Multi-model ensembles are one of the most common ways to deal with epistemic uncertainty in hydrology. This is a problem because there is no known way to sample models such that the resulting ensemble admits a measure that has any systematic (i.e., asymptotic, bounded, or consistent) relationship with uncertainty. Multi-model ensembles are effectively sensitivity analyses and cannot – even partially – quantify uncertainty. One consequence of this is that multi-model approaches cannot support a consistent scientific method – in particular, multi-model approaches yield unbounded errors in inference. In contrast, information theory supports a coherent hypothesis test that is robust to (i.e., bounded under) arbitrary epistemic uncertainty. This paper may be understood as advocating a procedure for hypothesis testing that does not require quantifying uncertainty, but is coherent and reliable (i.e., bounded) in the presence of arbitrary (unknown and unknowable) uncertainty. We conclude by offering some suggestions about how this proposed philosophy of science suggests new ways to conceptualize and construct simulation models of complex, dynamical systems.
Hydrological modeling outputs are subject to uncertainty resulting from different sources of errors (e.g., error in input data, model structure, and model parameters), making quantification of uncertainty in hydrological modeling imperative and meant to improve reliability of modeling results. The uncertainty analysis must solve difficulties in calibration of hydrological models, which further increase in areas with data scarcity. The purpose of this study is to apply four uncertainty analysis algorithms to a semi-distributed hydrological model, quantifying different source of uncertainties (especially parameter uncertainty) and evaluate their performance. In this study, the Soil and Water Assessment Tools (SWAT) eco-hydrological model was implemented for the watershed in the center of Vietnam. The sensitivity of parameters was analyzed, and the model was calibrated. The uncertainty analysis for the hydrological model was conducted based on four algorithms: Generalized Likelihood Uncertainty Estimation (GLUE), Sequential Uncertainty Fitting (SUFI), Parameter Solution method (ParaSol) and Particle Swarm Optimization (PSO). The performance of the algorithms was compared using P-factor and R-factor, coefficient of determination (R2), the Nash Sutcliffe coefficient of efficiency (NSE) and Percent Bias (PBIAS). The results showed the high performance of SUFI and PSO with P-factor>0.83, R-factor<0.56 and R2>0.91, NSE>0.89, and 0.18<PBIAS<0.32. Hence, we would suggest to use SUFI-2 initially to set the parameter ranges, and further use PSO for final analysis. Indeed, the uncertainty analysis must be accounted when the outcomes of the model use for policy or management decisions.
In this study, we investigate the uncertainties associated with land surface processes in an ensemble predication context. Specifically, we compare the uncertainties produced by a coupled atmosphere–land modeling system with two different land surface models, the Noah-MP land surface model (LSM) and the Noah LSM, by using the Maximum Likelihood Ensemble Filter (MLEF) data assimilation system as a platform for ensemble prediction. We carried out 24-hour prediction simulations in Siberia with 32 ensemble members beginning at 00:00 UTC on 5 March 2013. We then compared the model prediction uncertainty of snow depth and solid precipitation with observation-based research products and evaluated the standard deviation of the ensemble spread. The prediction skill and ensemble spread exhibited high positive correlation for both LSMs, indicating a realistic uncertainty estimation. The inclusion of a multiple snow-layer model in the Noah-MP LSM was beneficial for reducing the uncertainties of snow depth and snow depth change compared to the Noah LSM, but the uncertainty in daily solid precipitation showed minimal difference between the two LSMs. The impact of LSM choice in reducing temperature uncertainty was limited to surface layers of the atmosphere. In summary, we found that the more sophisticated Noah-MP LSM reduces uncertainties associated with land surface processes compared to the Noah LSM. Thus, using prediction models with improved skill implies improved predictability and greater certainty of prediction.
