Search NASA⌕ Search

SEARCH · Search NASA

Results for “mean squared error”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 325 records · Page 18

Machine learning and process-based modeling of spatiotemporal changes in active layer thickness across Alaska

Permafrost degradation poses a growing threat to infrastructure stability and ecosystem resilience in the rapidly warming Arctic. We investigated the spatiotemporal dynamics of active layer thickness (ALT) across Alaska by integrating field observations, environmental datasets, a physically based Stefan model, and machine learning (ML) techniques. Using weather projections from the Coupled Model Intercomparison Project Phase 6 under two Shared Socioeconomic Pathways (SSP 2-4.5 and SSP 5-8.5), we assessed ALT sensitivity to projected future weather conditions. The random forest (RF) model outperformed the Stefan approach in predicting ALT on the training dataset (R² = 0.84 vs. 0.53) but demonstrated lower generalizability on the test dataset (R² = 0.24 vs. 0.54). The root mean square error (RMSE) for the RF model for training and testing ranged from 14 to 22 cm, compared to 17 and 18 cm for the Stefan model. Variable importance analysis revealed that mean annual temperature and slope angle were the strongest predictors of ALT, accounting for 19% and 18% of the variance, respectively, followed by sediment transport index (14%) and stream power index (11%). Comparative analysis of baseline ALT predictions showed the Stefan model tended to project a thicker active layer (mean ± SD: 65 ± 16 cm), compared to the RF model (mean ± SD: 59 ± 8.8) cm). Both models indicated a latitudinal gradient in ALT, with shallower depths at higher latitudes. Projected ALT increases by 2100 were estimated at 3.3 ± 2.2 cm under SSP 2-4.5 and 5.9 ± 4.0 cm under SSP 5-8.5 for the ML model, whereas the Stefan model projected substantially larger increases of 13 ± 2.6 cm (SSP 2-4.5) and 28 ± 4.4 cm (SSP5-8.5). Spatial analysis showed the greatest ALT increases in northern Alaska, with relatively smaller changes in southern regions. These findings highlight the complex, multifactorial nature of ALT dynamics and the value of hybrid modeling approaches. As rising temperatures accelerate permafrost thaw, changes in ALT can disrupt ecosystems, damage infrastructures, and enhance the release of stored soil carbon, highlighting the urgent need for improved predictive capabilities to inform adaptation strategies in the Arctic.

Climate sciences↗

Scaling Observation Error for Optimal Assimilation of CCI SST Data into a Regional HYCOM EnOI System

South Africa currently possesses no operational ocean forecasting system for the purpose of predicting ocean state variables including temperature,salinity and velocity. Substantial initial efforts towards this goal have been made and resulted in a system using a regional Hybrid Coordinate Ocean Model (HYCOM) along with the Ensemble Optimal Interpolation (EnOI)assimilation scheme. Assimilating only sea surface temperature (SST) observations from the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) product into the system resulted in a degraded forecast. Aiming to address this, Climate Change Initiative (CCI) SSTs are assimilated into the system in an effort to improve the forecast skill. Observation errors in the assimilated product are used in the EnOI to determine whether more confidence should be placed in the model or observations in producing the analysis, but overconfidence in observations can shock the model and result in failure. To tweak the impact of the assimilation, a scaling factor is applied in the assimilation code. A scaling factor of 25 was found to produce a favourable result with lowest mean root mean square error (RMSE;1.098°C) between the model and observations over time. Postulating the error to be overconfident, a floor value is introduced in order to set a minimum value for the observation error thereby reducing confidence in the observations. These experiments fared less favourably with a floor value of 0.5 and a scaling factor of 15 producing the best mean RMSE (1.118°C).

Luyt, Hermann↗

TH-CD-207A-07: Prediction of High Dimensional State Subject to Respiratory Motion: A Manifold Learning Approach

