BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.cienciaenmexico.com//igisc-2026//speaker//7GNGZZ
BEGIN:VTIMEZONE
TZID:CST
BEGIN:STANDARD
DTSTART:20001029T030000
RRULE:FREQ=YEARLY;BYDAY=5SU;BYMONTH=10;UNTIL=20001029T080000Z
TZNAME:CST
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
END:STANDARD
BEGIN:STANDARD
DTSTART:20010930T030000
RRULE:FREQ=YEARLY;BYDAY=5SU;BYMONTH=9;UNTIL=20010930T080000Z
TZNAME:CST
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
END:STANDARD
BEGIN:STANDARD
DTSTART:20021027T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CST
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000402T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20000402T090000Z
TZNAME:CDT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20010506T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=5;UNTIL=20010506T090000Z
TZNAME:CDT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
END:DAYLIGHT
BEGIN:DAYLIGHT
DTSTART:20020407T030000
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4;UNTIL=20220403T090000Z
TZNAME:CDT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-igisc-2026-Z3ZF3B@pretalx.cienciaenmexico.com
DTSTART;TZID=CST:20261112T125000
DTEND;TZID=CST:20261112T130500
DESCRIPTION:**1. Introduction**\n\nReliable estimates of above-ground bioma
 ss density (AGBD) are essential for forest carbon accounting\, REDD+\, and
  conservation policies. Although Mexico's National Forest Inventory (INFyS
 ) provides the most reliable field-based AGBD data\, its limited spatial c
 overage and high acquisition costs constrain large-scale monitoring\, part
 icularly in tropical forests. Earth Observation (EO) products\, such as ES
 A CCI Biomass and GEDI-derived canopy height\, offer continuous spatial co
 verage but are affected by non-stationary relationships with AGBD and sens
 or saturation. Consequently\, integrating EO and field data requires model
 s that explicitly account for spatial dependence and uncertainty\, making 
 spatially varying coefficient (SVC) geostatistical models a suitable appro
 ach.\n\n**2. Research Gap**\nModels fitted over large\, heterogeneous doma
 ins often borrow strength across broad ecological gradients and tend to sh
 rink local predictions toward the national mean. While statistically effic
 ient at the national scale\, this can lead to regional miscalibration\, pa
 rticularly where Earth Observation (EO) covariates lose sensitivity. Both 
 ESA Climate Change Initiative Biomass (CCI Biomass)\, which becomes weakly
  informative beyond roughly 150–200 Mg ha$^{-1}$\, and GEDI-derived cano
 py height\, which saturates near the upper end of tall tropical canopies\,
  degrade precisely in dense tropical forests. Quintana Roo\, with semi-eve
 rgreen tropical forest\, secondary vegetation\, coastal gradients\, karsti
 c substrates\, and heterogeneous land-use histories\, exemplifies this con
 dition. A state-scale application therefore requires an independent modeli
 ng framework that jointly addresses response scale\, data quality\, and lo
 cal covariate behavior\, explicitly identifying where each EO source drive
 s predictions.\n**3. Methodology**\n\nWe develop such a reformulation for 
 Quintana Roo within the SVC–SPDE framework. On a transformed response\, 
 the model is written in the standard additive form\n\n$$ g{Y(\\mathbf{s})}
 =[\\alpha+\\tilde{\\alpha}(\\mathbf{s})]+[\\beta+\\tilde{\\beta}(\\mathbf{
 s})]x_1(\\mathbf{s})+[\\eta+\\tilde{\\eta}(\\mathbf{s})]x_2(\\mathbf{s})+\
 \epsilon(\\mathbf{s})\, $$\n\nwhere $\\tilde{\\alpha}\,\\tilde{\\beta}\,\\
 tilde{\\eta}$ are mean-zero Matérn Gaussian fields approximated by the SP
 DE over a triangular mesh. The database combines INFyS plot AGBD\, CCI Bio
 mass\, and GLAD-GEDI canopy height\, harmonized around 2019\, with covaria
 tes extracted as footprint-weighted means per plot. Because spatial covari
 ance is inferred from the data\, a single high-leverage record can distort
  local coefficient surfaces\, so we applied a quality-control protocol—c
 oordinate and unit verification\, removal of duplicates and implausible va
 lues\, and restriction to mainland forest. The SPDE mesh was confined to Q