Snow depth is the easiest snowpack property to measure in the field and is used to estimate the distribution of snow for quantifying snow storage. Often the mean of three snow depth measurements is used to represent snow depth at a location. This location is used as a proxy for an area, typically a digital elevation model (DEM) or remotely sensed pixel. Here, 11, 17, or 21 snow depth measurements were used to represent the mean snow depth of a 30-m DEM pixel. Using the center snow depth measurement for each sampling set was not adequate to represent the pixel mean, and while the use of three snow depth measurements improved the estimate of mean, there is still large error for some pixels. These measurements were then used to determine the variability of snow depth across a pixel. Estimating variability from few points rather than all in a measurement was not sufficient. The sampling size was increased from one to the total per pixel (11, 17, or 21) to determine how many point samples were necessary to approximate the mean snow depth per pixel within five percent. Binary regression trees were constructed to determine which terrain and canopy variables dictated the spatial distribution of the snow depth, the standard deviation of snow depth, and the sample size to within 5% of the mean per pixel. One location was measured in two years just prior to peak accumulation, and it is shown that there was little to no inter-annual consistency in the mean or standard deviation.
Spring snowmelt serves as the major hydrological contribution to many watersheds of the US West. Since the 1970s the conterminous western USA has seen an earlier arrival of spring snowmelt. The extremely low snowpack and early melt of 2015 in the Cascade Mountains may be a harbinger of winters to come, underscoring the interest in advancements in spring snowmelt monitoring. Target-of-opportunity and point measurements of snowmelt using meteorological stations or stream gauges are common sources of these data, however, there have been few attempts to identify snowmelt timing using remote sensing. In this study, we describe the creation of snowmelt timing maps (STMs) which identify the day of year that each pixel of a remotely sensed image transitions from “snow-covered” to “no snow” during the spring melt season, controlling for cloud coverage and ephemeral spring snow storms. Derived from the 500 m MODerate-resolution Imaging Spectroradiometer (MODIS) standard snow map, MOD10A2, this new dataset provides annual maps of snowmelt timing, with corresponding maps of cloud interference and interannual variability in snow coverage from 2001–2015. We first show that the STMs agree strongly with in-situ snow telemetry (SNOTEL) meteorological station measurements in terms of snowmelt timing. We then use the STMs to investigate the early snowmelt event of 2015 in the Cascade Mountains, USA, highlighting the protected areas of Mt. Rainier, Crater Lake, and Lassen Volcanic National Parks. In 2015 the Cascade Mountains experienced snowmelt 41 days earlier than the 2001–2015 average, with 25% of its land area melting>65 days earlier than average. The upper elevations of the Cascade Mountains experienced the greatest snowmelt anomaly. Our results are relevant to land managers and biologists as they plan adaptation strategies for mitigating the effects of climate change throughout temperate mountains.
Snow fraction is an important component of land surface models and hydrologic models. Information on snow fraction also improves downstream products retrieved from remote sensing: vertical atmosphere profiles, soil moisture, heat fluxes, etc. The uncertainty of the fractional snow cover estimates must be determined, quantified, and reported to consider the suitability of the product for modeling, data assimilation and other applications. The reflectances of snow and non-snow are characterized by a very significant local variability and also by changes from one scene to another. The local snow and non-snow endmembers are approximated by the Normalized Difference Snow Index with a high accuracy. The magnitudes of snow and non-snow Normalized Difference Snow Indexes are scene-specific and calculated on the fly to retrieve snow fraction. The development of the Normalized Difference Snow Index based algorithms to estimate snow fraction including a scene-specific approach taking local snow and non-snow properties into account is considered an optimal way to fractional snow retrieval from moderate resolution optical remote sensing observations. The Landsat reference data are used to estimate the performance of the fractional snow cover algorithms at moderate resolution and to compare the quality of alternative algorithms. The validation results demonstrate that the performance of the algorithms using Normalized Difference Snow Index has advantages. The advantages achieved in snow fraction retrieval lead to improved estimate of snow water equivalent and changes in snow cover state contributing to better modeling of land surface and hydrologic regime. The success of managing water resources on the whole depends on coordinating described investigations with the works of other researchers developing further enhanced models.