Purpose: The development of high dimensional imaging systems (e.g. volumetric MRI, CBCT, photogrammetry systems) in image-guided radiotherapy provides important pathways to the ultimate goal of real-time volumetric/surface motion monitoring. This study aims to develop a prediction method for the high dimensional state subject to respiratory motion. Compared to conventional linear dimension reduction based approaches, our method utilizes manifold learning to construct a descriptive feature submanifold, where more efficient and accurate prediction can be performed. Methods: We developed a prediction framework for high-dimensional state subject to respiratory motion. The proposed method performs dimension reduction in a nonlinear setting to permit more descriptive features compared to its linear counterparts (e.g., classic PCA). Specifically, a kernel PCA is used to construct a proper low-dimensional feature manifold, where low-dimensional prediction is performed. A fixed-point iterative pre-image estimation method is applied subsequently to recover the predicted value in the original state space. We evaluated and compared the proposed method with PCA-based method on 200 level-set surfaces reconstructed from surface point clouds captured by the VisionRT system. The prediction accuracy was evaluated with respect to root-mean-squared-error (RMSE) for both 200ms and 600ms lookahead lengths. Results: The proposed method outperformed PCA-based approach with statistically higher prediction accuracy. In one-dimensional feature subspace, our method achieved mean prediction accuracy of 0.86mm and 0.89mm for 200ms and 600ms lookahead lengths respectively, compared to 0.95mm and 1.04mm from PCA-based method. The paired t-tests further demonstrated the statistical significance of the superiority of our method, with p-values of 6.33e-3 and 5.78e-5, respectively. Conclusion: The proposed approach benefits from the descriptiveness of a nonlinear manifold and the prediction reliability in such low dimensional manifold. The fixed-point iterative approach turns out to work well practically for the pre-image recovery. Our approach is particularly suitable to facilitate managing respiratory motion in image-guide radiotherapy. This work is supported in part by NIH grant R01 CA169102-02.

60 APPLIED LIFE SCIENCES↗

Evaluation of Bayesian Sequential Proportion Estimation Using Analyst Labels

The author has identified the following significant results. A total of ten Large Area Crop Inventory Experiment Phase 3 blind sites and analyst-interpreter labels were used in a study to compare proportional estimates obtained by the Bayes sequential procedure with estimates obtained from simple random sampling and from Procedure 1. The analyst error rate using the Bayes technique was shown to be no greater than that for the simple random sampling. Also, the segment proportion estimates produced using this technique had smaller bias and mean squared errors than the estimates produced using either simple random sampling or Procedure 1.

Lennington, R. K.↗

Estimating cell populations

An important step in the diagnosis of a cervical cytology specimen is estimating the proportions of the various cell types present. This is usually done with a cell classifier, the error rates of which can be expressed as a confusion matrix. We show how to use the confusion matrix to obtain an unbiased estimate of the desired proportions. We show that the mean square error of this estimate depends on a 'befuddlement matrix' derived from the confusion matrix, and how this, in turn, leads to a figure of merit for cell classifiers. Finally, we work out the two-class problem in detail and present examples to illustrate the theory.

White, B. S.↗

Surface Turbulent Fluxes Over Pack Ice Inferred from TOVS Observations

A one-dimensional, atmospheric boundary layer model is coupled to a thermodynamic ice model to estimate the surface turbulent fluxes over thick sea ice. The principal forcing parameters in this time-dependent model are the air temperature, humidity, and wind speed at a specified level (either at 2 m or at 850 mb) and the downwelling surface radiative fluxes. The free parameters. are the air temperature, humidity, and wind speed profiles below the specified level, the surface skin temperature, the ice temperature profile, and the surface turbulent fluxes. The goal is to determine how well we can estimate the turbulent surface heat and momentum fluxes using forcing parameters from atmospheric temperatures and radiative fluxes retrieved from the TIROS-N Operational Vertical Sounder (TOVS) data. Meteorological observations from the Lead Experiment (LeadEx, April 1992) ice camp are used to validate turbulent fluxes computed with the surface observations and the results are used to compare with estimates based on radio-sonde observations or with estimates based on TOVS data. We find that the TOVS-based estimates of the stress are significantly more accurate than those found with a constant geostrophic drag coefficient, with a root-mean-square error about half as large. This improvement is due to stratification effects included in the boundary layer model. The errors in the sensible heat flux estimates, however, are large compared to the small mean values observed during the field experiment.