 uintana Roo rather than a national grid\, improving computational efficien
 cy.\n\nThe choice of $g$ matters once predictions return to the Mg ha$^{-1
 }$ scale\, motivating our departure from the square-root response used nat
 ionally. Writing $\\mu(\\mathbf{s})$ for the linear predictor\, the square
 -root link inverts by squaring\, whereas the logarithmic link inverts by e
 xponentiation:\n\n$$ g=\\sqrt{\\cdot}: Y=\\mu^2=\\dots+2\,\\beta(\\mathbf{
 s})\\eta(\\mathbf{s})\,x_1x_2+\\dots\, \\qquad g=\\log: Y=e^{\\alpha(\\mat
 hbf{s})}\,e^{\\beta(\\mathbf{s})x_1}\,e^{\\eta(\\mathbf{s})x_2}. $$\n\nSqu
 aring introduces covariate interaction terms\, so the marginal effect of e
 ach covariate depends on the other and on the intercept\, making EO effect
 s entangled. The logarithm instead factorizes biomass into independent mul
 tiplicative terms\, matching its power-law structure and preserving per-co
 variate effects. Zero-valued records were dropped\, as $\\log 0$ is undefi
 ned\; since $\\exp{E[\\log Y]}\\neq E[Y]$\, predictions were back-transfor
 med from posterior predictive samples. Accuracy was assessed with $R^2$\, 
 RMSE\, MAE\, and bias\, calibration with leave-one-out Probability Integra
 l Transform\, and prediction structure through covariate-contribution surf
 aces $C_{\\mathrm{CCI}}(\\mathbf{s})=\\beta(\\mathbf{s})x_1(\\mathbf{s})$ 
 and $C_{\\mathrm{GEDI}}(\\mathbf{s})=\\eta(\\mathbf{s})x_2(\\mathbf{s})$\,
  with relative dominance $D_{\\mathrm{CCI}}=|C_{\\mathrm{CCI}}|/(|C_{\\mat
 hrm{CCI}}|+|C_{\\mathrm{GEDI}}|)$.\n\n**4. Findings**\n\nThe logarithmic f
 ormulation substantially outperformed the square-root one\, returning an $
 R^2$ of 0.68 against 0.38\, RMSE of 26.7 against 46.9 Mg ha$^{-1}$\, MAE o
 f 21.9 against 33.2 Mg ha$^{-1}$\, and bias of 4.2 against 16.8 Mg ha$^{-1
 }$—a 78% gain in explained variance\, a 43% reduction in RMSE\, and a ne
 ar four-fold reduction in relative bias\, from 27.6% to 7.0%. These gains 
 reflect the positive\, right-skewed\, heteroscedastic nature of biomass: a
 ligning the linear predictor with multiplicative mechanisms prevents spati
 ally varying coefficients from absorbing distortions arising from an unsui
 table response scale. Applied locally\, the PIT diagnostic confirms better
 -calibrated behaviour once variance is stabilised and anomalous records ar
 e removed.\n\nBecause the log scale keeps covariate effects separable\, th
 e contribution surfaces admit a clean reading that the squared scale does 
 not. A coefficient map alone is misleading—a steep local slope contribut
 es little where its covariate is near zero—so weighting each slope by th
 e value its covariate takes reveals which EO source governs prediction at 
 a given point. Mapping dominance over Quintana Roo exposes a spatial divis
 ion of labor: zones where CCI Biomass carries the signal\, where canopy he
 ight does\, and where neither does and the latent field absorbs the residu
 al baseline. This matters because similar canopy heights can correspond to
  different biomass densities depending on composition and degradation. Hun
 ka et al. (2024) map spatially varying coefficients but stop there\; carry
 ing the analysis through to contribution and dominance turns a surface of 
 parameters into an operational map of per-sensor reliability—and is cohe
 rent only under a response scale that does not entangle the covariates\n\n
 \n**5. Contributions**\n\nThis study proposes an independent log-scale Spa
 tially Varying Coefficient–Stochastic Partial Differential Equation (SVC
 –SPDE) formulation for tropical biomass that integrates quality control\
 , a state-restricted mesh\, and posterior-sample back-transformation into 
 a unified workflow\, improving performance in high-biomass regimes where E
 arth Observation (EO) covariates saturate. Its main contribution is the sp
 atial decomposition of covariate effects using a logarithmic response that
  preserves separability on the natural scale\, enabling identification of 
 where each EO source governs prediction. The framework also supports poste
 rior predictive aggregation over administrative and conservation units wit
 h explicit uncertainty\, providing a transferable approach for biomass est
 imation.
