Pith. sign in

REVIEW 4 major objections 4 minor 41 references

Centroidal Voronoi Tessellation Based Methods for Optimal Rain Gauge Location Prediction

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Rain gauges can be placed optimally by solving a correlation-weighted centroidal Voronoi tessellation.

desk verdict A useful proof-of-concept for CVT-based rain gauge siting, but the central optimality claim rests on a circular comparison and an unvalidated density. read the letter →

arxiv 1908.10403 v2 pith:S7DZQEZV submitted 2019-08-27 math.NA cs.NA

Zichao Wendy Di , Viviana Maggioni , Yiwen Mei , Marilyn Vazquez , Paul Houser , Maria Emelianenko This is my paper · ORCID
classification math.NAcs.NA MSC 65D1865K10
keywords raingaugesCVTdecorrelationoptimalplacementprecipitationvariabilityPERSIANN-CCSVoronoitessellationnetworkdesign
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that rain gauge networks can be placed automatically by converting local precipitation variability into a spatial density and then solving a centroidal Voronoi tessellation (CVT) problem. The density is built from the average pairwise rainfall correlation at a single decorrelation distance, the separation at which correlation falls to $1/e$, and a truncated-Newton solver finds the gauge locations that minimize the weighted tessellation energy. Applied to satellite precipitation data over the Alto-Adige region of Northern Italy and Oklahoma, the method places more gauges in mountainous, low-correlation terrain and fewer in flat, high-correlation terrain, matching the observed physical controls on rainfall. The paper reports that the computed networks lower the tessellation error from $259.56$ to $15.02$ for Adige and from $8.61\times 10^4$ to $4.67\times 10^4$ for Oklahoma compared with the existing real gauge networks.

What carries the argument

The central object is the weighted centroidal Voronoi tessellation: a partition of the region into Voronoi cells whose generators are also the mass centroids of the cells with respect to a density $\rho(x)$. The load-bearing identity is the cell-size scaling $h_{V_i}/h_{V_j}=(\rho(x_i)/\rho(x_j))^{1/3}$, which turns a density contrast into a gauge-density contrast. The density itself is constructed as a power law of the local average correlation, $\rho(x)=r+R\left(\frac{C_{\max}-\operatorname{Corr}(x)}{C_{\max}-C_{\min}}\right)^{\alpha}$, where $\operatorname{Corr}(x)$ averages Pearson correlations on a circle of radius equal to the decorrelation distance; that distance is estimated from an exponential variogram model with nugget by the $1/e$ rule. The numerical work is carried by a truncated-Newton solver that minimizes the discrete CVT energy over the generators.

What would settle it

Take the same two regions, build the density from a training time period, solve the CVT problem, and compute the actual squared rainfall estimation error (2.2) of the resulting network on a held-out time period; if it is not lower than the error of the existing Adige and Oklahoma networks, the paper's central equivalence between minimizing (2.1) and minimizing expected error is falsified.

Watch

Extended reading notes

Core claim

The central claim of the paper is that the optimal placement of rain gauges is attained by minimizing the weighted CVT energy $$E(\{x_i\})=\sum_{i=1}^k \int_{V_i} \rho(x)\,\|x-x_i\|^2\,dx$$ with density $\rho(x)=r+R\left(\frac{C_{\max}-\operatorname{Corr}(x)}{C_{\max}-C_{\min}}\right)^{\alpha}$, where $\operatorname{Corr}(x)$ is the average Pearson correlation between location $x$ and points on a circle whose radius is the decorrelation distance. Because the relation $h_{V_i}/h_{V_j}=(\rho(x_i)/\rho(x_j))^{1/3}$ ties cell size to density, low-correlation regions receive more, smaller cells and therefore more gauges. The authors support this claim by comparing the optimized placements with the real Adige and Oklahoma networks, finding that the optimized networks concentrate gauges in complex terrain and that the tessellation error (2.1) drops from $259.56$ to $15.02$ for Adige and from $8.61\times 10^4$ to $4.67\times 10^4$ for Oklahoma. On the paper's own terms, this confirms that the CVT solution captures observed precipitation variability and that the resulting network is optimal for the cost function (2.1).