Lindsay, R. W.↗

Mini-TES Observations of Comanche Carbonate and Its Distribution

The discovery by the Spirit rover of outcrops rich in Mg-Fe carbonate [Morris et al., 2010] represents another manifestation of a diverse aqueous history in Gusev crater. In 2005, observations by the Moessbauer spectrometer (MB) on outcrops dubbed Comanche provided initial indication of Fe-Mg carbonate that was subsequently supported by analysis of elemental data from the Alpha Particle X-ray Spectrometer (APXS). The recognition of a carbonate component in thermal infrared spectra measured by the Miniature Thermal Emission Spectrometer (Mini-TES) was significantly delayed due to dust contamination of the instrument's optics. With the implementation of a viable dust correction, the Comanche spectra were revisited and presented clear and compelling evidence for a Mg-Fe carbonate component that could be as much as a third of the total mineral abundance. The data from all three instruments in combination are best matched by Mg-Fe carbonate with an abundance of 16-34 wt%. Mini-TES spectra were acquired for 12 targets at various locations on the Comanche (4-5 m long) and Comanche Spur (1-2 m long) outcrops, the latter being the location of the MB and APXS measurements. The two outcrops are spectrally comparable and share similar morphology and texture based on color images from the Panoramic Camera (Pancam). The highest quality Mini-TES spectrum comes from the larger Comanche outcrop on a target named Saupitty. Linear least squares modeling of the Saupitty spectrum employed a library of laboratory spectra tailored for consistency with the APXS and MB data and included spectra representing Martian dust, a slope spectrum to account for any temperature determination errors, and a blackbody spectrum to account for differences in spectral contrast between the laboratory and Mini-TES spectrum. Successful modeling of the Comanche Saupitty spectrum required one or more carbonate phases to obtain a good fit. Excluding all carbonates from the full starting library more than doubled the root-mean-squared error of the model fit (0.147% vs. 0.299%). Because Mg-Fe carbonate and Ca-Mg carbonate (dolomite) are so spectrally similar over the range used for modeling, both provide a comparable fit. However, Carich carbonates like dolomite are precluded based on APXS data and are inconsistent with MB results. The Comanche carbonate rocks are stratigraphically above a set of olivine-rich volcaniclastic rocks known as Algonquin class that mantle the Haskin Ridge feature of the Columbia Hills. Based on ~50 Mini-TES observations, the Comanche outcrops are the only rocks that host abundant carbonate. However, a target at the base of the larger Comanche outcrop appears spectrally transitional between the carbonate and olivine units. This transitional spectral character applies to additional outcrops a few 10s of meters away from Comanche that also appear stratigraphically transitional. Additional work will attempt to establish whether we are seeing an alteration horizon or depositional unit associated with the emplacement Comanche carbonate.

Ruff, Steven W.↗

SCOPE: Spectral cloud optical property estimation using real-time GOES-R longwave imagery

The output of ground-based, solar power generation systems is strongly dependent on cloud cover, which is the main contributor to solar power variability and uncertainty. Cloud optical properties are typically over-simplified in forecasting applications due to the lack of realtime, accurate estimates. In this work, we introduce a method, the Spectral Cloud Optical Property Estimation (SCOPE), for estimating cloud optical properties directly from high-resolution (5-min, 2 km) imagery from Geostationary Operational Environmental Satellite (GOES)-R, which is the newest generation of the GOES system. The SCOPE method couples a two-stream, spectrally resolved radiative model with the longwave GOES-R sensor output to simultaneously estimate the cloud optical depth, cloud top height, and cloud thickness during both day and night at 5-min intervals. The accuracy of SCOPE is evaluated using one year (2018) of downwelling longwave (DLW) radiation measurements from the Surface Radiation Budget Network, which consists of seven sites spread across climatically diverse regions of the contiguous United States. During daytime clear-sky periods, SCOPE predicts DLW within instrument uncertainty (10 W m -2 ) for four of the seven locations, with the remaining locations yielding errors of the order of 11.2, 17.7, and 20.2 W m -2 . For daytime cloudy-sky, daytime all-sky (clear or cloudy), and nighttime all-sky periods, SCOPE achieves root mean square error values of 23.0–34.5 W m -2 for all seven locations. These results, together with the low-latency of the method (1 s per sample), show that SCOPE provides a viable solution to real-time, accurate estimation of cloud optical properties for both day and night.