The bulk aerodynamic flux equation is often used to estimate snowpack sublimation since it requires meteorological measurements at only one height above the snow surface. However, to date the uncertainty of these estimates and the individual input variables and input parameters uncertainty have not been quantified. We modeled sublimation for three (average snowpack in 2005, deep snowpack in 2011, and shallow snowpack in 2012) different water years (October 1 to September 30) at West Glacier Lake watershed within the Glacier Lakes Ecosystem Experiments Site in Wyoming. We performed a Monte Carlo analysis to evaluate the sensitivity of modeled sublimation to uncertainties of the input variables and parameters from the bulk aerodynamic flux equation. Input variable time series were uniformly adjusted by a normally distributed random variable with a standard deviation given as follows: 1) the manufacturer’s stated instrument accuracy of 0.3°C for temperature (T), 0.3 m/s for wind speed (Uz), 2% for relative humidity (RH), and 1 mb for pressure (P); 2) 0.0093 m for the aerodynamic roughness length (z0) based on z0 profiles calculations from multiple heights; and 3) 0.08 m for measurement height (z). Often z is held constant; here we used a constant z compared to the ground surface, and subsequently altered z to account for the change in snow depth (ds). The most important source of uncertainty was z0, then RH. Accounting for measurement height as it changed due to snowpack accumulation/ablation was also relevant for deeper snow. Snow surface sublimation uncertainties, from this study, are in the range of 1% to 29% for individual input parameter perturbations. The mean cumulative uncertainty was 41% for the three water years with 55%, 37%, and 32% occurring for the wet, average, and low water years. The top three variables (z varying with ds, z0, and RH) accounted for 74% to 84% of the cumulative sublimation uncertainty.
Gross primary productivity (GPP) is very important in the global carbon cycle. Currently, the newly released estimates of 8-day GPP at 500 m spatial resolution (Collection 6) are provided by the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Science Team for the global land surface via the improved light use efficiency (LUE) model. However, few studies have evaluated its performance. In this study, the MODIS GPP products (GPPMOD) were compared with the observed GPP (GPPEC) values from site-level eddy covariance measurements over seven maize flux sites in different areas around the world. The results indicate that the annual GPPMOD was underestimated by 6%?58% across sites. Nevertheless, after incorporating the parameters of the calibrated LUE, the measurements of meteorological variables and the reconstructed Fractional Photosynthetic Active Radiation (FPAR) into the GPPMOD algorithm in steps, the accuracies of GPPMOD estimates were improved greatly, albeit to varying degrees. The differences between the GPPMOD and the GPPEC were primarily due to the magnitude of LUE and FPAR. The underestimate of maize cropland LUE was a widespread problem which exerted the largest impact on the GPPMOD algorithm. In American and European sites, the performance of the FPAR exhibited distinct differences in capturing vegetation GPP during the growing season due to the canopy heterogeneity. In addition, at the DE-Kli site, the GPPMOD abruptly produced extreme low values during the growing season because of the contaminated FPAR from a continuous rainy season. After correcting the noise of the FPAR, the accuracy of the GPPMOD was improved by approximately 14%. Therefore, it is crucial to further improve the accuracy of global GPPMOD, especially for the maize crop ecosystem, to maintain food security and better understand global carbon cycle.
Reservoir construction within a basin affects the process of water and sediment transport downstream of the dam. The Three Gorges Reservoir (TGR) affects the sediment transport downstream of the dam. The impoundment of the TGR reduced total downstream sediment. The sediment group d≤0.125 mm (fine particle) increased along the path, but the average was still below what existed before the reservoir impoundment. The sediments group d>0.125 mm (coarse particle) was recharged in the Yichang to Jianli reach, but showed a deposition trend downstream of Jianli. The coarse sediment in the Yichang to Jianli section in 2003 to 2007 was above the value before the TGR impoundment. However, the increase of both coarse and fine sediments in 2008 to 2014 was less than that in 2003 to 2007. The sediment retained in the dam is the major reason for the sediment reduction downstream. However, the retention in different river reaches is affected by riverbed coarsening, discharge, flow process, and conditions of lake functioning and recharging from the tributaries. The main conclusions derived from our study are as follows: 1) The riverbed in the Yichang to Shashi section was relatively coarse, thereby limiting the supply of fine and coarse sediments. The fine sediment supply was mainly controlled by TGR discharge, whereas the coarse sediment supply was controlled by the duration of high flow and its magnitude. 2) The supply of both coarse and fine sediments in the Shashi to Jianli section was controlled by the amount of total discharge. The sediment supply from the riverbed was higher in flood years than that in the dry years. The coarse sediment tended to deposit, and the deposition in the dry years was larger than that in the flood years. 3) The feeding of the fine sediment in the Luoshan to Hankou section was mainly from the riverbed. The supply in 2008 to 2014 was more than that in 2003 to 2007. Around 2010, the coarse sediments transited from depositing to scouring that was probably caused by the increased duration of high flow days. 4) Fine sediments appeared to be deposited in large amounts in the Hankou to Jiujiang section. The coarse sediment was fed by the riverbed scouring, and much more coarse sediments were recharged from the riverbed in the flood years than in the dry years. 5) In the Jiujiang to Datong section, the ratio of fine sediments from the Poyang Lake and that from the riverbed was 1: 2.82. The sediment from the riverbed scouring contributed more to the coarse sediment transportation. The contribution was mainly affected by the input by magnitude and duration of high flows.