Load-bearing premise

The load-bearing premise is that the correlation-derived density $\rho(x)$ in (2.6), computed from average correlation at one global decorrelation distance, is a faithful surrogate for the true expected squared rainfall estimation error (2.2), so that minimizing the CVT energy truly places gauges where precipitation is hardest to estimate.

Editorial extensions

If this is right

  • Gauge networks can be designed from satellite-derived precipitation fields alone, so the same automated procedure applies to regions with no existing ground instruments.
  • The power parameter $\alpha$ and the target gauge count $k_g$ let a network designer trade cost against spatial representativeness by setting the correlation threshold $C_{\mathrm{tol}}$.
  • Because the density is recomputed from updated time-series correlations, the method can track climate-driven shifts in precipitation variability rather than relying on fixed siting criteria.
  • The same density-to-CVT pipeline transfers to other siting problems, such as air-quality monitors or soil-moisture sensors, whenever local variability can be expressed as a density.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural stress test would replace the paper's single decorrelation distance with a spatially varying one; in complex terrain the true correlation length is likely shorter than the region-wide value, and the optimal density map could shift substantially.
  • Because the paper grades its own solution with the same energy (2.1) that the optimizer minimizes, an independent comparison should validate against held-out rainfall interpolation error (2.2); that would separate genuine predictive gain from mere cost-function tuning.
  • The paper's use of a 0.04$^\circ$ = 5 km conversion for both regions is a first-order approximation; a proper map projection would sharpen the density and gauge positions near coastlines and high mountains.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes an automated method for rain gauge placement based on centroidal Voronoi tessellations (CVTs). It constructs a density map from local precipitation correlations at an estimated decorrelation distance, then minimizes the weighted CVT energy (2.1)/(2.7) using a truncated-Newton solver. The method is applied to satellite precipitation data over the Alto-Adige region and Oklahoma, and the resulting locations are compared with existing gauge networks. The paper claims that the CVT energy is an approximation error and that the algorithm produces optimal gauge locations.

Significance. If the proposed density map were a faithful surrogate for precipitation estimation error, the method would be practically valuable, and the choice of two regions with different topography is a sensible testbed. The CVT/TN optimization machinery is standard and the algorithm is described clearly enough to reproduce. However, the optimality claim is not supported: the link between the rainfall-estimation objective (2.2) and the CVT energy (2.7) is asserted rather than derived, the key numeric comparison is circular, and the density parameters are chosen in an ad hoc manner with no sensitivity or independent validation. These are the paper's central claims, not presentation issues.