42 ENGINEERING↗

Projecting Future Energy Production from Operating Wind Farms in North America. Part I: Dynamical Downscaling

Abstract New simulations at 12-km grid spacing with the Weather and Research Forecasting (WRF) Model nested in the MPI Earth System Model (ESM) are used to quantify possible changes in wind power generation potential as a result of global warming. Annual capacity factors (CF; measures of electrical power production) computed by applying a power curve to hourly wind speeds at wind turbine hub height from this simulation are also used to illustrate the pitfalls in seeking to infer changes in wind power generation directly from low-spatial-resolution and time-averaged ESM output. WRF-derived CF are evaluated using observed daily CF from operating wind farms. The spatial correlation coefficient between modeled and observed mean CF is 0.65, and the root-mean-square error is 5.4 percentage points. Output from the MPI-WRF Model chain also captures some of the seasonal variability and the probability distribution of daily CF at operating wind farms. Projections of mean annual CF (CF A ) indicate no change to 2050 in the southern Great Plains and Northeast. Interannual variability of CF A increases in the Midwest, and CF A declines by up to 2 percentage points in the northern Great Plains. The probability of wind droughts (extended periods with anomalously low production) and wind bonus periods (high production) remains unchanged over most of the eastern United States. The probability of wind bonus periods exhibits some evidence of higher values over the Midwest in the 2040s, whereas the converse is true over the northern Great Plains. Significance Statement Wind energy is playing an increasingly important role in low-carbon-emission electricity generation. It is a “weather dependent” renewable energy source, and thus changes in the global atmosphere may cause changes in regional wind power production (PP) potential. We use PP data from operating wind farms to demonstrate that regional simulations exhibit skill in capturing actual power production. Projections to the middle of this century indicate that over most of North America east of the Rocky Mountains annual expected PP is largely unchanged, as is the probability of extended periods of anomalously high or low production. Any small declines in annual PP are of much smaller magnitude than changes due to technological innovation over the last two decades.

Meteorology & Atmospheric Sciences↗

An Assessment of Receiver Algorithms for Distributed Massive MIMO Systems: Investigating Design Solutions and Performance

This study investigates receiver design solutions for distributed Massive Multiple Input Multiple Output (D-m MIMO) systems, taking into account parameters such as number of access points as well as concerns related to channel estimates that use single-carrier frequency-domain equalization (SC-FDE). A significant contribution of this research is the integration of Low-Density Parity-Check (LDPC) codes to simplify coding complexity and enhance communication efficiency. The research examines different receiver designs, such as spatial antenna correlation and sophisticated channel estimation methods. The authors propose integrating LDPC codes into the receiver architecture to simplify computations and enhance error correction and decoding. Moreover, the paper examines performance evaluation measures and approaches, highlighting the trade-offs among complexity, spectral efficiency, and error performance. The comparative analysis indicates the benefits, in terms of performance, of incorporating LDPC codes and improving system throughput and dependability. We examine four distinct receiver algorithms: zero-forcing (ZF), minimum mean square error (MMSE), maximum ratio combining (MRC), and equal gain combining (EGC). The study shows that MRC and EGC receivers work well in D-m MIMO because they make the receiver system less computationally demanding.

Gashtasbi, Ali↗

Simulations of the monthly mean atmosphere for February 1976 with the GISS model

Three monthly mean simulations of the global atmosphere were computed for February 1976 with the GISS model from observed initial conditions on the first day of the month. In a replication experiment, two of these computations generated slightly different monthly mean states, apparently due to the schedule of interruptions on the computer. The root-mean-square errors of replication over the Northern Hemisphere were found to be about 2 mb, 20 m and 1 K for sea level pressure, 500 mb height and 850 mb temperature, respectively. The monthly mean 500 mb forecast results for February 1976 over the Northern Hemisphere are consistent with those from earlier GISS model experiments and again indicate some predictive skill at that level. Use of the observed monthly mean sea surface temperature (SST) field for February 1976 in place of the climatological SST field for February resulted in slightly improved simulations over the globe and Northern Hemisphere,

Spar, J.↗

Determination of design and operation parameters for upper atmospheric research instrumentation to yield optimum resolution with deconvolution, appendix 3

The Always-Convergent Iterative Noise Removal and Deconvolution Method of Ioup is applied as a single-filter in the transform domain to deconvolution with both narrow and wide Gaussian impulse response functions. The wraparound error for both cases is also studied. A method is developed by which one can find the optimum iteration number for single-filter iterative deconvolution of sampled data. The method employs the mean square error (MSE), the square of the difference between the deconvolved result and the input, for optimization. The MSE decreases as the deconvolution iterations proceed, but at the optimum iteration number, the MSE starts to increase. This procedure is repeated for signal-to-noise ratio of 10 to 150. The optimum iteration number and the MSE are plotted vs SNR. By knowing the SNR for a particular experiment, one can find the optimum iteration number and MSE.