We integrated a photosynthetic sub-model into the daily Century model (DayCent) to improve the estimations of seasonal changes in carbon fluxes at the Niwot Ridge LTER site and the Harvard forest LTER site (DayCent-Photo). The photosynthetic sub-model, adapted from the SIPNET/PnET family of models, includes solar radiation and vapor pressure deficit controls on production, as well as temperature and water stress terms. A key feature we added to the base photosynthetic equations is the addition of a variable maximum net photosynthetic rate (Amax). We optimized the parameters controlling photosynthesis using a variation of the Metropolis-Hastings algorithm along with data-assimilation techniques. The model was optimized and validated against observed net ecosystem exchange (NEE) and estimated gross primary production (GPP) and ecosystem respiration (RESP) values for AmeriFlux sites at Niwot Ridge and Harvard forest. The inclusion of a variable Amax rate greatly improved model performance (NEE RMSE= 0.63 gC·m−2, AIC= 2099) versus a version with a single Amax parameter (NEE RMSE= 0.74 gC·m−2, AIC= 3724). DayCent-Photo was able to capture the inter-annual and seasonal flux patterns for NEE, GPP, ecosystem respiration (RESP), and daily actual evapotranspiration (AET), but tended to overestimate yearly NEE uptake. The DayCent-Photo model has been successfully set up to simulate daily NEE, GPP, RESP, and AET for deciduous forest, conifer forests, and grassland systems in the US using AmeriFlux data sets and has recently been improved to include the impact of UV radiation surface litter decay (DayCent-UV). The simulated influence of a variable Amax rate suggests a need for further studies on the process controls affecting the seasonal photosynthetic rates. The results for all of the forest and grassland sites show that maximum Amax values occurs early during the growing period and taper off toward the end of the growing season.
Automatic multi-sensor image registration is a challenging task in remote sensing. Conventional image registration algorithms may not be applicable when common underlying visual features are not distinct. In this paper, we propose a novel image registration approach that integrates local self-similarity (LSS) and mutual information (MI) for multi-sensor images with rigid/non-rigid radiometric and geometric distortions. LSS is a well-performing descriptor that captures common, local internal layout features for multi-sensor images, whereas MI focuses on global intensity relationships. First, potential control points are identified by using the Harris algorithm and screened based on the self-similarity of their local surrounding internal layouts. Second, a Bayesian probabilistic model for matching the ensemble of the LSS features is introduced. Third, a particle swarm optimization (PSO) algorithm is adopted to optimize the point and region correspondences for maximum self-similarity and MI and, ultimately, a robust mapping function. The proposed approach is compared with several conventional image registration algorithms that are based on the sum of squared differences (SSD), scale-invariant feature transforms (SIFT), and speeded-up robust features (SURF) through the experimental registration of pairs of Landsat TM, SPOT, and RADARSAT SAR images. The results demonstrate that the proposed approach is efficient and accurate.