major comments (4)
  1. [Section 2.2, Eqs. (2.2)–(2.7)] The derivation of the main objective is missing. The paper starts from the expected squared approximation error (2.2), which is an integral over each Voronoi cell of 2ε²[1−Corr(x,xi)], and then replaces it with the CVT energy (2.7) using the Euclidean metric and a scalar density ρ(x) of the form (2.6). Equation (2.5), the cell-size relation, does not connect these objectives; it is a property of any CVT density. The effective correlation Corr(x) in Eq. (4.9) is an angular average at a single global decorrelation distance, so it discards the pairwise distance/direction dependence that (2.2) explicitly integrates. Consequently, it is not shown that minimizers of (2.7) minimize (2.2), and the term 'optimal' in the title and in Sections 4.2 and 5 is not justified.
  2. [Section 5, Figure 10] The numerical validation is circular. Figure 10 reports the CVT energy E from Eq. (2.1) for the algorithm's locations and for the real gauge locations. Since Algorithm 1 minimizes exactly this E, the optimized set is expected, by construction, to have lower or equal E; the comparison only demonstrates that the optimizer solved its own problem. It does not provide evidence that the resulting network better represents precipitation. The authors should compare against a metric independent of the optimized functional, such as the expected squared error (2.2), kriging, or leave-one-out cross-validation on withheld gauge records.
  3. [Section 4.2, Eq. (4.12), and Section 2.2] The density model is not anchored to an estimation-error model. The power-law form (2.6), the bounds r=10⁻⁶ and R=1, and especially the exponent α are selected by the heuristic criterion (4.12) that matches a target count of low-correlation grid points, not by any relationship to rainfall estimation error. No sensitivity analysis is given, although Figure 5 shows that α has a strong effect on the result. Since the entire placement is driven by this user-defined density, the conclusions are not robust to the free parameters.
  4. [Section 5, Tables 1 and 2] The proximity statistics are presented without a baseline, and the Adige table is internally inconsistent. Table 1 sums to 138 gauges, while Section 3 states that the Alto-Adige network has 192 gauges; this leaves 54 gauges unaccounted for. With roughly 138 (Adige) and 118 (Oklahoma) CVT locations, a large fraction of gauges falling within 10–15 km of some CVT point would likely occur under random placement as well. The paper should compare against a null model or random gauge configurations, and it should reconcile the gauge counts, before these tables can support any conclusion.
minor comments (4)
  1. [Table 2] Table 2 lists radius 15 twice, once with 49 gauges within and 69 not within and once with 86 within and 32 not within; one of these likely refers to a different radius and should be corrected.
  2. [Sections 2.2, 4.2, and 5] The default value of r is reported inconsistently: Section 2.2 and Section 5 use r=10⁻⁶ and R=1, while Section 4.2 says 'default values r=1, R=1'.
  3. [Section 4.1, Eq. (4.11)] The definition of the decorrelation distance d0 is ambiguous: the text treats d0 as a single global value per region, but Eq. (4.11) is written for a location-dependent Corrd(x); the paper should clarify whether d0 is global or spatially varying.
  4. [Appendix A, Eq. (8.13)] In Eq. (8.13), the variance-difference term [Var(x)−Var(xi)]² should be accompanied by a comment explaining how it vanishes under the constant-variance assumption used to reach Eq. (2.2), since the later approximation uses Var(x)=ε.

Circularity Check

1 steps flagged · score 6.0 of 10

Validation of optimality is circular: Figure 10 compares the CVT energy (2.1) that Algorithm 1 minimizes, so the energy drop is guaranteed by construction; the correlation-density surrogate for rainfall error is asserted, not derived.

  1. self definitional [Section 4.2, Section 5, Figure 10; Algorithm 1 lines 13-14]
    "As discussed above, optimal placement is attained by computing CVT that minimizes the approximation error (2.1). ... As shown in Figure 10, the approximation error is significantly decreased by running the optimization routine."

    Algorithm 1 minimizes exactly the energy E in (2.1) with density (2.6), and then reports that same CVT energy for the optimized generators. Comparing this value with the same functional E evaluated at real-gauge locations cannot independently validate optimality: any local minimizer produced by the optimization routine is expected, by construction, to have lower or equal E. The reported drops (259.56 to 15.02 for Adige; 8.61e4 to 4.67e4 for Oklahoma) are a restatement of the definition of the objective being minimized, not a check that the gauges minimize rainfall estimation error (2.2). Since the paper defines optimal placement as the minimizer of (2.1), the Figure 10 comparison is circular as evidence for the central claim.

full rationale

The main derivation chain is: precipitation time series to correlation map (4.9)-(4.10) to density (2.6) to CVT optimization (2.7) to gauge locations. The optimization step is mathematically sound and self-contained for the CVT energy; no circularity arises in the solver itself. The circularity enters at the validation stage: Figure 10 measures the CVT energy (2.1), the exact objective minimized in Algorithm 1, so the lower error for optimized locations is guaranteed by construction and cannot confirm that this objective corresponds to rainfall estimation error. The paper's statement in Section 4.2 defines optimal placement as the minimizer of (2.1), making the Figure 10 comparison self-definitional. A separate concern is that the surrogate density (2.6) is asserted, via (2.5), to achieve a similar effect to the Okabe/Du objective (2.2); no derivation is supplied, and (2.5) is merely a property of any CVT. This is an unproven modeling assumption rather than an additional circular step, but it means the quantitative optimality evidence reduces to the algorithm's own cost function. The self-citations in the paper ([6], [15], [19]-[21]) are not load-bearing for the central claim; the CVT theory citations [12, 23] are external and standard. Because the density construction and the proximity comparisons to real gauges (Tables 1-2) provide some independent content, the circularity is partial rather than total, giving a score of 6.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on hand-picked density parameters (r, R, alpha, C_tol, k_g) and a set of modeling assumptions about stationarity, the variogram form, and the validity of the power-law density. The density form is ad hoc, and the equivalence to the true sampling error objective is asserted rather than derived. No new physical entities are introduced.

