidimensional geometry rather than an ordered temporal lag. Functional kriging, spatial functional statistics, and spatially dependent principal components have been developed by Delicado et al. [13], Giraldo et al. [16], Martínez-Hernández and Genton [30], and Zhang and Li [42]; see also the broader second-generation framework of Koner and Staicu [25].

While the sparse-to-dense transition is well understood for independent functional observations, much less is known about how it changes under spatial dependence and simultaneous infill and increasing-domain sampling. A central difficulty is that spatial replication generates two different information scales. Additional nearby sites contribute new temporal observations, but their underlying functional trajectories may contain largely redundant information. Consequently, the total number of measurements and the effective number of spatial curves can grow at substantially different rates. As we demonstrate below, this distinction changes the temporal sampling density at which the leading estimation error switches from an observation-limited to a curve-limited regime. Understanding and quantifying this shift is the principal objective of the present paper.

At sites Si=(Si,1,…,Si,d)⊤∈RdS_i = (S_{i,1}, \ldots, S_{i,d})^\top\in\mathbb{R}^d, 1≤i≤mn1 \le i \le m_n, and times Tij∈[0,1]T_{ij} \in[0,1], we observe

Yij=η(Si,Tij)+σ(Si,Tij)ϵij,1≤i≤mn,1≤j≤Ni,(1)Y_{ij} = \eta(S_i,T_{ij}) + \sigma(S_i,T_{ij})\epsilon_{ij}, \qquad1 \le i \le m_n,\qquad1 \le j \le N_i, \tag*{(1)}

where η(s,t)=m(s,t)+X(s,t)\eta(s,t) = m(s,t) + X(s,t) and E{X(s,t)}=0\mathbb{E}\{X(s,t)\} = 0. The random field XX represents spatially dependent curve-to-curve variation and ϵij\epsilon_{ij} is an independent measurement error. We consider both a stationary mean m(t)=m(s,t)m(t) = m(s,t), shared by all sites, and a locally stationary mean surface M(u,t)M(u,t), where M(u,t)M(u,t) approximates the spatially varying mean m(s,t)m(s,t) after appropriately rescaling the expanding spatial domain to the cube [0,1]d[0,1]^d; see Kurisu [26]. This decomposition separates spatially dependent curve-level variation from observation-level variation generated by random sampling times and measurement errors, and this distinction is reflected explicitly in the convergence rates.

Our main contribution is to characterize how spatial dependence and the spatial sampling density affect the sparse-to-dense transition in functional mean estimation. We establish dependence-adjusted uniform convergence rates that reveal two distinct information scales. The observation-level component is governed by the nominal number of sampled measurements, whereas the curve-level component is governed by a dependence-adjusted effective sample size meff,n=mn/Θnm_{\mathrm{eff},n} = m_n/\Theta_n, where Θn\Theta_n quantifies the inflation caused by spatial dependence. The balance between these two components determines the temporal sampling density at which the dominant source of estimation error changes. Under increasing-domain sparse-space sampling, the effective sample size is of the same order as the nominal number of curves, recovering the familiar sparse-to-dense transition. Under spatial infill with persistent local dependence, however, the effective sample size is governed by the spatial domain volume and can therefore be substantially smaller than the nominal number of curves. Consequently, additional spatial observations may continue to improve the observation-level component while contributing comparatively little independent curve-level information. This shifts the temporal phase-transition boundary and reveals an interaction between spatial and temporal sampling that cannot be captured by simply replacing the nominal sample size with an effective sample size. For spatially varying means, spatial localization additionally affects the transition through the kernel-window volume or the spatial-spline resolution.

To develop this theory for different mean structures, we consider three spline-based estimators. A pooled least-squares estimator recovers the stationary mean, a local-kernel estimator recovers the mean curve at an interior spatial point, and a tensor-product spatial spline estimates the full mean surface. Their linear representations reduce the stochastic analysis to empirical Gram matrices and spatially dependent site-level scores and allow the same distinction between observation-level and curve-level information to be traced across the three estimation problems.

We complement the theoretical results by a simulation study that varies sample size, spatial geometry and temporal sampling density, together with additional distributional robustness experiments. Three environmental applications illustrate the estimators under regular-grid, irregular-network, and highly time-sparse sampling designs.