DTSTAMP:20261122T123849Z
LOCATION:Multifuncional (Remoto)
SUMMARY:A Bayesian SVC–SPDE approach to above-ground biomass estimation i
 n tropical forests: response transformation and covariate dominance in Qui
 ntana Roo\, Mexico - Katia Michelle Villarnobo Gonzalez\, Hugo Carlos-Mart
 ínez\, Jorge Omar López Martínez
URL:https://pretalx.cienciaenmexico.com/igisc-2026/talk/Z3ZF3B/
END:VEVENT
BEGIN:VEVENT
UID:pretalx-igisc-2026-9GDUH9@pretalx.cienciaenmexico.com
DTSTART;TZID=CST:20261113T111500
DTEND;TZID=CST:20261113T113000
DESCRIPTION:Background\nForest fires are a central ecological process in ma
 ny regions of Mexico\, particularly in temperate and montane ecosystems wh
 ere fire has historically shaped vegetation structure and regeneration dyn
 amics. However\, under current conditions of climate change\, land‑use t
 ransformation\, and expanding human pressure\, fire occurrence increasingl
 y leads to high‑risk scenarios for communities\, ecosystems\, and Protec
 ted Natural Areas (ANP). In the central region of the country\, decades of
  fire‑exclusion policies have contributed to fuel accumulation\, increas
 ing the likelihood of severe events. At the same time\, traditional fire
 ‑use practices\, regulatory constraints\, and heterogeneous patterns of 
 land occupation create spatially differentiated ignition conditions. Exist
 ing monitoring systems often rely on static or coarse‑resolution approac
 hes that do not adequately represent the spatial variability of fuels\, mi
 croclimates\, or anthropogenic drivers. These limitations underscore the n
 eed for methodologies that integrate meteorological\, fuels\, anthropogeni
 c\, and topographic predictors within a machine‑learning framework capab
 le of capturing the dynamic conditions that influence fire occurrence prob
 ability. Combining these components enables the construction of hazard map
 s that anticipate critical scenarios and support adaptive fire‑managemen
 t strategies. This work was carried out with the support of the Mexican Fu
 nd for the Conservation of Nature (FMCN) and the Development Bank KfW\, wh
 ose collaboration has strengthened ongoing efforts to advance geospatial t
 ools for fire‑risk assessment and sustainable landscape management.\nObj
 ective \nDevelop and implement a methodology for generating forest fire ha
 zard maps in the Central Region and Trans‑Mexican Volcanic Belt. The aim
  is to explain spatial patterns of fire hazard by evaluating methodologica
 l improvements\, including the integration of multivariate machine‑learn
 ing algorithms\, optimized predictor selection\, and threshold calibration
  for operational decision‑making.\nMaterials and methods\nThe study area
  covers 164\,203 km² across 13 states and 771 municipalities. It includes
  19 federal ANP that represent key conservation areas. The region exhibits
  strong altitudinal gradients\, diverse vegetation types\, and marked clim
 atic variability\, all of which influence fuel availability and ignition p
 atterns. Fire occurrence records from 2012–2024 were obtained from the n
 ational forest agency. A binary presence–absence dataset was constructed
  by filtering events within forested areas and generating pseudo‑absence
 s outside a 1‑km exclusion buffer around each ignition point. This appro
 ach reduces spatial autocorrelation and ensures that absence samples repre
 sent distinct environmental conditions. The final dataset includes annual 
 ignition points and corresponding absence samples\, each associated with v
 egetation type and year of occurrence. The independent variables were orga
 nized into four thematic components. The anthropogenic component included 
 distance to protected areas\, settlements\, and agricultural zones\, as we
 ll as population density derived\, representing human activities that driv
 e most ignitions. The forest‑fuel component incorporated spectral indice
 s from VIIRS imagery—such as GOSAVI\, NBR\, and NDVI—processed into 60
 ‑day composites to reduce atmospheric noise and capture short‑term veg
 etation conditions. The topographic component included slope\, terrain rou
 ghness\, and related metrics derived from the national elevation model and