free parameters (6)
  • r (lower density bound) = 10^-6
    Set by hand in Section 2.2 'to provide a proof of concept' and 'optimized for the use of TN CVT construction algorithm'; not determined by data or physical constraints.
  • R (upper density scale) = 1
    Chosen together with r to give R/r = 10^6; hand-picked, not derived from the data or from an external benchmark.
  • alpha (power exponent) = 3 for Adige; Oklahoma value not explicitly stated
    Selected via the heuristic (4.12) as the smallest integer such that the number of grid points with relative correlation below C_tol is at most k_g. This is a fit to the desired gauge count, not a derived quantity.
  • C_tol (correlation threshold) = 0.1 (default)
    User-defined input in Algorithm 1; not estimated from data.
  • k_g (number of gauges) = 138 for Adige comparison
    Specified by the user; the comparison in Table 1 and Figure 6 uses 138 real gauges, which mismatches the 192-gauge network described in Section 3 and suggests an unexplained selection.
  • decorrelation distance d0 = 45 km for Adige, 80 km for Oklahoma
    Estimated in Figure 3 by fitting an exponential variogram model (4.8) and applying the 1/e rule (4.11); fitted to data and used globally for the whole region in Algorithm 1.
assumptions (4)
  • domain assumption The precipitation field z(x) is a random process with constant mean and variance and correlation depending only on separation distance (weak stationarity).
    Invoked in Section 2.2 to justify the correlation-based energy (2.2) and its link to the CVT energy; the paper itself notes this 'might or might not hold in practice' and cites [6]. Complex topography in Adige makes this questionable.
  • ad hoc to paper The density function (2.6) with power-law form is a suitable model for the relationship between correlation and gauge density.
    The functional form (2.6) is introduced in Section 2.2 without derivation; the authors state 'there are many functional forms such a relation can take', making this an arbitrary modeling choice.
  • standard math The cell-size scaling relation h_Vi/h_Vj = (rho(x_i)/rho(x_j))^(1/3) holds pointwise and justifies interpreting R/r as the ratio of largest to smallest cell sizes.
    Equation (2.5) is an asymptotic result for CVTs with smooth densities in 2D; applying it locally to justify the density construction is heuristic, and the range r=10^-6 to R=1 produces a density ratio of 10^6 that may not be realizable for finite k.
  • domain assumption The exponential variogram model (4.8) with nugget effect adequately represents the spatial correlation of precipitation.
    Used in Section 4.1 to define the decorrelation distance; the exponential model is standard in geostatistics but is still a modeling assumption, and the fitted distance is used uniformly over heterogeneous terrain.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Centroidal Voronoi Tessellation Based Methods for Optimal Rain Gauge Location Prediction." pith.science (2026). https://pith.science/paper/S7DZQEZV

@misc{pith2026190810403,
  author       = {Pith},
  title        = {Pith review of: Centroidal Voronoi Tessellation Based Methods for Optimal Rain Gauge Location Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S7DZQEZV}},
  note         = {Machine review of arXiv:1908.10403}
}
read the original abstract

With more satellite and model precipitation data becoming available, new analytical methods are needed that can take advantage of emerging data patterns to make well informed predictions in many hydrological applications. We propose a new strategy where we extract precipitation variability patterns and use correlation map to build the resulting density map that serves as an input to centroidal Voronoi tessellation construction that optimizes placement of precipitation gauges. We provide results of numerical experiments based on the data from the Alto-Adige region in Northern Italy and Oklahoma and compare them against actual gauge locations. This method provides an automated way for choosing new gauge locations and can be generalized to include physical constraints and to tackle other types of resource allocation problems.