Statistical framework and methodology

We consider the model (1.1)(1.1) and assume that the spatial locations SiS_i are fixed and the within-site sample sizes NiN_i are treated as fixed whereas the observation times TijT_{ij} and the measurement errors εij\varepsilon_{ij} are random. For each fixed site, (ω,t)↦X(s,t;ω)(\omega,t) \mapsto X(s,t;\omega) is jointly measurable with respect to the σ\sigma-field generated by that curve and the Borel sets of [0,1][0,1]. More specifically, we make the following assumptions regarding the measurement locations and errors in model (1.1)(1.1).

Assumption 2.1 (Time design). The NiN_i define the number of measurements of each location. The variables {Tij:1≤i≤mn,1≤j≤Ni}\{T_{ij}: 1 \le i \le m_n, 1 \le j \le N_i\} are i.i.d. with density fTf_T on [0,1][0,1] satisfying 0<cf≤fT(t)≤Cf<∞0 < c_f \le f_T(t) \le C_f < \infty for all t∈[0,1]t \in[0,1], and the entire collection of time points is independent of the latent random field XX. The errors εij\varepsilon_{ij} are i.i.d., centered, variance one, and independent of the latent field XX and of the {Tij}\{T_{ij}\}.

Assumption 2.2 (Infill-increasing spatial design). For each nn, let L1,n,…,Ld,n>0L_{1,n},\ldots,L_{d,n}>0 and Sn=∏r=1d[0,Lr,n]S_n = \prod_{r=1}^{d}[0,L_{r,n}], the spatial domain with volume, Vn=∣Sn∣=∏r=1dLr,nV_n = |S_n| = \prod_{r=1}^{d}L_{r,n}, and Ln=Vn1/dL_n = V_n^{1/d}. For δn>0\delta_n>0, partition Rd\mathbb{R}^d into cubes Δn,ℓ=∏r=1d[ℓrδn,(ℓr+1)δn)\Delta_{n,\ell} = \prod_{r=1}^{d}[\ell_r\delta_n,(\ell_r+1)\delta_n), ℓ=(ℓ1,…,ℓd)∈Zd\ell=(\ell_1,\ldots,\ell_d)\in\mathbb{Z}^d, and set In={ℓ∈Zd:Δn,ℓ∩Sn≠∅}\mathcal{I}_n=\{\ell\in\mathbb{Z}^d:\Delta_{n,\ell}\cap S_n\ne\varnothing\}. Each ℓ∈In\ell\in\mathcal{I}_n contains exactly one spatial location Sℓ∈Δn,ℓ∩SnS_\ell\in\Delta_{n,\ell}\cap S_n. Thus {Si}i=1mn={Sℓ:ℓ∈In}\{S_i\}_{i=1}^{m_n}=\{S_\ell:\ell\in\mathcal{I}_n\} and mn=∣In∣≍Vnδn−dm_n=|\mathcal{I}_n|\asymp V_n\delta_n^{-d}.

Note that VnV_n measures the size of the observation domain, whereas δn\delta_n controls the spatial sampling resolution. Throughout the asymptotic theory, we consider an increasing-domain, that is Vn→∞V_n\to\infty, and distinguish two spatial regimes: (i) if δn≍1\delta_n\asymp1 we refer to this as an sparse space regime; (ii) if δn→0\delta_n\to0 we refer to this as a dense space regime.

Assumption 2.3 (Full-dimensional domain growth). There exist constants 0<cL<CL<∞0<c_L<C_L<\infty such that cL≤Lr,n/Ln≤CLc_L\le L_{r,n}/L_n\le C_L for every 1≤r≤d1\le r\le d and all sufficiently large nn.

Let M1,n=∑i=1mnNiM_{1,n}=\sum_{i=1}^{m_n}N_i denote the number of all observations and M2,n=∑i=1mnNi(Ni−1)M_{2,n}=\sum_{i=1}^{m_n}N_i(N_i-1), and set N‾n=M1,n/mn\overline{N}_n=M_{1,n}/m_n, and N‾2,n=M2,n/mn\overline{N}_{2,n}=M_{2,n}/m_n. Thus N‾n\overline{N}_n is the average number of observations per curve; its growth relative to the dependence-adjusted effective size determines the time-sampling regime. For the estimation of the mean function m:Rd×[0,1]→Rm:\mathbb{R}^d\times[0,1]\to\mathbb{R}, defined by m(s,t)=E[η(s,t)]m(s,t)=\mathbb{E}[\eta(s,t)], we will distinguish the cases of (spatial) stationarity, that is m(s,t)=m(t)m(s,t)=m(t) (for all ss) and a locally stationary mean that can be approximated, after rescaling the spatial domain, by a smooth surface M(u,t)M(u,t), see [26].

