Convenience Spatial Sampling Strategy For LAI Empirical Model Construction

Leaf Area Index(LAI)is a key biophysical variable influencing land surface photosynthesis, transpiration, and energy balance. The simplest and most practical way to retrieve LAI is to construct LAI empirical model based on remote sensing(RS) data, which needs strong support from the spatial sampling of LAI. However, existing spatial sampling methods are often high cost, or can hardly meet the accuracy requirements of RS models. A spatial sampling method for LAI empirical model construction is really desperately needed.Geared toward the above problem, this paper proposes a convenience spatial sampling strategy for LAI empirical model construction which combines the advantages of convenience sampling and stratified sampling. RMEI and sampling time are chosen as proper indicators for evaluating its accuracy and sampling cost. At the same time, the impact of several factors on the LAI modeling accuracy are analyzed and discussed in this paper, such as the width of sampling area, sample size, the number of stratum, the correlation between LAI and NDVI, and the way of allocating sample size to each stratum. We can get the following conclusions:(1) A convenience spatial sampling strategy for LAI empirical model construction is proposed. And its basic idea is that on the one hand, the convenience sampling is carried out by settling sampling points within road-centered sampling area with a certain width; on the other hand, when the frequency distribution of the NDVI in the sampling area is consistent with that in the survey area, as the consistency of the frequency distribution of LAI and NDVI, the stratified sampling can ensures the representativeness of the samples by using NDVI as ancillary variable.(2) The LAI convenience spatial sampling strategy consists of four steps: establishment of sampling areas, design of stratified sampling, selection of sampling units, arrangement of sampling points.(3) The empirical model has higher modeling accuracy and stability, which constructed based on the sample data collected by the convenience spatial sampling for LAI empirical model construction. Compared with the random sampling in sampling area, the mean and standard deviation of RMEI of LAI convenience spatial sampling are much smaller, the average RMEI reduces 19.37%, the standard deviation of RMEI reduces 25.67%. Because the ground sampling data from the stratified sampling based on NDVI data is distributed evenly in the range of LAI.(4) LAI convenience spatial sampling can greatly lowers the sampling cost. Compared with stratified sampling in the whole study area, this sampling method can reduces the sampling time by 22.9 hours on average, and reduces the sampling cost by 66.3% on average. Because sampling in the road-centered sampling area can greatly shortens the investigators’ walking distance in the field and effectively reduces the sampling cost.(5) The width of sampling area, sample size, the number of stratum, the correlation between LAI and NDVI, and the way of allocating sample size to each stratum all have great influence on the result of LAI convenience spatial sampling. ① With decrease of sampling area width, the improvement of modeling accuracy is not significant, but the stability of empricial modeling is increased and the sampling cost is declined. ② The accuracy and stability of LAI empirical model are significantly increased with the augment of sample size and the number of stratum. However, with increase of sampling size, although the modeling accuracy is improved, the sampling cost is increased dramatically. ③ Compared with the stratified sampling with proportional allocation, the LAI empirical model established by the stratified sampling with equal allocation holds higher accuracy and better stability. ④ The correlation between NDVI and LAI decides the efficiency and veracity of sampling results. When the correlation is low(correlation coefficient less than 0.6), the LAI convenience spatial sampling is unserviceable. And when the correlation is significant(correlation coefficient greater than 0.6), the LAI convenience spatial sampling method is applicable.