Figures

Figures reproduced from arXiv: 1908.10403 by the authors.

Figure 1
Figure 1. Illustration of the effect of the ratio of [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Topographic maps with exist￾ing gauge locations: (top) Oklahoma region; (bottom) Adige region. Precipitation climatology in the area exhibits strong spatial gradi￾ents, with mean annual precipita￾tion varying from ∼500 mm in the Northwestern region to ∼1700 mm in the Southeastern part ([26], [27]). A network of 192 rain gauges is available in the region. The rain gauges are distributed quite uni￾formly over the area… view at source ↗
Figure 3
Figure 3. Calculation of the decorrelation distance using 1 [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Density computation for the Italy region: (top left) Effective correlation map [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Comparing density computation for different [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Optimized rain gauge locations compared with real gauge data for the Adige [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Optimized rain gauge locations compared with real gauge data for the Oklahoma [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Comparing existing locations with optimal placement for the Oklahoma region. [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: Comparing existing locations with optimal placement for Adige region. [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Decay of the approximation error E (2.1) during TN iteration for Oklahoma region (left) and Adige (right). CVT-based location denoted as a red asterisk gives a smaller error value comparing to the error computed for the real-gauge locations (shown as a blue asterisk).…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 40 canonical work pages

  1. [1]

    Gottschalck, J

    J. Gottschalck, J. Meng, M. Rodell, P. Houser, Analysis of multiple pre- cipitation products and preliminary assessment of their impact on global land data assimilation system land surface states, J. Hydrometeorol. 6 (2005) 573598

    work page 2005

  2. [2]

    Hazra, V

    A. Hazra, V. Maggioni, P. Houser, H. Antil, M. Noonan, A Monte Carlo- 17 based multi-objective optimization approach to merge different precip- itation estimates for land surface modeling, J. Hydrology 570 (2019) 454–462

    work page 2019

  3. [3]

    Maggioni, P

    V. Maggioni, P. Meyers, M. Robinson, A review of merged high- resolution satellite precipitation product accuracy during the tropical rainfall measuring mission (trmm) era, Journal of Hydrometeorology 17(4) (2016) 1101–1117

    work page 2016

  4. [4]

    Anagnostou, V

    E. Anagnostou, V. Maggioni, E. Nikolopoulos, T. Meskele, F. Hossain, A. Papadopoulos, Benchmarking high-resolution global satellite rainfall products to radar and rain-gauge rainfall estimates, IEEE Transactions on Geoscience and Remote Sensing 48(4) (2009) 1667–1683

    work page 2009

  5. [5]

    C. Kidd, P. Bauer, J. Turk, G. Huffman, R. Joyce, K. Hsu, D. Braith- waite, Intercomparison of high-resolution precipitation products over northwest Europe, J. Hydrometeor. 13 (6783)

    work page

  6. [6]

    Emelianenko, V

    M. Emelianenko, V. Maggioni, Mathematical challenges in measuring variability patterns for precipitation analysis, in: H. Kaper, F. Roberts (Eds.), Mathematics of Planet Earth, Springer, 2019, pp. 49–61

    work page 2019

  7. [7]

    Agrell, T

    E. Agrell, T. Ericsson, Optimization of lattices for quantization, IEEE Trans. Inf. Theory 44(5) (1998) 1814–1828

    work page 1998

  8. [8]

    Thiessen, Precipitation averages for large areas, Monthly Weather review 39 (1911) 1081–1084

    A. Thiessen, Precipitation averages for large areas, Monthly Weather review 39 (1911) 1081–1084

    work page 1911

Show all 41 references
  1. [9]

    Ringler, R

    T. Ringler, R. Heikes, D. Randall, Modeling the atmospheric general circulation using a spherical geodesic grid: a new class of dynamical cores, Mon Weather Rev 128 (2000) 2471–2490

    2000

  2. [10]