Our approach is based on a B-spline estimator. More specifically, we fix an integer p≥2p \ge2, and let BJ(t)=(B1,J(t),…,BJ+p,J(t))T\mathbf{B}_{J}(t) = (B_{1,J}(t), \ldots, B_{J+p,J}(t))^{\mathsf{T}} denote the normalized B-spline basis of order pp on a quasi-uniform partition of the interval [0,1][0,1] with JJ interior knots. Throughout this paper, Jm=JmnJ_m = J_{m_n}, depends on mnm_n, denotes the time-direction spline resolution, and we use the notation Bm(t)=BJm(t)\mathbf{B}_m(t) = \mathbf{B}_{J_m}(t) where the dimension of this vector is Km=Jm+pK_m = J_m + p. For further details on B-spline theory and its use in regression, see, for example, de Boor [12].

Stationary mean

In this section, we consider the model (1) with a stationary mean, that is m(s,t)=m(t)m(s,t) = m(t), for all s∈Rds \in\mathbb{R}^{d}. The spatial locations do not enter the target, so observations from all sites can be pooled to estimate a single temporal curve.

Accordingly, pooling all observations gives the least-squares projection onto the time-spline space, that is

References

  1. [1]M. Benzi and N. Razouk. Decay bounds and O(n) algorithms for approximating functions of sparse matrices. Electronic Transactions on Numerical Analysis, 28:16–39, 2007.
  2. [2]M. Berger, P. Hermann, and H. Holzmann. From dense to sparse design: Optimal rates under the supremum norm for estimating the mean function in functional data analysis. Bernoulli, 31(3):1858–1888, 2025.arxiv.org/abs/2306.04550
  3. [3]D. Bosq. Linear Processes in Function Spaces: Theory and Applications. Springer, 2000.
  4. [4]S. Boucheron, G. Lugosi, and P. Massart. Concentration Inequalities: A Nonasymptotic Theory of Independence. Oxford University Press, 2013.
  5. [5]R. C. Bradley. Basic properties of strong mixing conditions. a survey and some open questions. Probability Surveys, 2:107–144, 2005.DOI
  6. [6]L. Cai and Q. Hu. From sparse to dense functional time series: phase transitions of detecting structural breaks and beyond. arXiv:2412.20858, 2024.arxiv.org/abs/2412.20858
  7. [7]L. Cai and Q. Hu. Unified theory of testing relevant hypothesis in functional time series. arXiv:2508.18624, 2025a.arxiv.org/abs/2508.18624
  8. [8]L. Cai and Q. Hu. From sparse to dense functional data: phase transitions from a simultaneous inference perspective. Statistics and Computing, 35:146, 2025b.arxiv.org/abs/2401.17646
  9. [9]T. T. Cai and M. Yuan. Optimal estimation of the mean function based on discretely sampled functional data: phase transition. Annals of Statistics, 39(5):2330–2355, 2011.DOI
  10. [10]G. Cao, L. Yang, and D. Todem. Simultaneous inference for the mean function based on dense functional data. Journal of Nonparametric Statistics, 24(2):359–377, 2012.pubmed.ncbi.nlm.nih.gov/22665964
  11. [11]M. Chen and Q. Song. Simultaneous inference of the mean of functional time series. Electronic Journal of Statistics, 9(2):1779–1798, 2015.DOI
  12. [12]C. de Boor. A Practical Guide to Splines. Springer, revised edition, 2001.
  13. [13]P. Delicado, R. Giraldo, C. Comas, and J. Mateu. Statistics for spatial functional data: Some recent contributions. Environmetrics, 21(3–4):224–239, 2010.DOI
  14. [14]S. Demko, W. F. Moss, and P. W. Smith. Decay rates for inverses of band matrices. Mathematics of Computation, 43(168):491–499, 1984.
  15. [15]H. Dette and W. Wu. Confidence surfaces for the mean of locally stationary functional time series. Statistica Sinica, 36:583–604, 2026.
  16. [16]R. Giraldo, P. Delicado, and J. Mateu. Ordinary kriging for function-valued spatial data. Environmental and Ecological Statistics, 18:411–426, 2011.DOI
  17. [17]S. Guo, D. Li, X. Qiao, and Y. Wang. From sparse to dense functional data in high dimensions: Revisiting phase transitions from a non-asymptotic perspective. Journal of Machine Learning Research, 26(15):1–40, 2025.