Urbanization is an irreversible trend worldwide, especially in rapidly developing China. Accelerated urbanization has resulted in rapid urban sprawl and urban landscape pattern changes. Quantifying the spatiotemporal dynamics of urban land use and landscape pattern not only can reveal the characteristics of social transfer and economic development, but also can provide insights into the driving mechanisms of land use changes. In this study, we integrated remote sensing (RS), geographic information system (GIS), landscape metrics, and gradient analysis to quantitatively compare the spatiotemporal dynamics of land use, urban sprawl, and landscape pattern for nine cities in the Pearl River Delta from 1985?2000. For the whole study region, urbanization was obvious. The results show an increase in urban buildup land and shrinkage of cropland in the Pearl River Delta. However, the nine cities differed greatly in terms of the process and magnitude of urban sprawl for both the spatial and temporal dimensions. This was most evident for the cities of Guangzhou and Shenzhen. Gradient analysis on urban landscape changes could deepen understanding of the stages of urban development and provide a scientific foundation for future urban planning and land management strategies in China.
The Central Asian Orogenic Belt (CAOB) is a giant suture zone produced by the reduction of the Paleo-Asian Ocean between the Siberian, North China, and Tarim cratons. The CAOB formed in three main stages, including continental accretion, late collision, and intracontinental orogeny. Strong crust-mantle interaction also occurred during these stages. The eastern China-Mongolia frontier area is an important part of the CAOB. Since the Caledonian period, this region has frequently experienced polycyclic tectonic reformation and intense magmatic activities. All of these geological activities lay the foundation for abundant metal resources. In the past, several large metal ore deposits have been founded there. However, the deep structure of different tectonic units and the fault distribution in the eastern China-Mongolia border frontier area are still not clear owing to the complex geological environment. Existing works in that region are insufficient for an in-depth understanding of the metallogenic deposits. The work discussed in this paper was carried out in the eastern China-Mongolia frontier area with measured gravity data along a profile and gravity data extracted from the WGM2012 earth’s gravity model for detailed fault distribution. In this study, empirical mode decomposition (EMD) and tilt angle analysis (TDR) were utilized for processing the gridded gravity data. The measured gravity data were inverted with a 2D inversion algorithm for density distribution along the survey line. The inversion result was used to verify the existence of known faults and describe their deeper extensions. Meanwhile, new faults were also identified along the survey line and then marked on the gridded data to get their horizontal distribution. These results provide significant information for the in-depth understanding of the tectonic units in the study area.
Research on how terrestrial ecosystems respond to climate change can reveal the complex interactions between vegetation and climate. net primary productivity (NPP), an important vegetation parameter and ecological indicator, fluctuates within any given ecological environment or regional carbon budget. In this study, spatial interpolation was used to generate a spatial dataset, with 1-km spatial resolution, with meteorological data from 736 observation stations across China. An improved CASA model was used to simulate NPP over the period of 2001–2013 by taking into account land-cover change in every year during the same period. We propose the grid-based spatial patterns and dynamics of annual NPP, annual average temperature, and annual total precipitation based on the model. We also used the model to demonstrate the spatial correlation between NPP, temperature, and precipitation in the study area with special focus on the impact of climate change in the early 21st century. Results showed that the spatial pattern of NPP over all of China is characterized by higher values in the southeast and lower values in the northwest. The spatial pattern of temperature indicates substantial latitudinal differences across the country, and the spatial pattern of precipitation shows a ribbon of decline from the southeast coast to the northwest inland. Most areas show an upward trend in NPP. Temperatures appear to decrease across the country during the global warming hiatus (1998–2008), and are accompanied by an increase in precipitation over most regions. The correlation between NPP and annual average temperature is weak. Alternatively, NPP and annual total precipitation are positively correlated in northern and central China at a coefficient above 0.64 (p<0.01) yet negatively correlated in the eastern parts of the Qinghai-Tibet Plateau and Sichuan Basin. Results can provide useful information for improving parameters for calibration of the terrestrial ecosystem process model.