Ioup, George E.↗

TU-H-CAMPUS-IeP1-01: Bias and Computational Efficiency of Variance Reduction Methods for the Monte Carlo Simulation of Imaging Detectors

Purpose: Variance reduction techniques (VRTs) are employed in Monte Carlo simulations to obtain estimates with reduced statistical uncertainty for a given simulation time. In this work, we study the bias and efficiency of a VRT for estimating the response of imaging detectors. Methods: We implemented Directed Sampling (DS), preferentially directing a fraction of emitted optical photons directly towards the detector by altering the isotropic model. The weight of each optical photon is appropriately modified to maintain simulation estimates unbiased. We use a Monte Carlo tool called fastDETECT2 (part of the hybridMANTIS open-source package) for optical transport, modified for VRT. The weight of each photon is calculated as the ratio of original probability (no VRT) and the new probability for a particular direction. For our analysis of bias and efficiency, we use pulse height spectra, point response functions, and Swank factors. We obtain results for a variety of cases including analog (no VRT, isotropic distribution), and DS with 0.2 and 0.8 optical photons directed towards the sensor plane. We used 10,000, 25-keV primaries. Results: The Swank factor for all cases in our simplified model converged fast (within the first 100 primaries) to a stable value of 0.9. The root mean square error per pixel for DS VRT for the point response function between analog and VRT cases was approximately 5e-4. Conclusion: Our preliminary results suggest that DS VRT does not affect the estimate of the mean for the Swank factor. Our findings indicate that it may be possible to design VRTs for imaging detector simulations to increase computational efficiency without introducing bias.

60 APPLIED LIFE SCIENCES↗

Developing an Effective Model for Predicting Spatially and Temporally Continuous Stream Temperatures from Remotely Sensed Land Surface Temperatures

Although water temperature is important to stream biota, it is difficult to collect in a spatially and temporally continuous fashion. We used remotely-sensed Land Surface Temperature (LST) data to estimate mean daily stream temperature for every confluence-to-confluence reach in the John Day River, OR, USA for a ten year period. Models were built at three spatial scales: site-specific, subwatershed, and basin-wide. Model quality was assessed using jackknife and cross-validation. Model metrics for linear regressions of the predicted vs. observed data across all sites and years: site-specific r2 = 0.95, Root Mean Squared Error (RMSE) = 1.25 °C; subwatershed r2 = 0.88, RMSE = 2.02 °C; and basin-wide r2 = 0.87, RMSE = 2.12 °C. Similar analyses were conducted using 2012 eight-day composite LST and eight-day mean stream temperature in five watersheds in the interior Columbia River basin. Mean model metrics across all basins: r2 = 0.91, RMSE = 1.29 °C. Sensitivity analyses indicated accurate basin-wide models can be parameterized using data from as few as four temperature logger sites. This approach generates robust estimates of stream temperature through time for broad spatial regions for which there is only spatially and temporally patchy observational data, and may be useful for managers and researchers interested in stream biota.

McNyset, Kristina↗

Simulations of the monthly mean atmosphere for February 1976 with the GISS model

Monthly mean simulations of the global atmosphere were computed for February 1976 with the GISS model from observed initial conditions. In a replication experiment, two of these computations generated slightly different monthly mean states, apparently due to the schedule of interruptions on the computer. The root-mean-square errors of replication over the Northern Hemisphere were found to be about 2 mb, 20 m, and 1 K for sea-level pressure, 500 mb height, and 850 mb temperature, respectively. The monthly mean 500 mb forecast results for February 1976 over the Northern Hemisphere were consistent with those from earlier GISS model experiments.

Spar, J.↗

LACIE - An application of meteorology for United States and foreign wheat assessment

This paper describes the overall Large Area Crop Inventory Experiment technical approach utilizing the global weather-reporting network and the Landsat satellite to make a quasi-operational application of existing research results, and the accomplishments of this cooperative experiment in utilizing the weather information. Global weather data were utilized in preparing timely yield estimates for selected areas of the U.S. Great Plains, the U.S.S.R. and Canada. Additionally, wheat yield models were developed and pilot tested for Brazil, Australia, India and Argentina. The results of the work show that heading dates for wheat in North America can be predicted with an average absolute error of about 5 days for winter wheat and 4 days for spring wheat. Independent tests of wheat yield models over a 10-year period for the U.S. Great Plains produced a root-mean-square error of 1.12 quintals per hectare (q/ha) while similar tests in the U.S.S.R. produced an error of 1.31 q/ha. Research designed to improve the initial capability is described as is the rationale for further evolution of a capability to monitor global climate and assess its impact on world food supplies.