  resampled to 1 km to ensure spatial consistency. The meteorological compo
 nent used Daymet reanalysis to compute 60‑day composites of vapor pressu
 re deficit\, accumulated precipitation\, accumulated radiation\, and tempe
 rature extremes\, representing atmospheric conditions that influence fuel 
 moisture and ignition likelihood.\nThe dataset was split into training (60
 %)\, validation (20%)\, and external testing (20%) subsets using stratifie
 d sampling to preserve class proportions. Variables were normalized to the
  [0\,1] range when required by the algorithms. A multi‑step procedure wa
 s applied: (i) Univariate AUC analysis to identify predictors with the hig
 hest discriminative capacity\, (ii) Collinearity assessment using Pearson 
 correlation and variance inflation factors\, (iii) Stability verification 
 using Y‑Scrambling to ensure that predictive relationships were not rand
 om\, (iv) Four predictors were selected for the final model: distance to A
 NP (adan)\, terrain roughness (teri)\, 60‑day vapor pressure water (vp)\
 , and spectral index gosavi.\nFive algorithms were evaluated: Random Fores
 t\, XGBoost\, Support Vector Machines with radial kernel\, k‑Nearest Nei
 ghbors\, and logistic regression. Models were trained using stratified 10
 ‑fold cross‑validation. Performance metrics included AUC\, accuracy\, 
 sensitivity\, specificity\, Matthews correlation coefficient\, and F1‑sc
 ore. The two best‑performing models (Random Forest and XGBoost) were opt
 imized using grid search. Final evaluation was conducted on the external t
 est set\, including threshold optimization using the Youden index. The fin
 al model was applied to a multiband raster stack of predictors aligned in 
 resolution and extent. Continuous probabilities were classified into five 
 hazard levels. Spatial residuals were computed to identify areas of over
 ‑ or under‑estimation.\nResults and discussion\nRandom Forest achieved
  the highest performance\, with an AUC of 0.906\, followed by XGBoost with
  0.895. Both models showed consistent behavior across folds\, indicating s
 table generalization. Threshold optimization increased sensitivity from 0.
 720 to 0.807 for Random Forest\, improving detection of ignition events wi
 th a moderate reduction in specificity. The selected predictors captured k
 ey spatial gradients: ignitions were more frequent within 50 km of ANP\, i
 n areas of intermediate terrain roughness\, under low atmospheric moisture
  (vapor pressure near 500 Pa)\, and in vegetation with intermediate values
 \, reflecting moderate biomass and canopy openness. The results highlight 
 the interaction between climatic drivers\, fuel conditions\, and human act
 ivities. The inclusion of distance to ANP captured spatial gradients assoc
 iated with land‑use restrictions and community fire‑use practices. The
  60‑day meteorological composites effectively represented short‑term a
 tmospheric conditions relevant to ignition. Limitations include positional
  uncertainty in fire records\, coarse spatial resolution of Daymet\, and t
 he absence of high‑resolution wind and humidity data. Future work should
  integrate hot-spots\, thermal anomalies\, and alternative climate dataset
 s such as CHIRPS or ERA5‑Land\, as well as medium‑resolution spectral 
 data from Sentinel‑2 or Landsat.\nConclusion\nThe study demonstrates the
  feasibility of integrating machine learning with meteorological\, spectra
 l\, and topographic predictors to generate dynamic fire hazard maps at reg
 ional scale. The methodology is scalable and suitable for operational impl
 ementation. Future efforts should evaluate multiple temporal windows\, inc
 orporate satellite‑derived ignition proxies\, and expand validation acro
 ss additional regions.
DTSTAMP:20261122T123849Z
LOCATION:Video Room
SUMMARY:Dynamic forest fire hazard mapping with artificial intelligence: An
  integrated geospatial framework for the Trans‑Mexican Volcanic Belt - A
 drian Israel Silva Cardoza\, Daniela Moctezuma\, Hugo Carlos-Martínez
URL:https://pretalx.cienciaenmexico.com/igisc-2026/talk/9GDUH9/
END:VEVENT
END:VCALENDAR