    Q. Du, M. Gunzburger, L. Ju, Advances in studies and applications of centroidal Voronoi tessellations, Numer. Math. Theor. Meth. Appl. 3 (2) (2000) 119–142

    2000

  3. [11]

    L. Chen, J. Xu, Optimal Delaunay triangulations, J. Comp. Math. 22 (2004) 299–308

    2004

  4. [12]

    Q. Du, V. Faber, M. Gunzburger, Centroidal Voronoi tessellations: ap- plications and algorithms, SIAM Review 41 (1999) 637–676. 18

    1999

  5. [13]

    S. G. Nash, A survey of truncated-Newton methods, Journal of Compu- tational and Applied Mathematics 124 (1) (2000) 45–59

    2000

  6. [14]

    Lloyd, Least square quantization in PCM, IEEE Transactions on Information Theory 28 (1982) 129–137

    S. Lloyd, Least square quantization in PCM, IEEE Transactions on Information Theory 28 (1982) 129–137

    1982

  7. [15]

    Z. Di, M. Emelianenko, S. Nash, Truncated Newton-based multigrid al- gorithm for centroidal Voronoi calculation, Numer. Math. Theor. Meth. Appl. 5 (1) (2012) 242–259

    2012

  8. [16]

    Burkardt, M

    J. Burkardt, M. Gunzburger, H.-C. Lee, Centroidal Voronoi tessellation- based reduced order modeling of complex systems, SIAM J. Sci. Com- put. 28 (2) (2006) 459–484

    2006

  9. [17]

    Zhang, M

    J. Zhang, M. Emelianenko, Q. Du, Periodic centroidal Voronoi tessella- tions, Intern. J. Num. Anal. Modeling 9 (2012) 950–969

    2012

  10. [18]

    Ringler, L

    T. Ringler, L. Ju, M. Gunzburger, A multiresolution method for climate system modeling: application of spherical centroidal Voronoi tessella- tions, Ocean Dynamics 58 (2008) 475–498

    2008

  11. [19]

    Emelianenko, Fast multilevel CVT-based adaptive data visualization algorithm, Numer

    M. Emelianenko, Fast multilevel CVT-based adaptive data visualization algorithm, Numer. Math. Theor. Meth. Appl. 3 (2) (2010) 195–211

    2010

  12. [20]

    Q. Du, M. Emelianenko, Acceleration schemes for computing the cen- troidal Voronoi tessellations, Numer. Linear Algebra Appl. 13 (2006) 173–192

    2006

  13. [21]

    Q. Du, M. Emelianenko, Uniform convergence of a nonlinear energy- based multilevel quantization scheme via centroidal Voronoi tessella- tions, SIAM J. Numer. Anal. 46 (2008) 1483–1502

    2008

  14. [22]

    L. Chen, M. Holst, Efficient mesh optimization schemes based on opti- mal Delaunay triangulations, Comp. Methods Appl. Mech. Engrg. 200 (2011) 967–984

    2011

  15. [23]

    Okabe, B

    A. Okabe, B. Boots, K. Sugihara, Spatial Tessellations; Concepts and Applications of Voronoi Diagrams, Wiley, Chichester, 1992

    1992

  16. [24]

    Q. Du, M. Emelianenko, L. Ju, Convergence properties of the lloyd algorithm for computing the centroidal Voronoi tessellations, SIAM J. Numer. Anal. 44 (2006) 102–119. 19

    2006

  17. [25]

    Brock, K

    F. Brock, K. Crawford, R. Elliott, G. Cuperus, S. Stadler, H. John- son, M. Eilts, The Oklahoma Mesonet: A technical overview, J. Atmos. Oceanic Technol. 12(1) (1995) 5–19

    1995

  18. [26]

    Maggioni, E

    V. Maggioni, E. Nikolopoulos, E. Anagnostou, M. Borga, Modeling satellite precipitation errors over mountainous terrain: The influence of gauge density, seasonality, and temporal resolution, IEEE Transactions on Geoscience and Remote Sensing 55(7) (2017) 4130–4140

    2017

  19. [27]