In the context of climate warming and urbanization, predictions and inundation simulations for future extreme precipitation have become highly active research topics. In this paper, using daily precipitation recorded at 10 meteorological stations in Shanghai for the period 1961–2010, the daily precipitation of each station during the period 2011–2099 was simulated by the statistical downscaling model (SDSM). And we examined the varying tendencies of future precipitation by the Mann-Kendall test. Further, the Soil Conservation Service (SCS) model and Pearson-III distribution curve were used to simulate the waterlogging duration and depth of future extreme precipitation in different scenarios with 3-, 5-, 10-, 20-, 50-, and 100-year return periods. The results show that: 1) Precipitation in Shanghai before the 2050s shows a trend of increasing and decreasing alternations, followed by a trend of decreasing and a marked decrease in about the 2070s. 2) In the 21st century, the waterlogging duration with return periods of 3, 5, and 10 years in Shanghai is predicted to last for less than 30 minutes, while the return periods of 20, 50, and 100 years last for less than 45 minutes. From the spatial distribution, the waterlogging duration to the east and south of the Huangpu River is predicted to be shorter than that of the west and north. 3) With the increase of the return periods, the depth of waterlogging is predicted to increase. The deepest inundated areas are Jinshan to the south-west of Shanghai, the east side of the Huangpu River, and Chongming Island.
The objective of this study is to evaluate satellite precipitation extremes of the Tropical Rainfall Measuring Mission (TRMM) 3B42 Version 7 product over China during the period of 2009–2013. Eight extreme indices are used to characterize precipitation extremes: monthly maximum 1-day precipitation (RX1day), monthly maximum consecutive 2-day precipitation (RX2day), monthly maximum 5-day consecutive precipitation (RX5day), simple daily intensity index (SDII), annual total precipitation amount for the wet days (PRCPTOT), annual wet days (R1), consecutive dry days (CDD), and consecutive wet days (CWD). The precipitation amount for indices RX1day, RX2day, RX5day, and PRCPTOT is well captured by TRMM 3B42-V7, as verified by lower mean relative bias and normalized root mean square error and the high spatial correlation coefficient. In contrast, the performance of TRMM 3B42-V7 in depicting the indices on intensity and duration (i.e., SDII, R1, CDD, and CWD) is not as good as its performance in depicting the precipitation amount indices. TRMM 3B42-V7 can reproduce extreme indices better in eastern China than in western China, and better in summer than in winter. Probability density function is also calculated better for RX1day, RX2day, RX5day, and PRCPTOT than for SDII, R1, CDD, and CWD. Investigation on the monthly time series of RX1day, RX2day, and RX5day at different spatial scales indicates that TRMM 3B42-V7 performs better at the large spatial scale than at the grid cell scale. Caution should be observed when the satellite-based extreme indices are used.
Proxies based on glycerol dialkyl glycerol tetraethers (GDGTs), a suite of membrane lipids occurring ubiquitously in soils, have generated increasing interest in quantitative paleo-environmental reconstruction. Hot and dry climates are likely to have occurred in the geological past; however, the limitations and applicability of these proxies to hot and dry environments are still unknown. In this study, we analyzed the GDGT distribution in the Turpan (TRP) basin of China, where the highest soil temperature can be approximately 70°C, and the mean annual precipitation (MAP) is 15.3 mm. We compared GDGT-based proxies among TRP soils, Nanyang (NY) soils, and Kunming (KM) soils; these three sites exhibit similar mean annual air temperature (MAAT) albeit contrasting temperature seasonality and MAP. Archaeal isoprenoidal GDGTs (isoGDGTs) dominate over bacterial branched GDGTs (brGDGTs) in most TRP soils; this is a characteristic GDGT distribution pattern for soils from dry regions globally. Another feature is the anomalously high GDGT-0/crenarchaeol ratio, which is generally attributed to the contribution of anaerobic methanogenic archaea by previous studies; however, these anaerobic archaea are unlikely to be highly abundant in the dry TRP soils, indicating that certain uncultured halophilic Euryarchaeota are likely to produce a significant amount of GDGT-0 that finally results in a high GDGT-0/Cren ratio. The changes in the salinity of the TRP soils appear to be an important factor affecting the MBT’5ME and the relative abundance of 6- vs. 5-methyl pentamethylated brGDGTs (IRIIa’). This is likely to introduce certain scatters in the correlations between MBT’5ME and MAAT and that between IRIIa’ and pH determined at the global scale. A comparison of the MBT’5ME-inferred temperature between TRP, NY, and KM soils does not indicate a significant bias toward summer temperature, indicating that brGDGT paleo-thermometers in soils could reflect the MAAT.