Hill, J. D.↗

SU-E-J-112: Intensity-Based Pulmonary Image Registration: An Evaluation Study

Purpose: Accurate alignment of thoracic CT images is essential for dose tracking and to safely implement adaptive radiotherapy in lung cancers. At the same time it is challenging given the highly elastic nature of lung tissue deformations. The objective of this study was to assess the performances of three state-of-art intensity-based algorithms in terms of their ability to register thoracic CT images subject to affine, barrel, and sinusoid transformation. Methods: Intensity similarity measures of the evaluated algorithms contained sum-of-squared difference (SSD), local mutual information (LMI), and residual complexity (RC). Five thoracic CT scans obtained from the EMPIRE10 challenge database were included and served as reference images. Each CT dataset was distorted by realistic affine, barrel, and sinusoid transformations. Registration performances of the three algorithms were evaluated for each distortion type in terms of intensity root mean square error (IRMSE) between the reference and registered images in the lung regions. Results: For affine distortions, the three algorithms differed significantly in registration of thoracic images both visually and nominally in terms of IRMSE with a mean of 0.011 for SSD, 0.039 for RC, and 0.026 for LMI (p<0.01; Kruskal-Wallis test). For barrel distortion, the three algorithms showed nominally no significant difference in terms of IRMSE with a mean of 0.026 for SSD, 0.086 for RC, and 0.054 for LMI (p=0.16) . A significant difference was seen for sinusoid distorted thoracic CT data with mean lung IRMSE of 0.039 for SSD, 0.092 for RC, and 0.035 for LMI (p=0.02). Conclusion: Pulmonary deformations might vary to a large extent in nature in a daily clinical setting due to factors ranging from anatomy variations to respiratory motion to image quality. It can be appreciated from the results of the present study that the suitability of application of a particular algorithm for pulmonary image registration is deformation-dependent.

60 APPLIED LIFE SCIENCES↗

Corrigendum to "Predicting biomass of hyperdiverse and structurally complex central Amazonian forests — a virtual approach using extensive field data" Published in Biogeosciences, 13, 1553-1570, 2016

Old-growth forests are subject to substantial changes in structure and species composition due to the intensification of human activities, gradual climate change and extreme weather events. Trees store ca. 90% of the total aboveground biomass (AGB) in tropical forests and precise tree biomass estimation models are crucial for management and conservation. In the central Amazon, predicting AGB at large spatial scales is a challenging task due to the heterogeneity of successional stages, high tree species diversity and inherent variations in tree allometry and architecture. We parameterized generic AGB estimation models applicable across species and a wide range of structural and compositional variation related to species sorting into height layers as well as frequent natural disturbances. We used 727 trees (diameter at breast height ≥5 cm) from 101 genera and at least 135 species harvested in a contiguous forest near Manaus, Brazil. Sampling from this data set we assembled six scenarios designed to span existing gradients in floristic composition and size distribution in order to select models that best predict AGB at the landscape level across successional gradients. We found that good individual tree model fits do not necessarily translate into reliable predictions of AGB at the landscape level. We observed systematic biases ranging from -31% (pantropical) to +39 %, with root-mean-square error (RMSE) values of up to 130-Mg ha -1 (pantropical), when predicting AGB (dry mass) over scenarios using our different models and an available pantropical model. Our first and second best models had both low mean biases (0.8 and 3.9 %, respectively) and RMSE (9.4 and 18.6 Mg ha -1 ) when applied over scenarios. Predicting biomass correctly at the landscape level in hyperdiverse and structurally complex tropical forests, especially allowing good performance at the margins of data availability for model construction/calibration, requires the inclusion of predictors that express inherent variations in species architecture. Furthermore, the model of interest should comprise the floristic composition and size-distribution variability of the target forest, implying that even generic global or pantropical biomass estimation models can lead to strong biases. Reliable biomass assessments for the Amazon basin (i.e., secondary forests) still depend on the collection of allometric data at the local/regional scale and forest inventories including species-specific attributes, which are often unavailable or estimated imprecisely in most regions.

54 ENVIRONMENTAL SCIENCES↗