    E. I. Nikolopoulos, M. Borga, F. Marra, S. Crema, L. Marchi, Debris flows in the eastern Italian Alps: Seasonality and atmospheric circula- tion patterns, Natural Hazards Earth Syst. Sci. 15 (3) (2015) 647656

    2015

  20. [28]

    K.-L. Hsu, X. Gao, S. Sorooshian, H. V. Gupta, Precipitation estimation from remotely sensed information using artificial neural networks, J. Appl. Meteorol. 36 (9) (1997) 11761190

    1997

  21. [29]

    Sorooshian, K.-L

    S. Sorooshian, K.-L. Hsu, X. Gao, H. V. Gupta, B. Imam, D. Braith- waite, Evaluation of PERSIANN system satellitebased estimates of trop- ical rainfall, Bull. Amer. Meteorol. Soc. 81 (2000) 20352046

    2000

  22. [30]

    K.-L. Hsu, S. Sorooshian, Satellite-based precipitation measurement us- ing PERSIANN system, in: Hydrol. Model. Water Cycle, Berlin, Ger- many: Springer-Verlag, 2008, p. 2748

    2008

  23. [31]

    K.-L. Hsu, A. Behrangi, B. Iman, S. Sorooshian, Extreme precipitation estimation using satellite-based PERSIANN-CCS algorithm, in: M. Ge- bremichael, F. Hossain (Eds.), Satellite Rainfall Applications for Surface Hydrology, Springer, 2010, p. 4967

    2010

  24. [32]

    Ciach, W

    G. Ciach, W. Krajewski, On the estimation of radar rainfall error vari- ance, Adv. Water Resources 22(6) (1999) 585–595

    1999

  25. [33]

    Ciach, W

    G. Ciach, W. Krajewski, Analysis and modeling of spatial correlation structure in small-scale rainfall in Central Oklahoma, Adv. Water Re- sources 29(10) (2006) 1450–1463

    2006

  26. [34]

    Cressie, Statistics for spatial data, John Wiley and Sons, 1993

    N. Cressie, Statistics for spatial data, John Wiley and Sons, 1993

    1993

  27. [35]

    Ciach, Local random errors in tipping-bucket rain gauge measure- ments, Journal of Atmospheric and Oceanic Technology 20(5) (2003) 752–759

    G. Ciach, Local random errors in tipping-bucket rain gauge measure- ments, Journal of Atmospheric and Oceanic Technology 20(5) (2003) 752–759. 20

    2003

  28. [36]

    R. E. Caflisch, Monte Carlo and quasi-Monte Carlo methods, Acta Nu- merica 7 (1998) 149. doi:10.1017/S0962492900002804

    1998 doi

  29. [37]

    Sanchez-Moreno, C

    J. Sanchez-Moreno, C. Mannaerts, V. Jetten, Influence of topography on rainfall variability in santiago island, cape verde, Intl. J. Climatology 34(4) (2014) 1081–1097

    2014

  30. [38]

    Johnson, C

    G. Johnson, C. Hanson, Topographic and atmospheric influences on precipitation variability over a mountainous watershed, J. Applied Me- teorology 34(1) (1995) 68–87

    1995

  31. [39]

    S. G. Nash, J. Nocedal, A numerical study of the limited memory BFGS method and the truncated-Newton method for large scale optimization, SIOPT 1 (1991) 358–372

    1991

  32. [40]

    Following [23], let z(x) be the random variable representing rainfall at location x∈W

    Appendix A: Optimization problem formulation. Following [23], let z(x) be the random variable representing rainfall at location x∈W . We can think of it as z(x) =m(x) +ϵ(x) with E(ϵ(x))] = 0, so that m(x) = E(z(x)) represents the average trend of the process over the given reg...

    1911

  33. [41]

    For more details we refer readers to [13]

    Appendix B: Truncated-Newton Algorithm for CVT calculation Here, we give a brief review of the truncated Newton algorithm. For more details we refer readers to [13]. To optimize a problem of the form min x f(x) 22 at the j-th TN iteration a search direction p is computed as an...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.