  18. [18]P. Hall, H.-G. Müller, and J.-L. Wang. Properties of principal component methods for functional and longitudinal data analysis. Annals of Statistics, 34(3):1493–1517, 2006.
  19. [19]S. Hörmann and P. Kokoszka. Weakly dependent functional data. Annals of Statistics, 38(3):1845–1884, 2010.DOI
  20. [20]S. Hörmann and P. Kokoszka. Consistency of the mean and the principal components of spatially distributed functional data. Bernoulli, 19(5A):1535–1558, 2013.
  21. [21]L. Horváth and P. Kokoszka. Inference for Functional Data with Applications. Springer, 2012.DOI
  22. [22]L. Horváth, P. Kokoszka, and R. Reeder. Estimation of the mean of functional time series and a two-sample problem. Journal of the Royal Statistical Society: Series B, 75(1):103–122, 2013.arxiv.org/abs/1105.0019
  23. [23]B. Huang, C. Liu, V. Banzon, E. Freeman, G. Graham, B. Hankins, T. Smith, and H.-M. Zhang. Improvements of the daily optimum interpolation sea surface temperature (DOISST) version 2.1. Journal of Climate, 34(8):2923–2939, 2021.DOI
  24. [24]N. Jenish and I. R. Prucha. Central limit theorems and uniform laws of large numbers for arrays of random fields. Journal of Econometrics, 150(1):86–98, 2009.pubmed.ncbi.nlm.nih.gov/20161289
  25. [25]S. Koner and A.-M. Staicu. Second-generation functional data. Annual Review of Statistics and Its Application, 10:547–572, 2023.DOI
  26. [26]D. Kurisu. Nonparametric regression for locally stationary random fields under stochastic sampling design. Bernoulli, 28(2):1250–1275, 2022.
  27. [27]S. N. Lahiri. Central limit theorems for weighted sums of a spatial process under a class of stochastic and fixed designs. Sankhyā: The Indian Journal of Statistics, Series A, 65(2):356–388, 2003.
  28. [28]S. N. Lahiri, T. S. McElroy, and D. Weinberg. Locally stationary spatial processes. Sankhyā A, 2025. doi: 10.1007/s13171-025-00403-y.DOI
  29. [29]J. Li and L. Yang. Statistical inference for functional time series. Statistica Sinica, 33(1):519–549, 2023.
  30. [30]I. Martínez-Hernández and M. G. Genton. Recent developments in complex and spatially correlated functional data. Brazilian Journal of Probability and Statistics, 34(2):204–229, 2020.
  31. [32]National Water Quality Monitoring Council. Water quality portal. U.S. Geological Survey and U.S. Environmental Protection Agency, 2021.
  32. [34]V. M. Panaretos and S. Tavakoli. Fourier analysis of stationary time series in function space. Annals of Statistics, 41(2):568–603, 2013.DOI
  33. [35]E. Rio. Covariance inequalities for strongly mixing processes. Annales de l’I.H.P. Probabilités et Statistiques, 29(4):587–597, 1993.
  34. [36]L. L. Schumaker. Spline Functions: Basic Theory. Cambridge University Press, 3 edition, 2007.
  35. [37]U.S. Environmental Protection Agency. AirData: Pre-generated air quality system data files. Air Quality System, 2026a.
  36. [38]U.S. Environmental Protection Agency. Ground-level ozone basics, 2026b.
  37. [39]U.S. Geological Survey. Nutrients and eutrophication, 2019.
  38. [40]A. van Delft and M. Eichler. Locally stationary functional time series. Electronic Journal of Statistics, 12(1):107–170, 2018.DOI
  39. [41]F. Yao, H.-G. Müller, and J.-L. Wang. Functional data analysis for sparse longitudinal data. Journal of the American Statistical Association, 100(470):577–590, 2005.DOI
  40. [42]H. Zhang and Y. Li. Unified principal component analysis for sparse and dense functional data under spatial dependency. Journal of Business & Economic Statistics, 40(4):1523–1537, 2022.
  41. [43]X. Zhang and J.-L. Wang. From sparse to dense functional data and beyond. Annals of Statistics, 44(5):2281–2321, 2016.DOI
  42. [44]Z. Zhou and H. Dette. Statistical inference for high-dimensional panel functional time series. Journal of the Royal Statistical Society: Series B, 85(2):523–549, 2023.arxiv.org/abs/2003.05968

Paper details

vlnc:W_6g2afprb
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

generative AI tools

View statement
Looking for related papers…
Contents